[This Transcript is Unedited]

Department of Health and Human Services

National Committee on Vital and Health Statistics (NCVHS)

Hearing on
Claims-Based Databases for Policy Development and Evaluation: Overview and Emerging Issues

June 17, 2016

Capital Hilton Hotel
Federal Room A
1001 16 St., SW
Washington, D.C.

TABLE OF CONTENTS

P R O C E E D I N G S       (8:04 a.m.)

Agenda Item: Welcome

DR. SUAREZ: Good morning, everyone, to the fourth day of NCVHS meetings.  I want to welcome you all here and on the phone to our hearing on claims-based databases for policy development and evaluation.  This is really a very exciting time for the National Committee addressing this issue. 

We have a very packed agenda.  We very much appreciate everyone’s work and adjustments on schedules to participate in the hearing, the testifier that came in person, as well as those that submitted testimony in writing which we will be acknowledging for the record.  As always, we start with the National Committee’s introductions.  I would ask every member of the National Committee to introduce themselves.  State your name, organization, membership in the National Committee and whether you have any conflict of interest on the topic.

I will start myself.  My name is Walter Suarez.  I am with Kaiser Permanente.  I am the chair of the National Committee and member of the various subcommittees and workgroups.  I don’t have any conflict.

MS. LOVE:  Denise Love, National Association of Health Data Organizations and co-leader of the APCD Council, which is learning collaborative.  I am a member of the full committee.  I am a member of the standards subcommittee and population health subcommittee, no conflicts.

MS. GOSS:  Good morning.  I am Alix Goss.  I am a member of the full committee, co-chair of the review committee, the standard subcommittee, and I have no conflicts.

MS. KLOSS:  Linda Kloss, health information management consultant, member of the full committee, co-chair of privacy, confidentiality and security subcommittee, member of the standards committee and no conflicts.

DR. RIPPEN:  Good morning.  Helga Rippen, I am on the full committee and the subcommittee on privacy and population health, and the workgroup on data.  I have no conflicts.

DR. COHEN:  Good morning.  Bruce Cohen from Massachusetts, I am a member of the full committee, co-chair of the population health subcommittee, member of the data workgroup, no conflicts.

DR. O’GRADY:  Michael O’Grady, I am a member of the full committee and in the population health subcommittee.  I am with O’Grady Health and NORC at the University of Chicago.

DR. PHILLIPS:  Good morning.  I am Bob Phillips from the American Board of Family Medicine, a member of the full committee and the subcommittees on population health and privacy, no conflicts.

DR. SUAREZ: Do we have any members on the phone?

DR. STEAD:  Bill Stead, Vanderbilt University, member of the full committee, co-chair of pop health, member of the review committee, no conflicts.

DR. MAYS:  Vicki Mays, University of California Los Angeles, member of the full committee pop and privacy.

DR. SUAREZ: Thank you.  Let’s see, just a quick note.  Anyone else on the phone, any member of the staff on the phone or any member of our agencies on the phone?

MR. LINCOLN: Mike Lincoln from VA.

MS. HINES:  Good morning, Rebecca Hines, executive secretary to the committee, National Center for Health Statistics.

MS. JACKSON: Debbie Jackson, National Center for Health Statistics, CDC committee staff.

DR. JONES:  Katherine Jones, staff to the committee.

(Intro audience)

DR. SUAREZ: Anyone else?  We will have the testifiers introduce themselves in a minute.  We appreciate your patience.  We are going to go ahead and start out.  I should mention for the members of the committee that we don’t have your printed packet of testimony.  Unfortunately, Marietta had a personal health family issue and couldn’t attend yesterday or is not attending today.  We won’t have the printed copies.  But we do have your email from yesterday that includes the agenda.  Then materials have been sent to you via email, as well.  We will have the opportunity to at least see that if you brought them.  If not, you can access them via email.

Some of the speakers are bringing their testimony, printed copies, so we will be passing those along as they make them available to us here.  Any other announcements or messages?  Okay.  We will go ahead and start. 

We are going to make a few introductory remarks.  I will start, and then I will pass it to Denise.  I think as I mentioned, this is a very exciting time for the National Committee.  We have been wanting to address the topic of data.  We are, after all, the oldest and advisory to the secretary on health data policy.  Certainly, data and data sources have been at the heart of our deliberations from all different perspectives.

As most of you know, we focus our attention on three, four or five major issues.  One is the standards and electronic and coded standards for submitting and exchanging data.  We focus on population health, and sort of at the other end of the spectrum, looking at how data can be used at different levels in the country for analysis and benefit of individuals, communities, population groups and certainly policymakers.

We also run all the things around privacy and security.  We look at the implications and policies related to protecting the health information that is being collected, used, exchanged.  We also look at how to make federal data resources available, accessible and usable.  Those are sort of our core areas. 

Across the board, the topic of today really touches on all of those.  We are fortunate to have representatives from our entire committee.  This is a national committee-wide hearing led by the full committee.  I am very pleased that we have really the perspectives of most of our members here present.

For all of those of us who have worked in data collection and data use for many years, we all can go back to the date when we didn’t have too much data.  We didn’t have too much data electronically either.  We started to think about how to gather data at different levels from different sources.  For many of us, we started back in the ‘80s.  This is the 1980s and even a little before that with collecting data from encounters and from the experience that consumers were having, that they were entering contact with health care systems.

Some of the initial databases were really about discharges from hospitals to understand really how was the hospital experience going, what was the utilization of hospital services, the availability, the accessibility of those services of different levels.  States started to really collect a lot of that data.

As we begin to advance in the ‘90s into more of how do we really expand that to include the ambulatory experience and other experiences, it challenges to poise to what kind of sources can we use.  It is in a state with 150 hospitals, for example.  It was released to go out to 150 hospitals and say send me every quarter your discharge dataset electronically using a standard.  As we move into electronic standards or the HIPAA, we started to see that opportunity to really automate and increase the quality and the frequency and availability of this data comprehensive to other data.

Then when you start thinking about the experience in ambulatory settings, then you jump to a scale issue from 170 points of submission to thousands in many cases or tens of thousands of points to try to capture that experience.  The idea came about to collect the data really from the other sources, which include health plans, and begin to capture the data at the health plan level, which certainly included both the inbound experience by virtue of the submission of health care claims and the payment process. Not the payment process, but the payment itself coming out of the health plan.  It included all the experience, not just hospital inpatient experiences, but also ambulatory, and ambulatory as it extends to clinical, medical, pharmacy, lab, dental, all sorts of sources.

Here we are with the current state of development of some of these claims-based data sources.  I think our agenda today is going to let us go look at a little bit of the past.  I mentioned a few things.  Where we are in the present and where we go into the future, particularly if we look at claims as an ongoing potential source of data.  Certainly we still very much depend on that type of data.

But if we look into sort of future models of care delivery, ACOs and others, and future payment models like alternative payment system, bundle payment, that do not depend too much or at all necessarily on an encounter-based or service-level, if you will, data on a fee for service model, but move more into the pay for performance model in which the collection becomes much more expansive because now we are going to be considering a lot of other data, not just the data that is contained in a regular claim.  It might be driven not by claims directly, but by other types of data sources. 

I think we are going to have an opportunity to really look into the future and see how and where this data from traditional claims will continue to support some of the activities we do, as well where are some of these other data sources.  I want to stop there and turn it to Denise for her introductory remarks then.

MS. LOVE: Welcome, everybody here.  I won’t go into the state stories too deeply because we have the states to tell their own stories.  We will open up with just a little level setting about what is happening in the state arena around all payer claims.  As Walter said, most states have some form of hospital discharge data reporting.  That data reporting, since the ‘80s, is pretty stable.  It is used for many things in population health, things that we didn’t even dream of when we started.  It was health planning and certificate of need. Now they are a major source of population health databases in states.

But states were running into this problem of there were no data on payment, on costs, and outpatient.  That is where all of the utilization was going.  The early states, and I will talk about that in just a moment, started filling these gaps.  Today is a conversation.  I look forward to the day-long conversation about how we are going about filling these data gaps.  It is not easy.  You will hear some stories about how challenging it is.  It is definitely moving, in my opinion, in the right direction to filling those gaps.

I want to say a little bit about the APCD council because you will hear about that today.  The APCD council is a learning collaborative. Joe Porter back here is co-leader with me in this learning collaborative.  It is a joint collaboration between University of New Hampshire and NADO.  We really just coalesce the states to work on common issues, technical assistance, shared learning.  We are catalyzing states to achieve mutual goals.  You will hear later today about one of them is standardization, which is much needed.

Why are states doing this?  I don’t need to tell this group here, but maybe on the phone.  I don’t know if you have my slides, but transparency, health care reform, high-tech, patient-centered medical home, some of the SIM grants and transformation grants, Medicaid demand new kinds of data, rate-setting through societal. These databases are developed to meet a myriad of needs.  The needs are emerging almost more rapidly than we can collect the data to fill those gaps.

What are they in the state arena? The database is typically mandated through law.  Medical, pharmacy, dental claims are collected with the eligibility and provider files.  We have public and private payers aggregated in these databases, the commercial carriers and then Medicare and Medicaid.

This depicts both what is in the APCD, but also some aspirational pieces, the future.  Tricare, VA, Indian Health Service and Federal Employee Health Benefits are ones that states would like to have.  Some states have a higher mix than others of these payers. But they are not our primary focus right now because it is just a different way of getting the data.

The states typically start out with a commercial collection and aggregation of the data of commercial payers in their state.  Then they will get part C and D Medicare in most states.  But then they will go and work with Medicaid and bring in Medicaid into the database.  Then Medicare parts A and B, we can talk about that a little later.  We have worked closely with CMS to work out a state research extract process by which states can request the detailed Medicare claims and make that a part of the all-payer claims database.

This is our May progress map.  Every time we do this map, it changes.  The dark blues are the full implementation states.  Those are the states with active aggregation of the claims data.  The darker blue, New York and West Virginia and Washington, I believe are about to go into production or go live.  The lighter blue states are states that are exploring possibly talking about legislation.  We have been out to do educational seminars.  They are likely to turn to a darker blue.

Then the two light blue states, California and Wisconsin, are states that have a regional collaborative on a voluntary basis.  The same data, just a little model.  There are different business models.  We will hear about some today at the national level.  But the state approaches typically are the state led.  You will see the bulk of the states have a state agency that is overseeing the development of the all-payer claims database with advisory committee typically in statute.

Public, private, we are seeing some emerging collaborations between the state and the private sector, Colorado, Virginia and Arkansas where the state has the authority, but delegates the day-to-day operations to a non-profit or academic center in their state.  Then again, I mentioned Wisconsin, California is depicting some private sector collaborations.

I won’t go through all of this because you will hear some of the interest by the various stakeholders that have key interests in APCD.  But these conversations in the states, and we have been to most of the states that are on that map in their internal discussions.  Stakeholders come to the table with different wishes and needs for the data, but also different concerns.  Those are vetted out and addressed in the states that are able to move into the darker blue.

They have figured out how to billed value in for these various stakeholders and how to address the concerns that people have about large-scale aggregation of claims data.  But you will hear today stories about how the states are transforming the data into information that is actionable to improve the health care delivery system in their states.

I won’t go into state experiences, but I would refer you to the APCDShowcase.org at the APCDCouncil.org website.  We have portals.  As states release a new website report, we populate that through these different categories in the website. 

In short, and you will hear again in the themes throughout the day, but the all payer claims databases, there are holes and we will talk about those, but it is an almost complete sample of a state’s insured population.  The large sample sizes provide more precise estimates for individual payers and for the population studies in the state. 

Larger numbers are protecting or help protect patient confidentiality when you are looking at certain populations and conditions.  The larger sample size really helps there.  They are filling information gaps for state agencies for payment reform planning and evaluation.  ACOs are a perfect example of how a statewide APCD can help the ACO track their total cost of care per patient and their efficiencies.  Transparency tools and clearer picture of health costs are emerging in these states. 

I just will refer you to the contact information for Joe Porter, myself and the other APCD council members.  This is truly a labor of love.  There is just no way to say it.  We do this, a lot of in kind and a lot of heavy lifting, without a steady stream of income.  It is the right thing to do.  The states need the help.

It is powerful to bring the states together to work on the common solutions.  If states are resource tapped.  They don’t have a lot of money, so they work together and share and transfer knowledge across the state lines.  It really does leverage the best practices and shared solutions and even shared tools across the states.

With that, I will just move on because we have some wonderful stories to tell.  We will move onto part one, the panel, on policy and reporting issues. This panel will talk about how the various entities are achieving value purpose and their structure of how they go about their data enterprise. 

We have the state representatives.  I will let them introduce, but from Massachusetts and Colorado CIVHC.  Then we have the HCCI Health Care Cost Institute, Eric Barrette and Blue Cross/Blue Shield.

DR. SUAREZ: We should let you all introduce yourselves as part of the testimony.  I think most of you usually introduce yourselves during the testimony.  We will let you do that.

MS. LOVE: We will start with Deb.  I think we can just kick off.  You can introduce yourself. 

Agenda Item: PART 1: Policy and Reporting Issues

Panel 1A: Value, Purpose, Structure, Public Reporting, and Policy Consideration

MS. SCHIEL: My name is Deb Schiel.  I am senior director of analytics for the Massachusetts Center for Health Information and Analysis, also known as CHIA.  Thank you for this opportunity to provide testimony related to benefits, structure and uses of the Massachusetts all payer claims database, as well as highlight some challenges associated with collecting, maintaining, analyzing and reporting such a complex dataset.

A little highlight on CHIA, we were created as an independent state agency through the enactment of chapter 224 of the Acts of 2012, an act improving the quality of health care and reducing costs through increased transparency, efficiency and innovation, six years after the passage of Massachusetts Health Care Coverage Reform Law, the precursor to the ACA.

The agency is funded through assessments and hospitals, ambulatory surgery centers and payer surcharges.  CHIA is tasked with detailed data collection from both providers and payers that relate to health costs, premiums, utilization, enrollment and financial performance.  The largest and most resource-intensive of the data collection is, of course, the APCD.

I think Denise has gone over the benefits of the APCD, but I will highlight we are totally on board with that.  It supports analysis and research related to the impact of health care reform initiatives.  It allows us to understand trends in health plan coverage for Massachusetts residents, understanding trends in both plan-paid and patient out of pocket costs for commercial insurance products, Medicaid delivery systems and program, understanding trends related to utilization, quality and access to health care services in Massachusetts.

It allows us to understand price variation by payer, product, region and health care systems.  Understanding how ACOs, patient-centered medical homes affect quality, cost, access to health care services across payers, health systems and populations of Massachusetts. 

An enhanced attribute of the APCD is the Master Patient index, which allows APCD users to track Massachusetts residents across payers and health insurance products.  For instance, research could track health plan migration patterns of residents as they move between plans and insurance products.

CHIA is also using the APCD to reduce the payer’s burden across state agencies.  For example, CHIA is currently developing reports source from the APCD so that payers will no longer have to submit enrollment and utilization statistics to the commonwealth division of insurance in the very near future.

Just to highlight on the scope of the APCD.  You will all be familiar with this.  Payers, including Medicaid, which is known as MassHealth in Massachusetts, provider monthly APCD submissions that include the following file types. Member enrollment, medical claims, pharmacy claims, dental claims, insurance product and provider information.

CHIA recently added an enhanced member eligibility file to the APCD specifically for MassHealth or Medicaid program.  This additional file is necessary because Medicaid enrollment information does not conform to commercial enrollment file specifications due to attributes that are unique to Medicaid programs, including daily eligibility determinations, disability status, dual-Medicare and Medicaid eligibility, as well as MassHealth’s unique delivery systems.  CHIA used this enhanced file, along with APCD claims, to publish a report related to MassHealth enrollment and cost for its primary clinician plan and fee for service plan.  I have included the link on our information.

Ongoing challenges of the Mass APCD, and everyone will be familiar with this, assessing, documenting and improving the quality of the data that is submitted by the payers is extremely challenging.  If you have experienced managing, assessing, documenting and mitigating data quality issues at a single health plan, imagine the challenges of this effort across multiple payers and insurance products.  This work is performed by CHIA staff and is not outsourced to external vendors in Massachusetts.

Publishing reports that are sourced from the APCD is critical in order to accelerate improvements to the quality of the data files that are submitted by the payers.  In other words, use it and you will improve it.  This improvement largely occurs because CHIA validates all data within a report prior to publication, which involves close collaboration between CHIA staff and payers in order to understand discrepancies and initiate corrective actions.

An example of a CHIA report that incorporates as agency-payer collaboration into the publication cycle is the Massachusetts Enrollment Trends Report.  Payers and policymakers have informed CHIA that they commonly refer to this report, and payers also provide suggestions and feedback on future enhancements.

We will be talking about the 42CFR part two a little bit later on this afternoon.  But I would like to highlight some of the challenges associated with this proposed rule.  If 42CFR part two proposed rule is adopted without changes, it would likely present considerable challenges to high priority research and analytic agenda for the following reasons. 

If payers excuse substance use disorder claims from the Mass APCD, it would impede the ability for policymakers, researchers, payers and providers to analyze and report on trends related to opioid use and treatment.  In addition, payers may use different algorithms to identify and/or excuse substance abuse disorder claims, which would reduce standardization of the APCD medical and possibly pharmacy claims. In an effort to mitigate these challenges, CHIA is proactively developing a standard approach for all payers to use to handle the substance use disorder claims.

Although there are ongoing challenges with supporting and maintaining the Mass APCD, it is the only dataset in Massachusetts that fully supports researchers, policymakers, our sister agencies and other APCD users in understanding the impact of health care reforms in terms of health care coverage, both public and private utilization cost and quality of health care services provided to Massachusetts.  Thank you again for this opportunity to provide testimony.

MS. ENGLISH: My name is Ana English.  I am the president and CEO for the Center for Improving Value in Health Care.  We are also the administrator for the Colorado All Payer Claims Database. 

Let me first give a little bit of an overview regarding CIVHC.  The genesis of CIVHC was through the recommendation of the Colorado’s Blue Ribbon Commission on Health Care Reform, which was in 2008.  The Colorado APCD was established through legislation in 2010.  Then CIVHC was subsequently named as the administrator of the APCD by the Colorado State Medicaid Agency, which is the Department of Health Care Policy and Financing.

We are independent.  We are a non-profit and we are non-partisan.  We are not structured within the state government by any means, other than through our relationship and oversight of the HCPF office.  Our goals are very simple.  It is to achieve triple aim plus one, which is basically the plus one for the benefit of Coloradoans.  Our goal is for better health, better care, lower cost and create greater transparency and access to data. 

Our areas of focus are payment reform and delivery system redesign.  We also have a very specific focus and mission as it relates to public price transparency.  The Colorado APCD contains information that allows us to inform a diverse group of stakeholders.  The Colorado APCD is considered Colorado’s most comprehensive claim data set available. We contain over 450, actually it may be closer to half a billion right now, of medical and pharmacy claims.  We represent well over 4 million unique lives of the Colorado total population.

We have added this year claims from self-insured payers, which currently, even with the ERISA role, we still have approximately over 700 to 750,000 additional covered lives under the self-funded segment.  The Colorado APCD as it currently stands represents approximately 80 percent of the insured lives of Colorado across the entire state.  This covers commercial Medicare and Medicaid, both fee for service as well as Medicare Advantage.

The key components that I think are really important to keep in mind is to any APCD, but in particular Colorado’s APCD success, as well as some of the corresponding challenges, is one and first and foremost is our governance model.  Because we are a non-profit and we are outside of the government realm, shall we say, we believe that we have to have a broad and diverse stakeholder. That stakeholder involvement is integral to our success.

We have a data release review committee that is made up of a broad diverse set of very highly capable policy, as well as privacy and security experts.  They ensure that we have compliance with HIPAA, HITECH, both privacy and security, as well as DOJ anti-trust guidelines. 

We also, through statutory mandate, manage and facilitate an APCD advisory committee of which the members of the committee are actually named by the governor and his designees. We also have the CIVHC board of directors, which is again multi-stakeholder, as well as representative across the entire state.  We believe that in each of these committees, it is extremely critical that we have multi-stakeholder representation, and that they are leaders within the health care industry.

Talking a little bit about one of the things that is extremely critical for us, and I know there is going to be more discussion related to this, is that we currently have and very much work through standard formats that are utilized by all of our submitters.  We ensure that we go through comprehensive reviews of which the submitters are very actively engaged for any proposed changes that we make.  We will make adjustments based on the recommendation of the submitters, as well.

Though Colorado is highly fee for service at this time, based on the trends for going forward, as well as some of the current situations that we have, there is very specifically a need to supplement with non-claim payments, in particular capitation and other incentive programs.

One of the key points that we really need to bring out is that having comprehensive population representation across the entire state is extremely critical.  The community-based needs and the local needs, not just from a statewide perspective, but within each of the varying markets and areas within the state, they have very different components, needs, as well as perspectives, from a provider perspective, a payer perspective, the level of competition, access to care, access to non-clinical social service, as well as social determinations that drive the overall care and status of that particular market.

We truly believe that we have got to, from an APCD perspective, from a data management and analytics perspective, have very solid, good representation within each of the markets, as well as have a greater level of awareness of what the needs are of those markets.  Having an APCD that has low population representation in certain markets lowers the perceived and real value of the APCD.

Currently having over 80 percent of the Colorado population of the APCD and under administration, it is considered to be one of the most comprehensive and valuable databases within the nation.  Yet, we must continue to push and obtain additional sources of data, such as the ERISA based self-funded federal employees and programs such as Tricare, as well as the uninsured encounter information, along with other pockets of the population.

The key that we believe to the APCD success is access to invaluable reporting of the information, and all within the framework of HIPAA HITECH, as well as DOJ anti-trust guidelines.  Our primary vision is that the majority of Colorado stakeholders are using the APC data to help inform opportunities for change. 

The key here, though, is the APCD is one large part of the overall health care data needs.  CIVHC is working with many partners throughout the nation, as well as the state, on initiatives to increase the value of the APCD by not only increasing the claims representation across the state, but by participating in initiatives for the development of social determinants database, benefit plan design matching up, as well as the integration of reporting based on clinical and claims data, among other areas.

CIVHC in the APCD receives no state operating funding and is required by statute to be sustainable through the non-public release of data and analytics, along with specific grant-specific programs and research projects.  We are mandated, and basically our success is through the creation of valuable reports that can be utilized and help drive change throughout the state.

CIVHC’s role in advancing the triple aim, and this is where we consider it to be the true value of the APCD, as well as the work that we are doing in Colorado, but our mission is to cultivate an advanced strategic initiatives that improve the health of Coloradoans, contain costs and ensure maximum value for health care received.

How is CIVHC achieving its mission?  We start by focusing.  One of our first focus groups is care delivery redesign efforts.  CIVHC is the managing partner of Healthy Transitions Colorado, a care transitions collaborative that convenes leaders in the care coordination field to share knowledge and best practices and avoid duplicate efforts statewide.  We have over 100 organizations that are a part of the healthy transitions Colorado collaborative.

CIVHC also convenes a statewide palliative care taskforce.  In 2015, it published the results of a survey highlighting gaps in access. This group reconvened again in March of 2016.  We believe using APCD data to demonstrate the return on investments, or ROI, for these programs is critical.  We, in conjunction with our partner organizations, utilize APCD data to help inform opportunities for new care delivery models, as well as the associated payment reform models that may ensue from that.

Specifically in regards to payment reform, some of the recent payment reform efforts, CIVHC has a representative on the state cost commission that is representing the voice of the APCD, the cost commission on Affordable Health Care.  We provide them with analytics on potential cost-saving opportunities throughout the state.  We have bundled payment analytics that is based on the CMS’ comprehensive care and replacement model.  This analysis shows wide variation across specific Colorado ranges and across major categories of services based on DRGs, such as inpatient stay versus post-acute care.

One of the key areas, in addition to payment reform, and basically a sub of that, is the public awareness through transparency.  As the administrator of the APCD, we take information that is submitted by the insurance company and make the data widely available and actionable to basically improve health care and lower costs. 

Who do we serve?  At CIVHC, we serve health care stakeholders across the entire spectrum of care.  In particular, we support state agencies and insurance companies.  One analysis and effort in particular was evaluating evaluation in the premium rating areas across the state.  In addition, we also support federal initiatives, such as SIM, TCPI and CTC. 

In regards to researchers and non-profit organizations, one in particular is that we have a study that is ongoing regarding demonstrating how good nutrition results in cost savings for patients with chronic conditions.  Actually, it is a very exciting program that is showing very significant results pre and post-services. 

For providers and consumers, we have supported increasing reimbursement rates for optometrist. Again, these are just some highlights and some examples.  For optometrists who treat Medicaid patients resulting in increased access to vision care and vision services.

The public consumer portal, which is CoMedPrice.org, contains cost comparisons by facility for procedures, cost and utilization information that is primarily used by legislators, as well as community-based organizations and governmental institutions, regional government groups, those that are looking for the big picture of the health care cost across the state of Colorado.  In this case, we look at it as an example, ER utilization, utilization by service category, total cost of care, chronic condition prevalence, just to name a few.

Another way that CIVHC publicly releases data is through our cost drive or spot analysis reports.  These reports allow Coloradoans to identify cost savings in health care for the state.  In one particular spot analysis, we identified that by reducing unnecessary emergency department visits, the result could be an $800 million cost savings for Colorado.

The analysis shows that if consumers went to the doctor’s office, clinic or urgent care for non-emergent conditions instead of the emergency department, over $1000 could be saved per visit on average.  Further analysis shows average cost to treat common ailments, as well as reasons that Coloradoans opt for emergency department instead of office visits is extremely telling and is extremely actionable, as well.

Next, I am just going to go ahead and shift very quickly at looking at some examples of our non-public releases of data.  These are primarily through custom requests.  Custom requests are based on different organizations’ needs.  Again, it is covering across the entire set of stakeholders.  These reports can inform on market share, utilization, bundled payments, total cost of care, variation in services across each of the different markets and comparing specific provider groups to others.

CIVHC’s custom data request portfolio can be accessed through www.comedpriceshowcase.org.  You can see some examples.  Those that have allowed them to be named are named, and others are listed, but without the named recipient. 

An example of one custom request was from Duke University in 2015.  Duke researchers used our data to understand the effects of competition in insurance markets.  They looked at geographic price variation and the effect of insured competition on premiums.  Duke chose Colorado’s APCD out of the 17 available APCDs because it was one of the most comprehensive. 

Another example of a non-public release of data is CIVHC’s ability to conduct bundled payment analysis or episode-based analysis.  The new CMS program, as most of you all know, the comprehensive care joint replace model began on April 1st, 2016.  CIVHC is able to provide side-by-side facility comparisons, including breakout of post-acute cost.  We have made that information available to inform hospitals on how to identify areas for cost improvements and enhanced patient care.

CIVHC is now implementing the 84 Prometheus episodes developed by HCI3.  We will be expanding that across all commercial Medicaid and Medicare.  CIVHC is always striving toward making the data available more digestible and providing the most comprehensive look possible at the health of Colorado. 

In 2016, as well as initiating last year, CIVHC has been enhancing its resources to provide the best possible picture of health care in Colorado.  How will we do this?  More claims.  We need more claims.  We need to continue to be more comprehensive across the entire population.  The addition of self-insured and dental claims in 2016 provides an even more comprehensive look of Colorado’s health, more custom reports. The increase in our internal workforce to fulfill customized reports and maximize non-public release of data. 

Enhance tools and access, we have developed a subscriber portal for the use of, in this case we are using the Tableau network software.  It is a data visualization tool that is used for developing interactive reports.  It allows us to basically develop more reports, more standardized reports, thus increasing access, as well as decreasing our cost to implement, as well as decreasing the cost for the users.  This allows us to meet our overall objective of increasing the access and again to achieve lower costs, as well.  I thank you for your time and allowing me to talk about the Colorado APCD and its administrator, CIVHC. 

MS. LOVE: We will move to a different model.  We saw one state that is inside the state in a state model, a state non-profit governance model.  Now we have some private sector data models that we will hear from.  We will leave time at the end for the committee Q&A.  I will turn it over to Eric Barrette.

MR. BARRETTE: My name is Eric Barrette.  I am the director of research at the Health Care Costs Institute.  I appreciate the opportunity to testify here today.  I would like to also add that my comments are from a multi-payer national database perspective, but they are my own.  They do not necessarily reflect those of HCCI, the executive director, the governing board or data contributors.

I will briefly introduce HCCI and its activities, and then focus on two areas that HCCI feels are very important.  One is data governance and the second is the value of a national multi-payer database.  I will conclude with a few of my own observations.

HCCI is a non-profit independent non-partisan research institute dedicated to reporting and promoting research on the drivers of health care costs and utilization.  We believe we are building out an essential part of the nation’s health services research infrastructure, but this is a joint effort.  It involves working with health insurers, policy leaders, states and federal agencies.

HCCI was founded in 2011 by Aetna, Humana, Kaiser Permanente and United Healthcare, which are four of the largest commercial insurers in the US.  But we are independent of them, with an independent governing board comprised mainly of academic health economists.  We do receive financial support from the insurers, but we also receive financial contributes from private foundations and licensing fees to cover the costs of providing access to the data.

I would also like to point out the data contributors do not gain access to any of the combined dataset or any proprietary or confidential analysis conducted with the data.  We currently hold claims with allowed payment amounts of actual prices for more than 50 million Americans from 2007 onward, which we make available for academic and non-commercial research.

The commercial claims come from all 50 states and D.C., and include employer-sponsored insurance, individual insurance and Medicare-managed care.  It is updated annually.  The research dataset is compliant with privacy and anti-trust requirements.  Additionally, by the end of 2016, we expect to hold through the qualified entity program 100 percent of part A and B Medicare claims, as well as many part D claims that are statutorily allowed through the program. 

HCCI’s day-to-day activities can be characterized into three main categories, through which HCCI seeks to carry out its mission.  First is public reporting.  All HCCI research products are made available for free in the public domain.  HCCI does not engage in any commercial proprietary research for any person or organization. 

The types of reporting we do include national reporting on employer-sponsored insurance, cost and utilization, as well as more detailed reports on topics such as cost and utilization in the diabetes population or out of pocket spending trends. 

Second, HCCI has a transparency initiative.  Guru.com is an independent, free, user-friendly source of price and now quality information for consumers of health care services.  Finally, we support research.  All researchers using HCCI data agree to use the data for academic research projects.  Currently, there is over 20 institutions using the data from some of the top public and private research institutes in the country, including Yale, Northwestern, Stanford, University of Michigan, Minnesota and the University of California Berkeley, as well as multiple government agencies, including the FTC, the Office of the Actuary, MedPAC and CBO. 

Among those research teams, there are over 30 active projects on a wide variety of topics, including multiple dissertations at some of those universities.  The topics include issues related to price transparency, ACOs, price variation, geographic variation, provider competition integration, health care workforce issues and state health policy impacts.

We also believe that HCCI contributes more broadly to the multi-payer database arena through its innovative approaches to data governance and public reporting, which I will speak about next.  HCCI has focused substantial time and energy on data governance. Data governance includes both working with the data contributors and data users in order to maximize the benefits and minimize the risks. 

There are numerous logistic and legal considerations for all parties involved.  The first issue is related to data holding, which affects the data contributors.  HCCI operates through a voluntary participation model as opposed to a mandatory model.  This includes buy-in from the participants up front, but also assures their interest and commitment to the project.

Contributors ensure complete and validated data is provided to HCCI.  HCCI then ensures that the data is securely stored and accurately received.  We engage the payers, actuaries and data teams on a regular basis to ensure the process is working well, and that any improvements that can be made are made, and that there is a common understanding to the process.

The second issue of data governance is data use.  This effects both sides, the contributors and the users.  HCCI works with the data contributors and researchers to ensure the broadest set of data use rights that still protect the data.  This means data contributors do not have control or veto power over the research agenda.  But researchers are prohibited from activities that pose a threat or risk of HIPAA violations, anti-trust violations or the release of any proprietary information.

Finally, there are issues of data access.  This is mostly relevant to the data users.  Because the size and security requirements of large datasets have implications for access, HCCI has moved into a data enclave environment supported by the University of Chicago and ORC.  This provides multiple layers of data security and access.  We have the ability to limit or end data access in a moment’s notice, limit the data merges and the use of the data, so only authorized datasets can be merged into our data after it has been reviewed by our external statistician to ensure that the data will remain statistically de-identified.

Final review of the results to ensure that no activities have been conducted that are not allowed.  Also, the enclave approach is scalable and requires fewer requirements of the users on their end.  Additional perspectives on the governance issues from individuals associated with HCCI are available in some published literature in the American Journal of Managed Care, as well as the health care’s blog.

In terms of the value of a national multi-payer claims database, assembling and analyzing data is not easy and it is not cheap.  At some point, there is a limit to the return on the data.  There are economies of scale to be achieved through larger datasets.  Even with sufficient standards, there are costs accrued by developing the databases in multiple locations multiple times. From my experience at the HCCI, as well as previous experiences, there is a high cost in time and effort in combining data. 

HCCI activities also include public reporting and benchmarking, research support and transparency, all of which benefit from large national data sources. The 50 state plus D.C. multi-payer standard is similar in some ways to the Medicare dataset and allows for national reporting and the development of national benchmarks.  In the past, HCCI data has been used to benchmark the Vermont APCD.  Through a state health policy grant program, researchers have used HCCI data to evaluate policies in one or more states versus other states and nationally.

A consistent dataset also benefits researchers by allowing for results to be replicated and institutional knowledge to develop around the data, which minimizes start-up costs for the researchers.  Through the academic partnership with dozens of researchers that are using the data, it allows for knowledge transfers, as well as the development of new collaborations.

Finally, from a methodological perspective in public reporting and through transparency, there is consistency with the national data.  We provide as much methodology related to the public as we can.  We are always striving to provide more detail.

I would like to conclude with two overarching observations.  First, data protection is critical, but it entails more than HIPAA compliance. Payers, health care providers and individuals are correctly concerned about what is going on with their data.  There are numerous regulatory ethical and logistical issues to consider.

Thought needs to be given to who gets the data, how they get the data and what it may be used for.  Also, thought needs to be given to what will actually happen with the data.  Approving data for certain uses does not guarantee that is what it will be used for.  This is why data governance is key. 

Second, building and maintain data is not easy and is not necessarily a money-maker.  You cannot presume that if you build it, they will come.  If you collect it, it can be used, or if you analyze it, it will be useful.  Collecting files from claims data warehouse does not directly result in a data set.  How the files are standardized, even among different fields, can have impacts.  For example, the same field could be reported as a 1-0 or a yes-no.  This has implications for research if this is not dealt with.

Second, how consistent the data feed is matters.  Is it always the same groups and members, the same contributors and consistent reporting periods?  If not, there are implications for researchers and their results, which will directly affect the policy implications.

Finally, there are tradeoffs to every decision made.  Receiving files with less run out or more frequently provides more timely data, but it is less complete.  Updating data more frequently improves the accuracy or representativeness.  In other words, it may be more complete, but then it reduces the ability to reproduce previous results.  There is a higher cost to updating the data.

Bigger files received less often versus smaller files receives more regularly impose different labor and capital costs.  These are types of considerations that need to be considered when developing a dataset.  Thank you for your time.

MS. LOVE: Thank you.  Next, we will hear from Joel Slackman, Blue Cross/Blue Shield.

MR. SLACKMAN: Thank you very much.  My name is Joel Slackman.  I am an executive director for policy at the Blue Cross/Blue Shield Association.  On behalf of Blue Cross/Blue Shield and its 36 plans, I would like to thank you for the opportunity to testify.

Let me start out by saying we think there are three reasons that now is not the time for this committee to be promoting the current centralized APCD approach.  First is that centralizing data, as some of the other speakers have already alluded to, across multiple payers is expensive.  It is time-consuming, and it is prone to issues of data integrity.

Second, and I think perhaps more important, is the elephant in the room that hasn’t been discussed yet.  That is the recent Supreme Court’s Gobeille decision that greatly diminishes the value of states’ centralized data repositories and, as you will see, opens a Pandora ’s Box of legal uncertainty.  Third, there is a better way.  States have a better way that is not fraught with the uncertainties arising from the Gobeille decision, a distributed data approach.

Let me expand on these points.  We at the association, like HCCI, have considerable experience with the challenges in assembling meaningful claims data across multiple independent companies.  Yes, Blue Cross/Blue Shield plans are all locally controlled independent companies.

Over the years, we have built an enormous cache of claims resources for analytic purposes, contains data on more than 107 million people covering every ZIP Code in the country.  We have used these data to provide insights into such issues as cost variations in knee and hip replacement surgeries and other sorts of public health-related studies.

It has taken us years of effort to achieve this ability to aggregate data in a meaningful way.  Maintaining the integrity and accuracy of centralized data is inherently challenging.  Even if centralized data are highly curated and scrubbed, which we believe isn’t necessarily the case with all state APCDs, issues come up where it is essential to call on the people who produce the data, which is difficult to do when the data are centralized.

This is a fundamental point that Mark McClellan made in his comments on CMS’ original proposal several years ago to create a centralized database for risk adjustment.  I will be coming back to this later in my testimony.  But more important perhaps than these technical issues are the legal challenges to centralized APCDs arising from Gobeille v. Liberty Mutual. 

In the short-term, I would argue that this ruling undercuts a major rationale for states investing in APCDs.  I am going to quote from Gobeille, an all-payer claims database that omits self-insured plans cannot be considered to be accurate or comprehensive.  This point was picked up by several amicus briefs in favor of Gobeille.  For example, the Harvard Law School Center for Health, Law and Policy Innovation wrote that without self-funded claims, APCDs will no longer provide an accurate portrait of the health of the general state population.

Well, losing self-funded claims is bad enough.  But Gobeille could ultimately deprive APCDs of claims from fully insured ERISA plans, too, virtually eliminating all commercial data.  Outside legal counsel advises that the language of the Supreme Court decision itself on its face appears to strike down the laws that relates to all ERISA-governed plans.  That is whether they are fully insured or self-funded or self-insured. 

At the least, future litigation is highly likely.  That casts a poll of uncertainty over any prospective APCD investment and should be considered by this committee.  The uncertainties also extend to any potential regulatory actions. 

What is a state to do?  We believe that distributed models provide a proven alternative that can work around the legal issues and uncertainties arising from Gobeille.  In the words of Dr. Richard Platt of Harvard, who was the principal investigator for Mini Sentinel, which turns into Sentinel, the nation’s largest distributed health network, which I will talk about in a moment, a distributed network can perform essentially all the functions of a desired centralized database without the disadvantages.

I would note that it is ironic that it is an amicus brief on Gobeille, the Harvard Law Center stressed the value of claims data by noting, quote, claims data allow researchers to study treatments once they are approved by regulatory bodies, such as the FDA.  Well, the FDA itself hasn’t created a centralized database.  It adopted a distributed model for such research.  That is today’s Sentinel system.

Sentinel, whose data partners include health plans such as Anthem, Blue Cross/Blue Shield, Blue Cross/Blue Shield of Massachusetts and other commercial payers like Aetna now has billions of claims.  It also includes a growing portion of EHR data, currently about 10 percent of all the data in Sentinel.

The FDA has used Sentinel not only for safety surveillance, but for public health analyses. For example, it looked at what is a major health problem, which is atrial fibrillation, and the use or rather the non-use of anti-coagulation.  As a result of this study, the FDA is now engaged with research partners in carrying out randomized clinical trials.

A distributed data structure like Sentinel containing claims information on more than 30 million lives in the individual and small-group markets is already operating in every state.  Let me explain.  Several years ago, CMS had to decide how it was going to obtain and process claims data for risk-adjustment calculations and also for reinsurance as is required under the ACA. 

CMS initially proposed, but then it rejected, a centralized approach in favor of a distributed model.  CMS implemented the so-called external data gathering environment or edge servers.  Issuers upload claims data to the edge servers.  The edge servers run CMS-developed software to execute the risk adjustment and then plan summarized data, not individual data, are reported back to CMS.  This is pretty much the data model that the FDA uses in Sentinel. 

I would urge the committee to look at edge servers as a valuable potential source of information for states.  It certainly merits further consideration.  These are data that are already standardized and could provide a lot of insights. 

Finally, let me end by mentioning the opportunities for partnership.  We have heard some of the briefs submitted in favor of Gobeille argued, and I will quote from one of them, that even if the majority of insurers cooperate to form a health care claims database, it will not deliver comprehensive data to researchers, limiting the usefulness of a private health care claims database. 

We absolutely disagree.  I would think Eric Barrette would, too.  As I noted at the beginning, the Blue Cross/Blue Shield Association has a vast repository of data that would deliver data equal incomprehensiveness, maybe even greater than any current APCD.  We are not keeping the data to ourselves.  We, like HCCI, have been entering into active partnerships with research institutions to mine these data.  In that same vein, state-sponsored research initiatives could partner with BCBSA to mine what is currently the nation’s largest store of commercial health care claims data. 

In closing, distributed data models and public-private partnerships, which are not mutually exclusive, deserve your consideration as alternatives to the current paradigm of centralized all-payer claims databases.  That concludes my testimony.  Thank you very much.

MS. LOVE: Thank you.  Now, we will enter the Q&A for the committee.  I think I will start.  My head has many questions spinning.  One of them is I guess I don’t understand the distributed model.  Can you give an estimate of what a cost for a distributed model would be across the 50 states?  How much does the Sentinel system cost?  How much was invested by FDA to implement that?

MR. SLACKMAN: You would have to ask the FDA that question.  I can provide the committee an analysis that we commissioned from Hewlett Packard back when CMS was looking at developing a centralized database.  I think essentially, the costs are similar.  Ultimately, I wouldn’t think the cost would be that different.

The essential point that I want to make, though, is if you are a state considering investing money, and by the way, over the last couple of years, there have not been any new state APCDS largely because of resource constraints.  If you are a state thinking about investing money, in light of the Gobeille decision, I would ask that a state might find it far more cost effective to invest in a distributed database or exploit the distributed databases that already exist in the current edge server environment. 

MS. LOVE: If you could explain how they work functionally, as I understand research and analysis are conducted.  You might have multiple tables in a database.  I don’t understand what do you mean by distributed database.  It would be really helpful to understand.

MR. SLACKMAN: The essence of a distributed database or a federated model is that the data stay with the owners of the data or the people who generated the data.  Generally, CMS has done this and the FDA has done this, create a common data model.  Everybody, as in some of the more advanced state APCDs like Massachusetts and Colorado standardize their data.  They upload it to this edge server, which is the equivalent of a data enclave. 

Then third parties, in the case of the edge servers of CMS, in the case of Sentinels of the FDA, creates software programs to analyze the data.  The data are analyzed in the edge server.  Then the aggregated results are sent to the third party. 

Now, this sort of analysis could be as simple as, let’s say, you are interested in a prevalence analysis like the state of Utah did with its APCD looking at the prevalence of, among other conditions, depression.  That is pretty simple.  You go into the APCD.  You run the data on the claims.  You get information because you have the diagnostic information.  You have information on providers and so on.  You get the aggregate results, and you combine them.

You can even do more sophisticated analyses. There has been considerable research into regression analyses across distributed data models and longitudinal analyses. Again, I would commend anybody who is interested in distributed data to look at the research and the papers published by Richard Platt, who is currently with Harvard and still heavily involved in the FDA.  Does that answer your question?

DR. COHEN: It is more like the research data center, where the data steward holds the data, but external users through a series of permissions can get access to use the analysis.

MS. LOVE: How are the denominators handled if the data are distributed?

DR. COHEN: I am sure Joel can explain that better than I can.  The denominators are the same.  Standard line population from which the database is drawn.

DR. RIPPEN: Actually, CDRNs, that is their model.  It is a distributed dataset that has standardized data elements.  You can run sophisticated SAS queries or whatever kind of queries that you like.  It is kind of an 80/20.  There are certain things that you actually need the data, especially when you do big data analytics.  It is a little bit more nuance.  But for most things, it actually is pretty effective.  It just doesn’t let you do everything.

DR. SUAREZ: I think the issue is really scaling up from a state or sub-state level dataset to a national level.  I think it is part of the question.  I think that is one of the points.

My question, and I know we have several people in line for questions, we will go through it.  I just wanted to mention, I have many questions about governance and data, governance and comprehensiveness of the data, sustainability of not just the state ones, but national ones, too. 

Probably the most significant one that I have as a question is really about the value that has been able to demonstrated with the data.  There are a lot of examples about the use of data and how data can be used and had been used for analysis on reporting.  I think in part, at the end of the day, we have the triple aim.  We have improved the health, improved the care, lowered the cost.  I would even add a couple extra to the triple, include improvement in care, improve access to health, certainly the population health and public health side of it.

So my question is really about that.  How much can we document and demonstrate that the reporting and the use of the data has contributed to or led to cost-reductions, quality improvements, improving access, ultimately the triple aim?  If there are examples of that, that would be certainly helpful, I think.

MS. ENGLISH: I think the key to the success, at least in the case of Colorado, as far as the key to the success of the APCD is three-fold.  One is we have to have the foundational component of data and analytics, which is a core for any APCD, whether it is being managed at a more higher scale level or whether it is at the state level.  That is integral to everything we do.  That is all about informing opportunities.

We have talked a little bit about regarding what analytics are incurring to inform opportunities.  But a driver, at least within Colorado, and I believe that is the case within many of the other states, is what then is done to affect the change.  That is why it is so critical that all of the stakeholders come together. 

We work very specifically in convenings of bringing together stakeholders.  We actually are pushing quite a bit more at the employer level.  When you look, for all intents and purposes, the payers are very much considered the payers.  But who is going to drive whether it is benefit plans, et cetera, putting pressure, helping to drive and feed into the consumer environment, as well?  It is the employers, as well.  We really are bringing each of those partners together, bringing convenings on whether it is payment reform or alternate methods of delivery system.

We also work as an organization taking that underlying information and analytics, again to inform opportunities.  But that is not enough.  The question I think is excellent and is right on in that we have got to be able to then bring those organizations together.  I have to say it is very local. It is very specific to the existing environment.  It differs in the metro areas versus in the rural areas as to what the needs are.  Access to care, the amount of competition, the pricing variation that occurs across the state, there are an extreme number of factors.

The only way that we are able to truly drive the change, and CIVHC is spending a lot of time specifically in this area of how do we get those stakeholders to own the data, own the results, and be disruptive.  That is the challenge I think we have.  We have very specific steps that truly are.

We gave some examples of how we are informing change.  But there are very specific examples where we are bringing the health systems together with certain particular groups, identifying what the potential savings are, and then having very specific actionable items that need to be driven by the businesses.

Again, CIVHC is a support environment.  I think we are playing as a support mechanism within the disruption and the change.  We have to get those that are in their business, between the payers, the employers, the providers getting together to say, this is the right thing to do.  It is a challenge, but it is also what needs to happen.  I don’t know if that answers your question.

DR. SUAREZ: I think that does help.  Thank you.  I think in the case of CIVHC, it sounds like you are not only being the data source, as you say, the most comprehensive database in the state.  You are also trying to be the agent of change, very different role from we are the data source, here is the data.  Someone else, go ahead and get together the employers and the health plans and the providers. Taking this data and this evidence, affect a change to improve access, to improve quality. 

MS. ENGLISH: There are different models that are out there.  In particular, you have APCDs that supply just the data and the analytics.  Then you have these regional health collaborative that are very closely aligned with using that data and pushing it out.  CIVHC is in a situation that we are a regional health collaborative and also is the administrator for the APCD.  We drive a lot of the analytics based on what the needs and what we are hearing from the stakeholder community.

DR. O’GRADY: I didn’t hear a specific.  I didn’t hear drug company X took a look at what the price of their drug was, used your data to identify populations who weren’t being served, and then therefore made a policy change that allowed for better care at a lower cost.  Do you have something that is a specific case that would address Walter’s concern of where somebody has actually used this real-life, real-data to make a policy change?

MS. ENGLISH: We have to step back.  At least from Colorado’s perspective, APCDs are very new.  We are continuing to gather the information.  The foundation has been strengthened and is, I have to say, extremely solid.  What we are doing now is we talked about adding the reporting.  We have been doing a lot of customized type of reporting and work.  We can give very specific examples of where payment reform has occurred because of analyses and work that has been done.  I have to say we are on the cusp.  I am speaking from Colorado’s perspective.

MS. SCHIEL: I would agree with that.  One of the cost savings, and it is a simple one, is the admin simplification in Massachusetts, where we are trying to offload some of the state reporting requirements from the payers since we have the data.  We will produce it for them.  That way, they can have a reduction in their resources in developing multiple reports from multiple state agencies.  That is just a simple one.  But you are right, we are on the cusp of being able to do this. 

MS. LOVE: To that end, we do have some states, and you may hear these stories later, using their APCDs to evaluate the investments in health care reform.  There is no other way to evaluate primary care effectiveness.

DR. SUAREZ: Thank you so much.  I think we are going to go around the table.

DR. COHEN: I can give you some more examples using the APCD from Massachusetts.  In my past life, I was doing research in the Department of Public Health.  There is an ongoing example I am still involved with and one that we are promoting.  One is the one that we are actively doing now.  Massachusetts has introduced legislation to look at the impact of our new casino gambling.  We are actually using APCD to identify persons who have diagnoses of gambling addiction.  We are going to monitor that over time to track the impact and identify risk factors using APCD data. 

Another project that I have been involved with, Massachusetts has the highest multiple birthrate in the country.  But there is no general population-based data around infertility, except for the National Survey of Family Growth, which is a very limited dataset.  Using the pharmacy claims and some of the diagnostic and procedure information and different CHIA APCD files, we are developing an approach to estimate population-based infertility.

Of course, both of these examples rely on the comprehensiveness and representativeness and completeness of APCD, which is a core issue that I think needs to be addressed, along with the other issue around standardization across states of data elements that are available, so that we don’t have 50 different flowers blooming.  We have a system. 

I think of the vital statistics system as a perfect analog where there are actually 57 different jurisdictions.  But they agree on definitions, so that we can create a national database and do policy at the local level all the way up through the national level.  I think it is really important because in previous days this week, we have heard of limitations emerging around survey data, cutbacks in federal resources. This is capitalizing on the existing data that are being generated. 

I will leave aside the issue of whether distributed model or centralized database is the best approach.  But these data are being generated for the entire population.  You know my focus is on population health.  Are there data ubiquitously available that can tell us more about population health in the community?  APCD certainly has that potential to be used not only for cost in health services evaluation, but to estimate risks and morbidity in ways that we have no other data available. 

I am very bullish on the possibility of the use of APCD.  But things need to happen that I think NCVHS and the federal government can promote around creating a more comprehensive approach that is representative and standardized.  I am sorry I am going on so long.

MR. SLACKMAN: Two issues that have concerned me.  One is the insurers and plans, commitment to collecting race ethnicity data on their databases and how you all deal with that issue.  The other is sort of a unification of the concept of episodes of care because these claims databases are, for someone with my background, very difficult to use for classic public health kinds of research activities. I would love for all of you to address those two issues, if you could.

MR. SLACKMAN:  I will be happy to talk from a plan perspective. Disparities, important information not collected on the claims.  Plans have used various techniques like geographic analyses, ZIP Code analyses.  Anthem in particular has been a leader in doing that.  But I think that is a really tough nut to crack if you are relying only on claims data, which I think leads to another question.  We are increasingly living in a rich ecosystem of data on people from physicians, EHRs, from various other data sources that give insight into the social determinants of health. 

If I were a state or a policymaker interested in information about disparities in health across different populations, whether it is sexual orientation or race or ethnicity, I don’t know that I would look at claims.  There are other places to look.  I am sorry, your second question?  Oh, episodes of care.

There are a lot of episode groupers out there.  We have a number of plans that have been using Prometheus.  I think Prometheus has proven to be so clunky and data-intensive that the plans have abandoned it.  I know that Blue Health Intelligence, which is a spin-off of Blue Cross/Blue Shield has been working on its own group strategy.

We currently make available to plans a cost-transparency tool that offers pricing and some quality information on about 1100 conditions.  We are working on developing episode grouper, Steve Bandeian, who some of you may know used to be with AHRQ and is now working for BHI on that effort.  That is another area where there are lot of different approaches to developing episode groupers, no standardization.

MS. ENGLISH: As far as race, we just actually did a study.  We received race on the eligibility file, which is optional.  Not all plans collect it.  If it is not collected by the plans themselves, it is basically only as good as how much you can populate. 

Then on the claims, we actually did some studies going back with the hospitals.  The reality is it is just not verifiable or can be used as a true source of truth because there is just a lot of filling in the blanks, then necessarily validation of the data itself.  It is a problem that we have.  We do have work that is going on regarding the creation of a social determinants database, which will come directly from the patients themselves.  But again, the population whereas our claims represents around 80 percent of our insured population, the social determinants is going to take time to get populated because it is going to get billed.

Regarding episodes of care, yes, there is clunkiness.  But truly, the HCI3 is a standardization.  We have evaluated a number of different episodes.  I believe that we are one, if not the first APCD, that is truly putting all of its data through the episode system.  Our intention is that it is starting more with the procedural, though we are loading it for all 84.  There will be some of the episodes that are highly valid.  The procedural ones are giving us a lot of good solid information.

Then we will be moving to understanding and validating the chronic and the different categories of episodes.  We have to start with some level of standardization that we apply to all.  One of the issues that are out there is that if there are different methodologies, it is very different to compare across particular population of one payer uses a different one from another.

MS. SCHIEL: CHIA has been testing a brand approach to identifying race and ethnicity through tracking of census track and things like that.  They are trying to handle it that way.  I don’t have an update on how that is going. 

In terms of episodes, you are absolutely right.  APCD is, at this point, data in the raw as we call it, but we are working.  Actually, we have some models of data.  They are aggregate data states that are going to be far easier for people to use that is going to include episode and both on inpatient and across episodes.  We are hoping to release aggregate data files that are far simpler for people to use and not have to come and do all the aggregation and the analytics.  It will be more analytically ready for folks. 

MR. BARRETTE: I am going to change hats and speak as a researcher.  Yes, race and ethnicity is not in claims.  There are approaches to dealing with it, whether you use census data or aggregated survey data.  There are many other things that are not in claims that would be helpful.  It is important to remember that claims is one tool.  I don’t know anything about people’s eating habits or their exercise habits. That will affect their health.  That will affect their health care. 

Claims alone will never include everything we need them to.  Eventually, if we can get more better race ethnicity data, that would be great. But then we just have more stuff to get after that.  I think we should keep that perspective that it is never going to be comprehensive.

In terms of the episode groupers, having used a variety of them, and now HCCI is using it.  We have about 295 care bundles on Guru.com.  Again, they help, but there are many ways to do it.  There are many things to consider when you are putting things together.  Not everyone’s episode of care will match what is in a group.  There is a lot to be learned in terms of how to group data.

I think up until now, a lot of the episode groupers have been based on data that is available, which is Medicare.  That might not translate to commercial data.  Our episode groupers have been validated in commercial claims.  But I think there is a lot more to learn.  They should hopefully be getting better with access to data.

DR. RIPPEN: Just having experience from a South Carolina perspective as far as value of kind of an all-claims database.  We are focused on the outpatient.  It was really to get everyone to agree what major cost issues there were, like say hospital readmissions.  Then also to assess the effectiveness of an intervention because that at least was available.  I think there are a lot of different uses for it. 

But I think that with the elephant in the room also how does that link into true clinical care.  If we think about electronic health records or what is happening from a clinical perspective, and then claims which might be, well, what is happening on the pricing, I think the big question is what is ultimately the role of each, especially since we are becoming more and more electronic. 

As you start thinking about determinants of health, is that something that is really associated with clinical outcomes and health outcomes as opposed more to the billing.  Then if you think about short-term versus long-term, short-term is you use whatever the heck you have and try to get as much information as you can.  You apply it to health things, even though we all know the strengths and weaknesses.  Know the limitations of claims because claims is what people are at least costing, not necessarily the cost of.  That is the price of.

Where do you think we are in this transition?   Is the all-claims database really intended to serve as the proxy for now?  In the long-term, there might be a marriage between the different datasets. Then what does that look like?  That is the first question.  Then I have a second question, but I will ask this one first because I have a hard time remembering questions. 

MR. SLACKMAN: The Office of the National Coordinator put out last year its interoperability roadmap.  In there, they said the convergence or integration of administrative and claims data and EHR data is extremely important.  However, it is out of scope of the roadmap.

We have commented numerous times to HHS and ONC that an interoperability roadmap should not put the convergence of administrative and claims data outside its scope.  Now, I know you are going to hear from ONC today who may talk about that.  From a plan perspective, first of all, just combining data across EHRs, everyone knows the problem.  To call it a headache is an understatement for plans.

Then second of all, even if EHR data were interoperable, integrating them with claims data, lining up the providers, diagnoses, et cetera, is really tough.  We hope that this is an issue that the NCVHS takes up since your recommendations, your advice to HHS could be very valuable in advancing this important issue of converging and integrating these data. 

MS. ENGLISH: What we are doing is we are looking at health care data, most extremely complex and just the process of gathering all the information from the APCD has been a fairly big effort. But the reality is that we also have to deal with clinical information, social determinants, benefit plan design, et cetera.  One of the key issues is that we are looking at, and we have initiatives in place, with several behavioral health clinics on how can we go ahead and merge.

There are some big initiatives in Colorado regarding integration of primary care.  The challenge you run into is the lack of standardization on the clinical side.  What we normally get to that we can at least get some consistency is the numerators versus the denominators, which really is not the ultimate goal.

Again, I think pilots are going to need to be put in place, standardization is going to have to really move at a much faster place before it can truly be put on that overall load. But there are initiatives from a reporting perspective and where we are trying to merge the retrospective claims information with more current clinical information to truly get a better picture.

MS. SCHIEL: Trying to do the same thing with very select quality measures as a pilot.

DR. RIPPEN: Or even going back to the second kind of theme of the questions which is centralized or decentralized. Actually you may want to consider, in South Carolina, Health Science South Carolina, actually has the clinical data and actually works with the all-claims database for specific studies.  It isn’t all together.  It is based on appropriate use.  Again, the more data, obviously the more nuance and challenging.  Besides, you will see it in the standards.

The second one then goes to centralized versus distributed, having lived in both words. Centralized, three different systems, four different systems, how do you standardize it?  What is the data model, but then, also the distributed.  Ultimately, what I have found, and I would like your views on it.  As it relates to cost and transparency, I mean ultimately it is about trust.  Everyone has to deal with the standards.

Where you think kind of the challenges are with all of them because whether it is distributed or not, you have to deal with the cost.  You have to deal with the standardization.  You have to deal with the governance challenges.  What do you think the big differences are between a distributor versus a consolidated centralized as it relates to that?

MR. SLACKMAN: Let me handle the trust issue because I think the FDA sentinel and the CMS edge servers illustrate different approaches to trust and a distributed environment.  In the FDA sentinel, the FDA develops the software for whatever analytic question it is looking at.  It gives the software to the issuers.  They are responsible for running the software against their data that are standardized.  They send the results back to the FDA.

In the risk adjustment model, CMS did not trust issuers to run their own software.  In fact, issuers didn’t necessarily trust other issuers to run software.  What CMS does in this environment is CMS goes into the edge server.  CMS runs the software.  It does the analysis.  It doesn’t rely on the data provider.  Then CMS extracts the results.  So two different approaches to handling trust or lack thereof.

MS. KLOSS: I have two specific questions.  For Ana and Dab, given your mandate, you seem like you would be in a perfect position to sort of amp up the consumer education about health data.  I was wondering if you could comment on what you have been able to do so far in that regard.  I think we are all concerned that we want to make sure consumers are well engaged with their data.  You seem well-positioned to be able to advance that cause.

MS. SCHIEL: I think our publications are geared towards.  First of all, we produce reports that we want to make sure that consumers can understand them.  We do have a consumer website, transparency website that we are working on.

I think we engage with folks like Health Care for All, an advocacy group for the citizens of the commonwealth to weigh in on what our priorities are and what kind of publications we are going to release.  We get good feedback from them and support.  We will continue to engage them and use them, especially in the development of our consumer website. That is where we are planning on going right now. 

MS. ENGLISH: From a Colorado perspective, we have done focus groups. We have developed the websites based on the focus groups.  We have initiated claim talk type of blogs that are geared more towards the consumers.  So much of what we do in the health care industry is geared towards others that are in the industry, which the language is very different to the general consumer.

One of the challenges that we run from a CIVHC perspective and within Colorado, and I think industry-wide, is where do consumers normally go.  They are used to going to their provider, to their physician, to the hospital.  They don’t know what the numbers are, as well.  What we have done is we have taken a multi-faceted approach and more regarding spot analyses for the consumer, still maintaining the website.

We are now working with employers to push our data onto their sites.  We are working with digital vendors in which the creation of our public data, so that it gets pushed into their applications versus assuming everybody is going to come to CoMedPrice.org.  There is a number of different things that we are doing to try to push the information out.  But in order to truly get to price transparency, increasing consumer awareness, we have to also get to these other stakeholder groups that touch the consumer.  Again, multi-faceted approach, it is not just about the website.  It is how we are getting the information out.

MS. KLOSS: I have a quick question.  I guess it the same vein.  There are some best practices emerging.  I am particularly interested in Eric’s comments about data governance, and how those best practices and learnings get accumulated and shared, so everybody isn’t reinventing this important wheel.

MR. BARRETTE: The biggest way they are accumulated is through experience.  The first payer we interacted with probably was a much different experience than the third.  But more so from the researcher side, the first research license we negotiated with the university was a completely different experience than the 15th.  We have started to describe.

There is a journal article on AJMC describing sort of ideas, kind of concepts of data governance.  I am not sure the best way to disseminate some of this.  I am not saying this is a proprietary business information.  It is just there is not a good avenue to get it out there.  Hopefully, events like this and other types of conferences or opportunities to discuss these issues will send it around.

MS. GOSS: I have a couple of things I wanted to ask from a point of just ensuring that I am hearing you correctly.  I am focused on the distributed versus centralized model.  Having experience in both worlds, as well, and I appreciate my other committee members’ questions related to this.  I want to make sure I hone in on a few things.

Is anybody aware of a state that is currently doing a distributed APCD model?  I didn’t hear one.  Okay.  I did hear that there is some federal examples related to distributed models.  I think I read in your testimony, Joel, that the experience with your blues plans and enormous cash claims data.  When you use the word, cash, I often think centralized. Are you centralizing that data?  Or are you using a distributed model?

MR. SLACKMAN: The current data that Blue Cross/Blue Shield has is a centralized data model.  If you talk to the architects, if you ask them if you had to do it again, what if you used a distributed approach, informally they may have said yes.  But one of our main points in talking about the differences between distributed and centralized is in the political context.  Think about a state that wants to harness the value of claims data, doesn’t have an APCD yet.  It faces a choice of centralizing data or distributing, or using a distributed approach. 

I think there are pros and cons on the technical side. But I think really complicating this, and I will go back to the Gobeille decision, states face enormous impediments right now.  Certainly they face a major impediment to getting self-funded data.  I know, Colorado, this is the year you are supposed to get self-funded data.  Is that right?  Most health plans are ERISA plans.  ERISA self-funded is anywhere from 50 to 60 percent of the claims.  It is a huge issue. 

Why would a state invest money in building the information for centralized warehouse?  We are talking about a lot of money, if there is legal uncertainty about whether it is going to have access to any commercial claims data.

MS. GOSS: I think that the money is one aspect.  But I think having forged consensus around health information exchange models and also having lived within the HIPAA world at a detailed level, there are a lot of aspects related to why decisions are made the way they are within the state’s arena related to governance, trust, et cetera.  I am trying to also understand sort of the current landscape.  I think I heard HCCI, Eric, I think you supported this idea of a distributed model.

MR. BARRETTE: I wouldn’t characterize it that way.  I think we see value in rather than each state collecting data from every insurer, and let’s assume now that all the insurers voluntarily participate in every state, that is 50 states plus DC, maybe Puerto Rico, collecting data from the insurers or the payers, I should say, 50 sometimes. 

Alternatively, we are getting all of this data from every state in one place.  In some ways, if there is a distributive model where everything goes the same way to one location, and then the last step, there is just a marginal cost of combining it all into one.  Rather than doing that 50 times, just do it once.  Put it in one place.  It is accessible.

MS. GOSS: I noted you had a centralized model.  You are a proponent of centralizing the data analytics, but you are interested in seeing a sort of an edge server model for a landing zone, so that data can be pulled when it is needed by the centralized analytic hub?

MR. BARRETTE: I am not thinking in terms of the technical specifications of the architecture.  If we think of how the data currently exists in our enclave at NORC, if a researcher from Colorado wants to access Colorado data only from all of our insurers and our holdings, we can create a space on the enclave.  We can create a table of Colorado data. They can access that data. 

But we haven’t had to collect all of the Colorado data separately from all of the Massachusetts data, separately from all of the Minnesota data.  We got one data feed from United, one from Aetna, one from Humana.  It all got put in one place. Then we can just disaggregate it from there.

MS. ENGLISH: If I can add to that, through standardization, we have on the medical side and the pharmaceutical side, we have over 40 submitters that are submitting.  If they have at least 1000 covered lives within the state of Colorado, they are to submit claims to the Colorado all-payer claims database.  That is medical, dental, pharmacy, eligibility files, as well as provider detail.

The key here is when you are looking at up just bringing it in and managing the data, the work that goes into ensuring that you are matching up providers, that is called local knowledge input.  A big differentiator between managing it with a focus toward how the data is then going to be distributed and used to those that are going to be affecting the change, that is those providers, those payers that are managing within those specific markets.

I can tell you there are 800 lines for one ambulatory that do not just match up according to very specific criteria.  There is a lot of work and effort to make that information usable within each of the various markets. When you are dealing in the KIA’s, we have got to understand the makeup within those particular regions.  I am not even just saying at a state perspective, but within each of those pockets because they are very different.

Half a dozen employers may represent X percent within the metropolitan areas, but they may represent only 5 percent within these smaller regions, which means you are not looking at these smaller regions because they are a different makeup and a different set of players.  That is one of the things of knowing and understanding where they say health care is local.

Yes, we do have a centralized database.  Massachusetts and the other states do have centralized database, but how we are managing it is the only way to be successful and to make it useable through the various stakeholder group is really understanding those nuances and working with it very heavily.

MS. GOSS: We heard Helga ask the question about the intersection of the clinical and administrative financial data.  They have very different standards, very different purposes. But they really ultimately are going to need to come together.  When you look at what data quality, data structure and transport evolution we are going through within the EHR realm, we can’t lose sight of the various models for health information exchange, which you depending upon the culture, the politics, the money, the trust, et cetera, there are buckets.

There are states that centralized.  They are centralized inside government.  There are states that are centralized that are in a non-profit private sector.  There are states that are federated inside Medicaid and other private sector organizations that do that federated governance and technical architecture efforts.

I think we are going to have to grapple with the fact that we have got historical architectures from HIPAA with modern technology, a lot more standardization and also need for respect for the local community’s culture.  I think we have got some things we are going to need to discuss on that.

DR. PHILLIPS: This is just such a curious thing to me that this distributed versus state-based argument got set up.  I don’t actually think it is real.  There are very specific needs and use cases and complexities at the state level.  That is where most of health insurance innovation and health demonstrations are taking place. 

There is such a utility (coughing).  Ana, I thought your points about the understanding about what is happening on the ground and the ability to improve the value of the data is incredible.

The distributed model has its own important value and benefits, and certainly for supporting research. Rather than pitting them against each other, why can’t they be supportive?  I think if there is a distributive model that can offer state extracts, they offer the opportunity to tell you more about physicians than your claims allow you because they look across all claims.  They also have the ability to look at your claims data and enhance its value because they have on the ground experience and knowledge about your data you just can’t have.

I don’t know why there can’t be a partnership of distributed and state-based efforts.  Eric, I really appreciate it, but in this day and age, setting up a table is, boy, we can go much further than that in terms of data extracts.  I don’t mean to pick on you, but in a cloud-based environment, we can do so much more with sharing data to enhance value back to you than a table would represent.  I would just encourage you all to think about how these two things can play together.

MR. BARRETTE: I will clarify.  When I said table, I am thinking in terms of the way table would be set up in SQL.  It would essentially be data.  It would be a, we will say, data subset.  I will add I completely agree that the institutional and local knowledge within not even a state, but within a county or a region, helps a lot when working with the data.  By taking the data into a centralized location, we are not saying that sort of local knowledge isn’t necessary, there are just efficiencies to be gained by taking it in once and then letting people that have the local knowledge take it from there, rather than putting together 50 times.

DR. PHILLIPS: I am agreeing with you.  I think the distributed model could support the states, but still have the to take advantage of your efficiencies.

DR. SLACKMAN: I think your points are well taken.  You remind me, and this gets back to Dr. O’Grady’s question about what is the ROI in effect, of these.  Really, one of the problems that we have seen with APCDs across the country is that they are often not well-defined use cases.  Now, every market is different.  Massachusetts has statewide health reform.  Colorado is thinking about it.  Their political cultures have a very different need for data from Kansas, which has an all-payer claims database and has not done anything with it or Tennessee.

I would think that whether we are talking about state or federal policymakers who are looking at imposing costs on the private sector, whether it is through collecting data in a centralized fashion or a distributed fashion, have well-articulated use cases before they start collecting the data or gaining access to the data.

DR. MAYS: I think probably about 90 percent of what I wanted to ask may have been covered by Alix and Helga.  I want to make sure I understand something.  The arguments for centralized versus distributive have been really put on the back of research. Much of what we as researchers are trying to do is looking at a particular question. 

When we have to do this research, it is often we are funded to be able to do certain things and not others.  The models really differ as to where the work push goes to some extent.  I am really trying to understand from the research perspective in a centralized model.  There are a lot of things that you can specify, that you would ask for, that would push out the innovation to some extent in terms of trying to do standardization.  Help me understand in the distributive model how that would be done. 

DR. SLACKMAN:  Very simply, and as Dr. Platt has said, there is essentially no difference between a centralized or distributed in being able to conduct research. 

DR. MAYS: No, not in conducting the research. I am actually talking about the work that is to be done before you kind of have to do that research in terms of standardization of variables.

MR. SLACKMAN: Whether the data are made available in a distributive federated approach or a centralized approach, it is essential that the data from across multiple payers be standardized in a meaningful way.  That is why I talked coming up with a common data dictionary, so that you are comparing apples and apples.  That absolutely has to happen ahead of time. 

When Blue Cross/Blue Shield several years ago brought plans together to create its first national data repository, we are not a command to control association.  It took a long time to get plans to agree that they would all define, say, maternities and the infant the same way.  Some are counting as two admissions, some as one. 

That is just one example where as they say garbage in, garbage out.  You as a researcher depend on the integrity of the data.  Whether we are talking a distributed or a centralized model, there has to be a very good thorough process of guaranteeing the integrity of the data.  Massachusetts, you would say, it has been a long process.

MS. SCHIEL: It is producing reports.  You show it, and you see something that could be a programmer on the health plan side that just is a new programmer and miscoded how they are classifying small groups versus large groups, just a simple error.  Then we catch it on our routine edits.  We can see right away, oops, this is going to affect our reporting.  We go back to the payer and say, great, we are going to correct that for you.

It really is I think the advantage of producing reports, working with the payers and then producing a report that the payers also find useful and informative.  We are adding maps on coverage across the region in Massachusetts of where the health plans have their largest coverage, heat maps and things like that.  That is based on what the payers have asked us to do.  It is using the data and actually saying, you thought you did this right.  But there is something very different about your data that really threw this off.

MR. SLACKMAN: Let me just add that the same approach, FDA, CMS have programs that run edits to see whether, in fact, the data that are on the distributed platform meet the standards. Conceptually, whether a third-party is doing it in a centralized approach or in a distributed approach —

MS. HINES: Would it be fair to say that regardless of approach, to Bob’s point, regardless of whether it is distributed or centralized, having standards that everyone agrees upon consistently would serve all ends?

MS. LOVE: Which take us into later discussion today.  I think that is one area we all agree on.

MS. GOSS: Can I make one point, though?  The standards are key, but what I heard was having a neutral intermediary that can do that kind of data analysis and can really drive the governance is really the crux of the issue here. 

MR. SLACKMAN: That is why I referred to Mark McClellan.  There are questions about whether a distributed approach or a centralized approach might have advantages for data integrity.  I mean, once the data go to a third party, if you find a problem, it is hard to do the root analysis to figure out what was going on.

MS. LOVE: In the interest of time, this has been a fabulous discussion.  We could really go for another hour and a half without breathing.  I feel some duty to the other speakers and also some bio break maybe, ten minutes or so.

What I am taking away from this is the passion of people that care about data and care about compiling data.  I think we embrace all approaches.  I really predict someday we will have this hybrid approach.  Things will evolve.  But I think the passion in this room is great. I appreciate your comments. Thank you.

DR. SUAREZ: We are going to take 10 minutes of break while we transition to the new panel.  We will come back.

(Brief recess.)

Agenda Item: Panel 1B: Data Suppliers, Users – Policy Consideration

DR. SUAREZ: I think we are going to start again.  Thank you for the flexibility to all our panelists. I know we have someone that is going to be on the phone and joining.  I also know that for the next panel, which was scheduled to start at 11:00, we have one person that has to start at 11:00.  We are going to bring in her into this panel that we are going to be listening to.  She is Doris Lotz.  I don’t know if she is on the phone.  She is not on the phone yet.  She needs to be on at 11:00.  We will bring her into this panel. 

I think I am going to turn it to Denise, who is going to get us started with this next panel.

MS. LOVE: So, to continue the discussion flowing from the morning that talked about structure and governance and some of those issues, now we are talking about the policy considerations and the states and the plans and others with all-payer claims databases from the perspective of the state, the plan, the national database and employers and an ACO representative, and their perspectives as stakeholders and submitters and users of all-payer claims databases.

With no further ado, we will start out with Ben Steffen with the Maryland Health Cost Commission. I may have said that wrong.  MHCC is how I know it. I will let you introduce and start.  Then just say a brief introduction where you are from for the people on the phone.  We will just go down the line.  We do have a little complication because our conversation ran over.

At 11:00, we are going to do a stop.  We have a medical director calling in at 11:00 for the next panel.  We have to accommodate.

DR. SUAREZ: Well, I think what I said is we are bringing Doris Lotz into this panel at 11:00.  She will be part of this panel. We will have the discussion.

MS. LOVE: So, no further ado, Ben.

DR. STEFFEN: Thank you.  My name is Ben Steffen.  I am executive director of the Maryland Health Care Commission.  The Maryland Health Care Commission has a range of authorities in Maryland government, including collecting broadly information on cost and quality of which the APCD is one key element.  We also report on the cost and quality of health plans, hospitals and long-term are facilities. 

We have responsibility for diffusing the adoption of health information technology.  We have a historic key role typically in state government, which is developing, planning models for health care facilities in the state.  You know that under the guise of certificate of need regulation.  We have somewhat of a different perspective.  We are not solely a data organization.  We are also a planning and policy organization.  It is probably going to come as not too much of a surprise that I am going to talk about that.

First, I want to sort of level set very quickly that when we talk about APCDs, we are talking about a number of transactions, not only claims.  Most APCDs also include eligibility files.  We are all moving in these directions, talking about non-claim-based reimbursement transactions.  No state that I am aware of has really come up with a concept of a standard.  But we all recognize that the claims, as I think Nile Brendan said, the exhaust fumes of the health care system are increasingly becoming less comprehensive in terms of the scope of reimbursement that they cover today.

Importantly, eligibility is key to all of the APCDs that I am aware of because it allows you to generate the denominators on which measures such as total cost of care can be calculated.  Really, they are essential also for capturing demographics and characteristics of the products themselves currently.

There are moves afoot for developing plan-specific transactions.  We know that certain things like levels of deductibles, co-pays, co-insurance, et cetera are also important.  But those are TBD as we move forward, but very important.

The Maryland Health Care Commission’s APCD really got started in the last century, in 1996, with some health care reform that was passed at that point.  It got started as a result of a vision that we would embark on.  Physician rate setting, just as we had a long-standing hospital rate-setting system. 

The focus was initially on physician services, with the vision that we would, with the help of the federal government as they rapidly developed a national patient identifier, be able to link with the hospital data that had been longstanding collected in the state.  We are still waiting for prompt action on the part of the federal government of the part of the national patient identifier.  But let me tell you this.  We have moved ahead in other directions.

I think the key point I would want to make here is that all of the earlier discussion has important points.  But as we move through this, states are focusing on initiatives specific to their state.  I would strongly recommend that you take that into consideration.

The recent Supreme Court decision, regardless of where you are, certainly throws a wrench into some of the initiatives that we have in Maryland, either on your way today or planned. The focus has been on our database for policy, great and small, good and bad.  There have been some instances over the last decade where I have wrung my hands and said that I really wish we didn’t have to use it for this purpose.  But at the same time, the database has provided insight to policymakers.  I will highlight just a few.

Certainly, the Maryland Health Care Commission modeling work that had been done at MedPAC on payment adequacy has done, I think, a pretty good job in characterizing the level of reimbursement, particularly in the provider community across specialties.  The APCD, as it contains detailed information on amounts paid, allows us to characterize reimbursement levels for different specialties and reimbursement levels across payers large and small.  It is particularly very useful for demonstrating the level of what I would characterize as payment inadequacy for our primary care providers in the state of Maryland.  Several years ago, we pegged it at 92 percent of Medicare reimbursement.  Certainly, many would argue Medicare is not generous.  By contrast, some private payers are considerably below that.

We have also done things that I think really shine the light on what is happening in health care on the small scale.  We have looked at things like air ambulance reimbursement.  We have examined certain variations and payments for different types of mental health providers within the state.  We have tried to explain their differences.

We have also more broadly looked at issues and particularly we have used the APCD to further a multi-payer PCMH program.  It was the dataset that was used for attribution on this multi-payer model.  It was also used for developing shared savings calculations that were then applied back to the private payers. They were suspicious that the APCD could be used for shared savings. We proved the case that we could.  This was a pilot program not taken to scale, but it definitely proved the case that was feasible.

We also, I think in terms of what was mentioned earlier, the APCD was used as we developed our state health benefit exchange.  Particularly it was used to assess the state’s policy for reinsurance.  It was considered, but ultimately rejected as a source for the readjustment calculations.  The state ultimately opted to follow the federal model for risk adjustment.

I would quibble a bit with the statement made by my colleague from Blue Cross/Blue Shield about the model that (indiscernible) was using for risk adjustment.  That is a single purpose dataset for risk adjustment.  I think what APCDs are broadly focused on are a range of issues, a data system that serve multiple public health costs, public policy and consumer applications.  While I think what CMS is doing is very important, it has not yet been demonstrated and had broad applications.

Certainly as we move forward, and many of you know that we have a hospital rate-setting system in Maryland, in 2014, we embarked on a new hospital payment model establishing global budgets under which all hospitals were to operate.  All hospitals were to be under that budget by 2016.  The state accomplished that by the end of 2014.  We had also set up a number of very tough criteria for us to keep the new contract.  Remember historically, Maryland was protected under federal law.  We were exempted from PPS and IPPS.  We gave up that authority and were rolled under a new demonstration.

We are currently operating within the parameters of that model.  But the plan is, going forward, that the entire state and health care economy, $56 to $58 billion would be brought under global budgets in the next version of the model.  The APCD is envisioned to be a key element of our monitoring efforts.  Not now.  We are aware of total outpatient spending.

But we will be moving rapidly to try to bring the APCD up to date in terms of timeliness of collection, in terms of completeness, so that we can use it as a monitoring tool as we move into version 2 of the model.  We have been successful so far, but we have a lot of work to do.  The data systems are going to have to be developed in parallel. It includes not only the APCD, but other administrative and clinical data systems going forward.

We are not the leaders in terms of making information available to the consumers and other stakeholders.  But with the support of the federal government, particularly CCIIO, we have used our CCIIO advanced rate review grant to develop additional consumer tools and to work with other stakeholders, particularly our insurance administration in developing the APCD for rate review.

I would applaud my colleagues from Massachusetts who have moved forward.  I don’t think it was emphasized enough.  Very far in making the APCD the trusted source for insurance rate review in the states, something the insurers in that state have bought into.  It rarely represents a significant administrative savings to stakeholders, as well as complete information on which the insurance administration can assess the legitimacy of insurance premium increases.

Certainly an enormous benefit, and I wouldn’t want to discount that.  Certainly something that as long as insurance regulation is the authority of a state government, we need to further the ability of those state governments to have the data tools to effectively do their work if they so choose.

Lastly, I think the discussion on seeing the APCD as part of the data architecture is key.  I have always seen the NCVHS as a forward-looking entity.  We need for you to do that today.  Not only would we hope that you would make recommendations on how the Department of Labor would move forward with a standard for collection of information from self-insured ERISA plans, but you would think about the APCDs fit in the broader data architecture.

The Health Care Commission is fortunate in that we deal both in the claim world and in the clinical world through our responsibility for working with our state HIE.  I think those landscapes need to converge sooner rather than later.  There are enormous challenges.  Data standards don’t align.

We are taking some experimental steps in working with our administrative networks that act as the intermediaries between practices and hospitals and other health care providers.  In capturing those raw, unadjudicated transactions and directing them to our HIE’s encounter notification service to provide early alerts to providers on what is happening with patients for which they have responsibility, but do not necessarily have great vision or great control.  I think there are challenges. I know that Dr. Suarez has significant interest in that.  We have worked with him over the years.  I think there is an important need to do that.

To make that happen, we need to align data standards. But we also need to go back to that nagging issue of a patient identifier.  In Maryland, we have enrolled patients with the action of the federal government and are moving forward, working with our HIE in developing the common patient identifier encrypted, but nonetheless uncommon, so that ultimately, we are well positioned to link clinical and administrative data.

First off and foremost for the support of clinicians in the delivery of care, but secondly for some of the secondary uses that we have talked about earlier.  I think the idea of how we do this in terms of collection is important.  I am definitely a person that wants to save money.  We have a guaranteed budget to do this.  But it is not as ample as it might be. 

Certainly the work with the federal government, the CCIIO funding, be one example is very important.  We would like to see that expanded and continued.  But I think states have a key role.  Don’t forget that insurance companies themselves are not a single entity.  In our case, in Maryland alone, CareFirst sends us data feeds from four data submitters. The idea that Aetna, Cigna or any of the other data plan would be submitting a single file to a national entity.

At this stage, it is probably unrealistic and certainly misleading our own submitters in the state were they to submit nationally would be submitting three, four or five submissions because their adjudication engines are largely different.  That stems from the fact that adjudication is not a priority big investment area for health plans.  They, too, have major challenges in reengineering some of those efforts. I think many of the points were very valid. But simply centralizing data submission would not eliminate the fact that we are still dealing with multiple transactions.

Lastly, keep in mind that most plans have multiple data submission requirements, not only to states but to self-insured employers.  We would perhaps limit some of the transaction submission requirements. Certainly by going to one standard, we would not restrict them to 50 only.  It is important, I think, to keep in mind. But I certainly understand the challenges that health plans face.  Maryland, like other states, really want to work for the most effective, most efficient solution going forward.  We want to put most of our money that we have not towards collection, but towards dissemination, helping the stakeholders use this information in effective ways in improving the health care system.  Thank you very much.

MS. LOVE: Thank you, Ben.  This is the hardest thing to do, but we are going to have to move quickly because we have someone calling in at 11:00.  We may have to do some micro adjustments as we go through.  Just be mindful of the time because we want the robust dialogue at the end, as well. 

MS. HARRINGTON: Good morning.  I am going to speak very quickly because I am going to try and get it all out in five minutes.  I am from the state of Maine.  Maine is a state where we were awarded the SIM grant.  We have a multi-payer patient-centered medical home demonstration going on that includes the Medicare beneficiaries.  We were awarded a couple of years ago two societal grants to help us expand on our cost and quality information that we put out for the public.  We have a lot going on.  We have a long history of a public-private collaboration in these type of reform initiatives. 

I think what has become clear over the years is that if you don’t have the data to assess and measure how you are doing, how do you how to proceed?  How do you know where to course correct?  How do you know what to change in your market? 

The Maine Health Data Organization, actually we were, I am happy to say, the first APCD in the country.  As I said, we have been in operation collecting claims data since 2003.  We have been using that data to measure and drive change. 

We are governed by a 21-member board.  We are a state agency, an independent state agency.  That is why there is a board.  I serve at the pleasure of that board, not at the pleasure of the administration or governor.

The agency by statute is responsible for creating and maintaining a useful, objective, reliable and comprehensive health information database that is used to improve the health care quality for Maine people and to promote transparency of the cost and quality of health care in the state of Maine by procedure, by payer, by facility.  Our statute gives us a broad set of responsibilities. 

What I would like to leave you with are sort of these three aspects. One, the structure of the Maine APCD as a public model provides, in our opinion, the greatest amount of transparency, accountability, comprehensiveness, as well as fair and equal access to our data for all users. 

The MHDO is using used to support all kinds of initiatives to support the triple aim.  We believe that health care is local.  We believe that the collection and the use of data is local.  I want to give you quick example.  One of the things that I think is critically important in this work is you have got to be working with your data users.  You need to understand from the users what do they need.  Collecting data for the sake of collecting data, nobody has time for it.  We need to focus.  We need to know exactly what we need to collect to advance some of these things that we are doing.  It became very clear a couple of years ago that the users, with our patient-centered medical home demonstration, were developing these practice reports for the primary care doctors, using our data so they can see the rate of readmissions, the rate of EDUs.  The docs came back in the pilot and said, you know what?  That information is great.  But here is the thing, I need to know who my frequent fliers are in the ED.  Even though they have all this technology in their offices, they don’t use it.  They don’t know how to coordinate and put the information, take the data and turn it into information.  They want that from us.  So we went to our legislature.  We got the support.  The bottom line is we passed a bill.  Nobody said we could do it and we did it.  We passed a bill with a unanimous vote out of our committee, our Health and Human Services Committee.  We can now release individually identifiable data under very defined purposes.  We are not covered entities, so we don’t necessarily have to comply with HIPAA.  Although we have said we will use that as best practice.  Now we can produce reports for our providers at the individual identified level.  Individuals have the right to opt out.  There is a whole lot that goes under this.  But the fact of the matter is, it is local.  It is a local issue.  It was a local solution. 

The other thing just to underscore is the importance of our public reporting.  You have heard this morning, and you will probably continue to hear, how comprehensive these data.  We collect 97 percent of all our claims data in the state of Maine.  It is an all-payer claim database in the sense that we have Medicare, Medicaid and our commercial data. 

How that data is being used, again, to not just support some of the initiatives that I just talked about in terms of SIM and patient-centered medical home.  If you have got my testimony in front of you, on page four, there is a list.  This is just a quick overview of some of the uses.  We have over a billion health care records.  Every month, that number continues to grow.  As I said, we have been releasing data for over 10 years.

Some of the organizations, they are not just in the state of Maine, but some of the national groups that are using our data, for example, is the CDC.  One of the things that we implemented in our state, and I now know I have gone over five minutes, but just quickly, is an opioid prior auth program.  We think locally, we have seen based on the data some improvements in the use of opioids for a Medicaid population.

The CDC wants to do this national study to look at this to determine whether or not this might be best practice that can be shared across the country.  They are using our data to do that type of analysis. The main office of the attorney general uses our data to look at mergers, with anti-trust laws that are going on or anti-trust issues.  They use our claims data to make the assessments when they are being asked to look at those things.

Then some of our largest health systems, two of our largest health systems in the state use our data for their planning efforts. They have spent millions of their own dollars to build the infrastructure, to take in our data and to coordinate it and use it with other datasets that they have to help them in the work that they do in terms of planning, in terms of looking at market needs, looking at their costs.  Are they competitive in the market?

Lastly, public reporting, so we have this new website.  In the very back of my testimony, there is a brochure.  Just look at the very last page.  This gives you a visual of this website that we were able to expand on with again the support of CCIIO, where we convened a consumer advisory group.  They helped us build this.

We are reporting for over 200 procedures.  We are using a grouper.  The MEG grouper we were using for 11 of the inpatient procedures that we report, like knee replacements, colonoscopies.  You can get information on what the cost is by facility, by procedure, by payer. For over 200 procedures, that website is going to continue to grow.  But for the data that we have in the APCD, we would not be able to create this type of information. 

I will just close with what I think we absolutely believe in our market.  Bottom line is we think, and Arthur Nelson said this, the price of light is less than the cost of darkness.  We absolutely believe that.  The stakeholders have spent millions of dollars to invest in our infrastructure.  It is because there is an ROI on that investment.  With that, I will close.  I am sorry, I did go over five minutes.

MS. LOVE: Bernie, you are up.

MS. INSKEEP: My name is Bernie Inskeep.  I am the program director for all-payer claims database activities at United Health Group.  I am pleased to be with you today.  Thank you for inviting me.

For the questions that we were provided, the first two questions, examples of benefits and benefits to the state and public health, I think I am going to skip over those.  They are in my written testimony.  They are rather short as compared to my list of issues. I think that the states have really addressed those quite well.

One of the things that I wanted to talk about and really stress is the question most significant issues in implementing claims-based databases and APCDs.  My first two bullets are both the most significant issue because they were tied for first. I will talk about the first first one.

It really starts prior to implementation.  What we do not see and what we see a lack thereof is really a pilot testing process to confirm that objectives can be met with the data collection.  We have certain APCDs.  We have one that is six years old, I believe, and is on version seven.  They just keep tweaking in order to meet objectives. 

Many times, the objectives can be vague.  The approach that uses a use case scenario could really aid both in the data submission guide development and the testing.  But it could really reduce the multiple revisions.  It could help the data collection process.  It would ultimately benefit the state and the APCD.

The other most significant issue implementing all-payer claims databases relates to the variability of the format, the process and the seemingly limited understanding of health plan data, along with the late involvement of health plans or actually no involvement while developing data submission guides and requirements.

Some states, we have been involved early and regularly.  We greatly appreciate that and thank those states. Then in other states, it is very authoritarian and presented to us as a finished product.  That makes it very challenging for those items that we just don’t have.

There are reasons that we have things like lack of identifiers.  Sometimes the expectation that the data and the health plans maintain every piece of data versus the actual data that is submitted on a claim and retained are very different. If you think about each state having their own clean claim laws, they are between five and seven fields.  We are held to that standard to pay the claim on a timely manner. 

From a business perspective, and to be a compliant insurer within that state, we have to pay those claims.  But then the state would like us to reach 100 percent on a certain field that is not part of a clean claim. 

The other thing that I wanted to bring up in terms of variability is that health plans primarily retain data in reporting data warehouse.  We generally retain data according to HIPAA minimum standards for the purposes of our business operations as a health insurer.  The expectation that insurers have every field on a claim at 100 percent is a somewhat dubious assumption.  It is something that we have concerted conversations on a regular basis with many states.

In asking the next question of how APCDs are supported for the business and sustainability model, we have heard a lot about the ongoing SIM and CCIIO grants, and how the federal government is helping to help fund those, as well as states obtaining their own funding either through budgetary purposes or sales of data or the other activities that are accomplished.

I just wanted to put forth the fact that health plans spend millions of dollars complying with this.  In terms of a sustainability model, this cost particularly national payers millions of dollars. So we have just a significant concern as a data submitter that is in multiple states of the ongoing and escalated costs that are a result with a lack of a standard today. 

Right now, we have increased our staffing.  We have created a process to support these data.  But these data that we are trying to support don’t necessarily support the functions of health plan operations.  Again, that is another cost that we are bearing or, if you would like to think about it, contributing to this effort.

I would just want to lay an idea out there. As SIM grants, CCIIO grants and other federal grants are considered, I would like to just suggest that maybe those be considered with the adoption of a standard.  What we don’t see today are states that actually buy into the standard.  We have every state talking about the standard.  We have included all of the fields that are in the standard.  Yet, we have 15 states and 15 different standards.  Just a little bit of incentive may help that.

In terms of a technical challenge, one of the most significant technical challenges for us is really both the structure of the files, the formatting of the files, the different submission types and then the file requirements.  UHG currently submits over 2700 files in 15 states.  Each state has a different format.  Each data submission guide is unique.  Each state requires a ground-up build of these files in every single new state, no exceptions.

Then in states that have a provider file of all of the fields that they are collecting, only 10 percent are common between all of the states. Then the variability of the file QA process, it can create an enormous burden of reanalyzing, repolling, rechecking and resubmitting data sometimes for years after the original due date. 

We have one state in particular that is continuing to QA data from 2014, which we thought was in past status.  We are currently looking at resubmission of 2014 and 2015, which are in two different formats, which again is technically challenging and difficult. 

Most states update their data submission guides every year.  Some states actually update them multiple times a year. All of these efforts are out of sync with each other.  Everyone is on their own timeline.  They have their own programs and their own funding and their policies and procedures. From a health plan perspective, it makes our programming and our operations and our ability to comply more and more challenging.  It also creates a potential for decreased quality and inefficiency.

Then the single most difficult format is really the PACDR X12 format. This format was recently developed.  It has been adopted in one state.  It takes two transactions, an 837 and an 835.  It kind of gloms them together into something different. 

Neither of these transactions are typically retained by a health plan.  We don’t have a bunch of 837 transactions sitting there.  We don’t have a bunch of 835 transactions sitting there.  We can just throw them together and call it done. 

Our data comes in as a three-dimensional object.  It is transformed into a flat file.  Think of this piece of paper.  Then what we have to do is do origami and try to make it into a three-dimensional object again.  This is really challenging for us.  The programming time is very elongated. It is really fraught with a lot of challenges.

It also creates a really large file.  It adds expense not only to payers, but it also adds expense to vendors. Then the size of the file prevents transmission if there are too many records. Then we have got to parse up the files because they are too big. 

Also the X12 transaction creates the need for the receiving vendor to deconstruct those transactions in order to store them in order for somebody to use them.  We are going from flat file to three-dimensional object, back to flat file.  I think hopefully I have made my point.

The X12 process also adds a transaction with is a response file.  So health plans have to build this mechanism to receive this transaction that we don’t normally receive.  It is just fraught with a lot of challenges for health plans.

Then the role of claims-based databases including APCDs for the reformed health care system, I think as a practical matter, as APCDs are currently configured today, we see that they are somewhat more suited to support some risk adjustment activity models, established by reform and, to a limited extent, some review of rate filings again dependent on the population in the APCD.  If you have the wrong population, and you are trying to do rate review on rates that are approved by another state, it will skew your data.

So health plans, like our health plan, we are currently working with states.  We are trying to assist in collecting alternative payment model costs.  Since many of the payments are paid outside of the claim processing system, they are not standard fee for service.  We can’t tack a field onto the technical specifications today.  

What we have to do is actually have a supplemental report.  Then some states are trying to take that supplement report and then add fee for service to that report, even though they have the fee for service in their APCD.  There is a lot of redundancy in some of this work.

I just wanted to mention that the industry should consider the electronic record datasets to be incorporated as well.  I think that was talked about very thoroughly earlier. But health care certainly is supportive of that.  I hope my perspective was helpful.  I hope it was quick enough.

MS. LOVE: We will go to Leanne.  Then we will go to Doris, who is online and has some time constraints.  Then we will come back and finish off our panel starting with Jessica.

MS. GASSAWAY: Good morning.  My name is Leanna Gassaway.  I am the senior vice president of state affairs at AHIP.  I am going to skip the AHIP description.  I think you all know who we are and who we represent. 

I want to start off the conversation by just saying that we have appreciated and value data, both clinical and claims and transactional datasets.  They do play a very important role in enhancing the quality and affordability and availability of health care.  That was the very reason that we actually asked and submitted a letter to NCVHS to hold this very hearing today.  We thought it was so critically important that we were coming up to this nexus of data technology reform that made this conversation an absolutely necessity at this juncture.  We are very happy to be here.  APCDs do play a role in that health data landscape.

I want to go through three quick things.  First, just a little background and level-setting of why we believe that growth of APCDs has happened over the last several years and why that growth is going to continue.  First of all, there is simply a need. There is a need for states that want to better understand and manage the cost and quality of the data that is being delivered in their state.  It is plain and simple.  We also support those goals and don’t see that stopping.  If anything, it is gaining more importance as we move into more and innovative payment and delivery system reform options.

The second driver is federal funding.  It has been mentioned several times. All of the various grants, whether it is SIM grants, rate review grants, beacon community program grants, Medicaid funding, targeted funding, surrounding claims analysis, all of that is spurring a lot of activity in this space. We also don’t see that slowing.

Third, that there is this movement towards value-based models of payment and delivery reform.  As those national models transition to a more value-based system in both the public and private space, there is a demand on quality related data. I think it has been mentioned already that is one of the things that is lacking in a claims database is the quality of care that is being delivered.

You know what has happened.  You don’t know how well it happened. You don’t know why it happened.  You just know that it happened.  So that needs to be recognized that we need to partner with other data sources that exist.  I think that has been mentioned, as well.

Secondly, as we look as kind of why this is all happening, we move into the challenges that have resulted from a lack of standardization.  For the 17 APCDs that exist today, our members are submitting those formats, as Bernie has articulated for United Health Care, 15 or 17 different formats on a different frequency.

The absence of standardization of both the content and the format and the use, it was resulting in two things.  One, it results in apples to umbrellas.  It is not even apples to oranges. It is apples to umbrellas comparison of the data.  From state to state, you can’t make comparisons that Minnesota had this happen and Maine had this happen and California had this happen because the data wasn’t collected in the same format.  We need to have some definitional standardization, data collection and use of standardization that there can be comparisons across state lines.

Secondly, as has been discussed, these databases are expensive.  They need to have a sustainability model built into them.  One thing that is of great concern to our health plans is not only are many of our health plans data suppliers, but they pay for the privilege of submitting that data.  We need to have a more sustainable model moving forward.  There are considerable costs that are incurred to ensure compliance with each state’s specific requirements. Just for an example, one of our large insurers reports that between the implementation of the software, the staff and the testing, it can be upwards of $500,000 per state.

If you are a small plan that is in even three states, you are looking at a considerable amount of money that is being spent just in-house to get the data to the state for their use.  That is not necessarily incorporating the fact that you then pay an assessment along with other healthcare participants to sustain that over time.

How do we move forward from here?  Our recommendations kind of come from the five questions that my third-grader has learned when he is reading, which is what, how, when, why and who.  First, what.  First and foremost, standards should be set regarding what data is collected.  We need to have again common definitions, a common understanding of what data is being collected, so that it is standardized across the various APCDs and other voluntary efforts that are out there.

Secondly, how the data is collected. We believe standards should be set that the data, whether it is aggregated including the methodology for determining data integrity, as well as to the respect for formatting used in reporting.

Third, when, similarly consideration should be given in establishing greater consistency about the frequency of data.  Is it monthly?  Is it quarterly?  How far is the reconciliation period drawn out?  How long is the claims run out, et cetera?  We need to have some frequency standardization.  That gets into the file size for many of these files.  If you go to a quarterly submission, that is a large file.  If you go to a monthly submission, it is a smaller file, but it is more often having to be pulled out of a dataset. There are pros and cons that need to be considered.

Fourth, why.  We strongly recommend that consideration be given in developing standards regarding how data may be used and who has access to it.  Many of these enabling statutes have very broad and vague terms about why the data is being collected, who can access the data and for what purposes it can be used. I think that especially as we move more into identifiable data, we need to have some very robust and candid discussions about who gets access to the data and what it is going to be used for, and the level set on that. 

Then lastly, I would be remiss if I didn’t talk about cybersecurity threats.  In this environment that we are living in, we need to have an honest discussion of standards of how the data is going to be protected and how it is going to be kept confidential.  We are working in a lot of other venues on cybersecurity issues, whether it is at the NAIC, whether it is at the federal government.

Health data is a very valuable dataset in the black market, as well as anywhere else on the planet right now.  It is even far greater in value than identity data.  It implores us as folks in the health care data set to take that very seriously and ensure that the HIPAA standards must be followed and they must be applied in essence to keep that data very secure.

As advisors to the secretary of Health and Human Services, we believe that NCVHS can play a very effective role in endorsing a common framework for data collection and reporting.  That dataset would then be incorporated in federal funding requests, in establishing and maintaining an APCD, whether it be through SIM grants, Medicaid grants, rate review cycle grants and the like.  This standard would become the standard that states and other parties that would be given federal grants would be held to as recommend by the NCVHS body.  We think you are uniquely positioned to make that happen and appreciate your time on this.  Thank you.

MS. LOVE: Is Doris on the line?  We are going to cut to the next panel ahead of time because of time constraints and Doris’ availability.  Is Doris on the phone?  Thank you, Doris.  Do you want to introduce yourself and share your comments briefly?  We are sort of in a time crunch.  We appreciate you taking time out of your day to share your testimony with the National Committee.

DR. LOTZ: My name is Dr. Doris Lotz.  I am the chief medical officer for the New Hampshire Department of Health and Human Services.  As CMO, I have a responsibility for the quality of department programs and services.  I provide clinical leadership to the state’s Medicaid program.  I have direct oversight of health care data and analytics.

New Hampshire has a mature all-payer claims database.  You have heard enough about the kind of standard applications and uses.  I think you are all impressed with those uses.

But we are restless in New Hampshire.  We just don’t want an APCD with a limited application.  We want to push the envelope.  In 2007, the New Hampshire Insurance Department was the first in the nation to implement a consumer-accessible health cost website, which draws from the APCD and which I have recently participated in incorporating quality data to better prepare providers and plans on both cost and quality. 

This moves us closer to understanding the value of health services and enabling consumer choice based on value.  In the DHHS quality shop, we are very committed to public transparent and usable data.  We built the Medicaid quality information system and its companion website. 

In addition to MQIS holding our managed care organizations publicly accountable for their quality outcomes, we allow any user, consumer, advocate, ourselves, of course the HHS, to easily dive into aggregate data, such as HEDIS, HAPs, operations data much more.  We can select certain populations or programs of interest, such as children or behavioral health services.  We can better understand the Medicaid program strengths and opportunities. 

This data is updated continuously and undergoes automatic system analysis for validity, and active and automatic surveillance against goals.  Any outliers are subsequently reviewed by staff for department action.  MQIS lets us stay on top of hundreds, maybe thousands of measures. Managing large data is not an impossible task.

This committee is interested in the next frontiers for APCD.  I would submit the following three.  The need for further integration of datasets across all payers, including Medicare.  You have heard a little bit about that already.  Greater access to data on the social determinants of health and the barriers and resources available to improve outcomes. 

We are all familiar with the drivers of health being primarily our social circumstances, behavioral patterns in our environment.  That data needs to be captured and made actionable, in addition to the claims data. 

Last, the need to continue to address the administrative burden that creates barriers to data collection and use.  You have heard a little bit about that certainly in the last couple of speakers. Fortunately, New Hampshire is beginning to address all three of these areas.

Regarding Medicare data integration, CMS is partnering with states to provide Medicare data and also provide support to integrate that data into data sets for use.  New Hampshire is participating in a CMS innovator accelerator program for Medicare/Medicaid data integration.  Working with support from CMS and their technical assistant vendor, SEI, the state is incorporating the Medicare dual eligible data into both the APCD and the MQIS system I just referred to, so that Medicare data on duals will also be transparent and accessible.

We will begin using the data immediately to develop a comprehensive beneficiary profile to better understand the population.  We are incorporating active dual eligible management into our managed care program.  Remember, you can’t manage what you don’t measure. 

There are two immediate challenges. You have heard about them already in the brief time I have been able to listen.  Data standardization and redundant data, I don’t think you have heard that yet.  First, nomenclature differ across systems, provider identifiers and the national provider identifier is insufficient to the task and member identifiers differ.  These differences need to be resolved. SEI is developing a master provider and member index for New Hampshire that will also be usable by other state Medicaid programs. 

Second, there are significant claims duplication.  Since the HHS is responsible for payment of some of these services, there is a potential for duplicates when HHS has a claim for the patient and then receives an expanded dataset.  We are working through solutions to officially resolve the issue of redundant claims. 

Regarding the capture and use of data on the social determinants of health in a public private partnership, New Hampshire Medicaid is working with Blue Cross/Blue Shield Association to present an easy to navigate website that integrates clinical data with social determinants of health and allows for augmented case management for targeted population.  The website, when completed later this summer, will identify community assets and innovations at granular level, so the health practitioners and local and regional stakeholders can affect change in their neighborhoods.

New Hampshire would have been hard pressed to develop this resource on her own as the Blue Cross/Blue Shield Association has integrated over 21 different data sets, some public, others purchased, including things like Google Data Traffic to generate drive time, NIH data on alcohol consumption, the US Department of Agriculture data on grocery stores and grocery store bar scan data to look at consumption of fruits and vegetables.  US Geologic land survey data to see where sidewalks exist and parks, and HRSA and SAMHSA data to look at primary care in mental health clinics among many other data elements.

We will layer in data about the New Hampshire Medicaid population and augment the website with various community and other programs.  The user will be able to see both where there are opportunities and resources in an intuitive and elegant Google maps interface at a remarkably detailed level. 

Regarding the administrative and monetary costs of data collection and use, as part of developing Data New Hampshire, we elicited the support of our external quality review organization to collect hybrid data, so we could report on required CMS quality measures.  Not all data, as you have heard, can be gleaned from administrative data.  Some must come from clinical resources to be actionable. 

We found that the cost of collecting data from clinical charts was about $50,000 per measure.  The cost to field a cap survey is at least $60,000.  It is very easy to spend over half a million dollars on a few data elements before you even begin to work with them.

Also for providers to pull data for each requester, be that multiple payers or for their credentialing or for recertification, it is just too much.  It is not what we want to spend our limited health care dollars and human resources on.  We need to speed the development, adoption and use of electronic clinical quality measures, so that we can easily extract clinical data for use, along with the cost and other administrative data. 

We need to move to data repositories, where providers can submit quality data once and obligate all other users to draw from that repository.  We can potentially do that as a complement to our APCD, so that we can once again bring cost and quality data together to understand value.  This plan receives strong stakeholder support and is part of both our New Hampshire state innovation model and our delivery system reform incentive payment, our 1115 waiver. 

I have spoken from the point of view of a public policy maker for the state of New Hampshire and specifically for the Medicaid program.  It goes without saying that providers need easier access to data across systems and datasets to better practice medicine for the individual in front of them.  I think the APCD offers many solutions to support them.  I hope the committee will hear more about that from others.

As a policymaker, I see data as a public good and an APCD as an important foundation to work from.  More needs to be done to make APCD datasets easier to access and use.  But we need to comprehensively bring datasets together from payers and providers, but also be inclusive of community services and resources.  It is only then we will be able to be data-driven in improving the health outcomes for individuals, for populations and to ensure value at both cost and quality for New Hampshire and across America.  Thank you for the opportunity to speak with you.

MS. LOVE: Thank you, Doris.  I invite you to stay on the line during the Q&A, if you have the time.  We will move right now to Jessica. 

MS. BROOKS: Thank you for having me here this morning.  I am very excited to be here.  My name is Jessica Brooks.  I am the CEO and executive director of the Pittsburgh Business Group on Health.  Christy is the other PBGH, Pacific Business Group. 

We have been in existence since the 1980s and represent over 90 employer member organizations across various sectors.  We do also have an additional 40 or so associate members that make up the health care industry, so to speak.  Pharmaceutical companies, health plans, providers, consultants, et cetera. 

We represent over 2 million lives, as well as $5.2 billion in annual health care spent.  I am also equally pleased to represent PBGH employer members as a representative on the health innovation in Pennsylvania price and quality transparency workgroup, as well as the all-payer claims database subcommittee, which is helping to determine how Pennsylvania continues to move toward an all-payer claims model. 

I will use that as a disclaimer that APCD was not even in our language five months ago.  Our employers don’t use that term.  We have had to get up to speed rather quickly to represent their matters and their voice on this committee.  There is a lot that we do need to learn. 

Also, when I reference APCD, realize that I am also talking about potentially an employer-led data strategy versus a state-wide strategy, as we are still determining exactly our position on that. I can say the results of our being at this table and on that committee is because there is a sense of urgency.

We don’t believe time is on our side, as what we consider employers as payers, to wait for a state-wide strategy to be able to move forward and make short-term decisions that will have an immediate impact, as well as long-term decisions that will help us collaborate better with the health care industry.

On behalf of the employer members of PBGH, I am pleased to share their business concerns which are driving their desire to make claims data available in an actionable manner, so they can make informed decisions about the cost and quality of health care they are buying for their employees and families, as well as serve as more effective fiduciaries of their employee benefit plans.

Our mission as a business coalition is to champion outcomes-based cost-effective health care by way to improve the delivery cost and quality of health care and employee benefits in our region.  In order to do so, we believe that an employer-led data strategy is the key to address change in transparency in health care. 

As I discussed in our efforts, I will cover three important areas.  One, why we embarked on the effort to create a regional claims data strategy.  Two, what we expect to achieve, and three, what our position is regarding all-payer claims databases. 

First, our reasons for embarking on a regional claims data strategy, employers need access to their own information regarding payments, providers and other aspects of administering the health plan benefit, which is held by typically third-party administrators. This is vital to plan sponsors to exercise the fiduciary duty established by ERISA and substantial case law.  To ensure that the plan is being run in the best interest of the plan’s members.

Having access to such information would allow employers to review their costs, the quality of health care being delivered, and identify areas of improvement in terms of both lowering and controlling costs, and increasing the quality of care available to plan members.  What we don’t want is to run the risk, although most employers haven’t said they are going to jump out of the health care game by way of public exchanges or other forms to find contribution.

This is always annually multiple times a year being considered by leadership that are accountable to stakeholders and needing to maintain their businesses’ operation.  The risk of employers not having a seat at the table is a high risk.  Further, past efforts to address change in transparency in health care historically have been made by single employers acting individually and often working across the silos of information with the proverbial left hand not knowing what the right hand is doing.

The lack of a concerted and concentrated effort to truly optimize data results and inefficiencies, which in turn lead to fragmented and decision-making, and buyer’s remorse, as the costs of health care continue to rise for both employer and the employee.  The quality of care provider continues to be confusing and misunderstood.

My employer members have not leveraged their collective buying power or influence partly because they have lacked insightful benchmark data, which would allow them to drive lower costs, improve health and target benefits programs that more specifically meet their needs in our region and also in the various markets that they reside.

Another challenge we are addressing is getting beyond the billboards and marketing campaigns.  As employers make benefit decisions on behalf of employees, they also have to combat the perceived value versus the reality value.  For provider networks that sit below the 50th percentile in a cost and quality dataset, the regional employer should know that fact one, two, work with a provider network to set performance improvement goals, and three, use that data to measure progress toward these goals.   

As such, PBGH recently contracted with a national data analytics partner, Innovu, to aggregate raw member data and to develop regional performance benchmarks to measure access to high quality and affordable health care in the region.  This brings us to point two, what we expect to achieve in the barriers. 

Employers tend to use this information in two ways.  One to set performance expectations of cost and quality with the providers and health plans in the region, and to inform each employer of how its improved their own benefit offerings to their employees.  Along with being part of the collective voice for southwestern Pennsylvania and achieving true insight into the issues driving across quality and population health.  PBGH and its employer members also receive detailed insights to improve and understand not only specific health and benefit plans, including designs, but how those plans stack up to others in terms of cost, design, performance thresholds, predictive analytics, provider profiles, risk stratification and other analytic parameters both nationally and regionally.

Additionally, employers can use such information for renewals and new services, advanced plan design and benefit program management with current regional benchmarks and trends, enhanced leverage with vendors by having comparative data and prevent loss of claims data if vendors are changed.  We also acknowledge though, that the integration of several other elements are needed to overcome obstacles, including alignment of interest among employers and organizations to manage and deliver benefits and collaboration across employers, providers, health plans and advisors to jointly discuss best practices and actual priorities based on factual data. 

Whether through a statewide claims data effort or a regional data program such as I am speaking about that we are implementing in southwestern PA, there are technical challenges which must be addressed.  I am not going to go through all the details, but data errors in quality have been mentioned.  Data integration is a key point for us.  Our claims database is inclusive of, of course, medical and pharmacy data.  But total human capital data by which employers can collect claims information, including workers compensation, dental, vision, absence management information, biometrics screening data, 401k and financial well-being.  We are monitoring that closely, the impact of the cost of health care and the financial strain on citizens, and what employers can do about that.

The core challenges that data integration evolve around technical complexity and resistance to sharing information, the data especially in health care comes from a vast number of legacy systems with varied levels of sophistication, accessibility and quality.  In addition, vendors are reluctant to share critical elements due to concerns with providing too much business transparency or, in other words, their competitive intelligence.

It is especially difficult to automate a scalable integration solution without the right expertise and technical approaches. Some of the competitive intelligence includes allowed charges by providers or even allowing the aggregation of data between employers.  This makes it especially difficult to automate and move forward as employers are desiring.  Member matching is also an issue.  Eligibility vendors, as you all are aware, as well as security, which was mentioned with the speaker before me.

The key to this regional data strategy is accessing the raw member claims data.  Unfortunately, access to the employer’s own data has been blocked largely with the TPAs and health plan ASO contracted by self-insured employers.  Third-party administrators use non-disclosure agreement, confidentiality agreements, business associate agreements and other similar documents to restrict the usage of employer data by the employer and its designated agent and other third-party service vendors, or make it very costly and sometimes prohibitive for them.

Over the years, we have been looking at a variety of different issues, such as price variation, (coughing) procedures, what employers can do individually and collectively by benefit design component.  I like to share a quick story, without having a robust integrated quality system in our marketplace, what a large employer has been able to do by leveraging independent third-party information and drive down costs.  Why we think all-payer claims database or an employer-led regional claims database can help us truly have a sustainable impact on health care.

This story is one of the largest regional employers in our marketplace, which is a school consortium consists of over 50 school districts, 43,000 belly buttons.  They have recently narrowed their network mostly by choice, but also because of the local market dynamics and increased consolidation in our market. They then leverage third-party independent benchmarks on provider quality variation and tiered their plan design.  By having insight into data, they recognize the highest quality hospital had 33,352 services, nearly 300 admits and cost them $4.9 million in total costs.  The lower quality hospital ranked number 32 had a similar amount of services, 362 claims and cost them $15 million in total costs. They were $10 million. 

After plan changes, their plan resulted in a $21 million profit from 2014 to 2016, whereas they previously budgeted $2 to $5 million deficit.  This is recent.  This is new, but this is one large employer that has been able to leverage data to truly impact change and work with providers in a different way.

I wish to wrap up with a few lessons which I have learned in working both with on the regional state level.  Several things have become abundantly clear that the collection, aggregation and use of cost and quality data cannot be left up to health plans or to providers. Rather, self-insured employers should have the loudest voice in the matter and a guaranteed right to their claims data.

They should also not be burdened with pushback from their contracted TPAs and ASOs.  Our members feel that in in order for businesses and their employees to benefit from transparency regarding cost, access and quality, there must be unfettered and expedient access to critical claims data.  With that, I would like to share that we are open and want to collaborate with the providers and health plans, but believe we have to have a leadership role in this.

Any tool that allows companies to put their data to work for their employees, giving them actual insights, they need to understand and improve the ability to influence true change in health care is a good thing and one our employer members fully embrace.  Thank you for your time today.   

MS. LOVE: Thank you, Jessica.  Kristy, we are really tight on time.  We appreciate brevity.  We do want to hear from Pacific Business Group on Health and how your views on the data fall out.  I will turn it over to you right now.

MS. THORNTON: Good morning.  My name is Kristy Thornton.  I serve as manager of Transparency at the Pacific Business Group on Health, the other PBGH.  I would like to really express our appreciation to the committee for convening this hearing on what I will call APCDs and then multi-payer claims databases, which I will refer to as MPCDs. 

The Pacific Business Group on Health or PBGH is a coalition of large health care purchasers, both private employers and public agencies who drive improvement in quality and affordability across the US health care system.  PBGH consists of 65 organizations that collectively spend more than 40 billion per year purchasing health care services for over 10 million Americans.

Our members include national employers such as GE, Walmart, Boeing, Tesla, Target, Disney, Intel, Bechtel, Chevron, Wells Fargo and Safeway, as well as public sector employers such as CalPERS and the city and County of San Francisco.

Increasing transparency in health care is a key priority for all of PBGH’s members. Our organization houses the California Healthcare Performance Information System, or CHPI, the large MPCD in the state of California, a state that does not have an APCD.  CHPI combines data on the health care experiences of more than 12 million people from three private health plans and Medicare.

PBGH is also a significant contributor to the Center for Healthcare transparency, which is a national non-profit organization seeking to make information on the relative cost and quality of health care available to 50 percent of the US population by 2020.  Last month, the Network for Regional Healthcare Improvement and the Pacific Business Group on Health co-hosted a national employer leadership seminar where employers from across the nation gathered to discuss strategies for leveraging regional data and partnership to obtain higher-quality more affordable health care.

My testimony today is informed by these efforts.  I am going to focus on why APCDs and MPCDs are a critical resource for purchasers, and how some purchasers are already using them.  Then I will conclude with a couple of areas for the committee to consider with regard to improving the ability of APCDs and MPCDs to meet the purchaser community use case.

So low-value care strains public budgets.  It impacts the ability of businesses to compete in the global economy.  This is something we are very acutely aware of.  Purchasers recognize the need for change.  They want to be able to identify and reward better value care. 

Reward mechanisms, including value-based payment, benefit and network design, are becoming increasingly available.  However, the information to identify high-quality affordable providers are often very difficult for purchasers to access or use, as my colleague just talked about.  APCDs and MPCDs are one of very few objective, reliable resources that are available to support purchasers value-based initiative. 

Our work shows that private employers in public agencies are using APCDs and MPCDs in several different ways.  First, they are a critical resource for employers to engage in high-value network and benefit design strategies, including APOs, tiered networks and narrow networks. Purchasers making network and benefit design decisions want the highest level of confidence that the performance measures that are being used to evaluate these networks are valid and reliable, and few purchasers are large enough to achieve this on their own.

APCDs and MPCDs offer neutral third-party information on the quality, cost and resource use of providers. Notably, Minnesota Community Measurement has been particularly active in this area for more than a decade, partnering with large purchasers including the Minnesota State Employee Group Insurance Program, to create a three-tier provider network based on claims and total cost of care, as well as other quality measures.

In this program, employees choosing high-value providers were rewarded with lower premium payrolls on a deduction.  Using this model, over one-half of employees migrated to the highest-value providers.  The network cost trend was minimized.

APCDs and MPCDs are also an important resource for purchasers engaging in alternative payment models.  Analyses from APCDs and MPCDs provide purchasers with standard episode of care definition and transparency into the episode cost in the wider market place outside of their own purchasing, which can inform the develop and operation of bundled payments.  As we heard earlier today, CIVHC has made some really exciting progress in this area of looking at an APCD and bundled payment arrangements for orthopedic care.

APCDs and MPCDs also provide information to purchasers for global payment methodology.  As was mentioned just a moment ago, commercial payers have used performance metrics in global payment arrangements, but purchasers lack visibility into the details of these arrangements.  They really cannot evaluate their effectiveness.

APCDs and MPCD use transparent methods of data collection, analysis and reporting that allow progress to be measured in a standard way.  This gives purchasers the confidence that they really need and want that these types of value-based payment models are truly paying for value. 

An example that I would also like to offer is in the state of Washington, who recently used MPCD data to establish two value-based medical plans for state employees that hold providers financially and clinically accountable to 19 measures of performance through their statewide common measure set.  I also want to mention that APCDs and MPCDs can stimulate the health care market through more indirect channels, as well, that are beneficial to them. 

Cost and quality reporting websites, we started to talk a little bit about that.  I had mentioned just a few examples, such as the California Office of the Patient Advocate, New Hampshire Health Costs, Get Better Maine and many others are using APCD and MPCD data.  This public transparency motivates provider improvements.  It enables better consumer choice.

I have just talked about four ways that APCDs and MPCDs are being used by purchasers and can be used by purchasers.  More directly for network and benefit design, including ACOs, tiered and narrow networks, alternative payment models development and monitoring, and then kind of more indirectly to motivate providers and enable consumer choice.

I will conclude by briefly mentioning some of the things that might be helpful for this committee to consider in terms of making APCDs and MPCDs providing more utility for purchasers.  First, generate results on a named provider basis, especially individual physicians, practice sites and facilities. Performance information at these levels provide the strongest support for network and benefit design and provide actionable information to consumers.

Second, include cost information.  Many state-mandated APCDs do have this, but many MPCDs do not.  Purchasers simply cannot address the value of it without cost information.  Third, seek out and engage purchasers and business coalitions as APCD and MPCD stakeholders members.  I know that many have done that.  I do think there is more opportunity there because this will increase the purchaser uses of data.  It will really help translate APCD and MPCD information into action. 

Lastly, as others have mentioned, I think it is important for APCDs and MPCDs to collaborate with other regional sources of data to include clinical and patient-reported data, to look at the most complete picture of performance.  I would encourage the consideration of regional health improvement collaborative again to be this local convener. 

I would also mention if the committee has interest to look at the claims and clinical data integration pilots that were completed by the Center for Healthcare Transparency last year in Cincinnati and Utah.  With that, thank you so much for this opportunity.  I will pause there.

MS. LOVE: Thank you so much, Kristy.  I will turn it over to Eric.

MR. BARRETTE: Thank you.  My name is Eric Barrette.  I am still the director of research.  I will say my disclaimer from panel 1A still stands.  I was asked to speak in this panel.  I looked at the title, and I decided I could best comment on terms of a user.  My testimony is from the perspective of a researcher.  That is how I prefer to spend my time.  I will have a few brief comments.

First, I think it is important to think about what claims data are.  They are an important source of information about prices and utilization of health care services.  But they represent financial transactions.  Although there are clinically relevant information included in a claim, it is not a clinical record.  This is important when thinking about using the data, but also important in thinking about building the datasets.

Second, how can we use claims data?  Well, we use them for research just like any other dataset.  But there is a variety of cleaning, processing, aggregating and many other methodological techniques needed to prepare an analytical dataset for an analysis.  Just as methodologies exist to support clinical trials and research design and survey research, there are many methodologies that are applied to claims data, both to construct dataset for analysis and to analyze the data.  When preparing a dataset or putting together an APCD or a multi-payer claims database, it is important to keep that in mind.  Someone ultimately needs to use this data.

In terms of research applications, because claims data collect information on services provided, and the datasets often include prices paid for those services, claims data are a common source of studies for price or utilization.  Now, other studies can be done, but in terms of thinking about how these datasets can be used, that is one of the most common uses.

Although claims datasets may be big, claims are small.  The fact that the data is produced by an individual-level interaction quickly becomes apparent when you are partitioning the data by age, gender, geography, diagnosis and then services.  Immediately at this point, sample sizes can be reduced to the levels that are insufficient for reporting or even for analysis.

I think in terms of using claims data and interdisciplinary approaches necessary, and I think the disciplines that are needed are institutional knowledge, statistical training, experiencing common sense.  Even simple as assigning patients to geographies can be complicated by the fact that is what is required to be collected for membership may not align with the level at which the analysis is being done.

ZIP Codes are a good example of this.  ZIP Codes are used by the postal service and are part of your address, but they don’t correspond uniquely with counties.  Although ACA rating areas or plan offerings may be at the county level, ZIP Code isn’t.  One ZIP Code can cross three or four counties.

Finally, a lot of thought should go into the analysis and the results reported.  The research question and specific aims need to be clearly identified.  Thinking about those before designing the APCD can be very helpful.

I will quickly mention a few limitations and risks.  One, there are limitations from a research perspective in terms of how the results can be generalized or the strength of the conclusions.  Using data with billed charges rather than allowed charges may be used to assess costs.  But the strength of that conclusion would be weak.

Alternatively, examining allowed prices in one state may produce robust results for that state that don’t generalize to another state.  Mitigating this risk or acknowledging the limitations is ultimately the responsibility of the researcher.  But it helps when the data with fewer limiting factors is available to work with and then the limitations of the data are known.

Second, claims include sensitive financial and personal information.  That creates large risks from the chance that it may be disclosed either intentionally or unintentionally.  It is important to have levels of governance and monitoring and enforcement.  There is trust required of the researchers, as well as the data contributors, to provide access to the data for appropriate uses.

Finally, research especially in health economics, health services research or health policy that evaluates programs and policy needs to be reproducible.  That includes the methodology, as well as the data itself.  A good example of this is the numerous amount of studies that have been done with Medicare data.  Because there are only small changes over time in the claims, and those changes that do occur while documented, it is easy for other researchers to follow on from previous studies.  There is a risk to researchers, both to their results and the credibility when their claims data vary from source to source and over time.

A few concluding observations, the more detailed information in the data the better. But more importantly, the more detail about the data is what is most important.  Understanding the data is the first and most important step to using the data.  We can collect more and more variables or fields.  But if that isn’t done through standardized process or what is actually collected is unclear, it is very difficult to use the data for reliable analyses.

Second, a knowledge base about the data is valuable to new and existing users.  Being able to draw on lessons from other users is very helpful.  Finally, I will point out that access to commercial claims data for research purposes is a relatively recent advancement.  There is still a lot to be learned from claims data in the commercial population, as well as using those claims data.  Access to consistent data, even with certain restrictions or limitations of use, is better than trying to work with small varying disparate datasets or no data at all.  Thank you.

MS. LOVE: Thank you.

MS. TURNEY: Thank you.  My name is Cheryl Turney.  I am a senior director for All-Payer Claims Database Analytics with HealthCore Inc.  HealthCore Inc is a wholly-owned, independently-operated research subsidiary of Anthem Incorporated.  HealthCore conducts a broad array of cutting edge health research studies and analyses which produce much needed real-world evidence on the safety and effectiveness of biomedical products for both internal and external finds.  I am here really representing both Anthem and Health Core.

Health Core is one of the nation’s leading health benefits companies with over 72 million people served by its affiliated companies, including more than 38 million enrolled in its family of health plans.  I have been the lead for the Anthem APCD implementation since I have been there, which is about six years. Anthem submits APCD data in 12 states, including three states in which the submission is voluntary. 

We have worked very closely and very hard to be collaborative with all of our entities where we submit APCD data.  We support the implementation and ongoing maintenance of these resources. 

Anthem, in its effort to fully support these efforts and promote high-quality, high-value care, hope that sharing our experience here today and the recommendations with NCVHS will go to improve the effectiveness and policies and standards as you make recommendations out of this committee.

Based on our extensive experience in both as a data submitter and as a sophisticated research organization, we also believe that perhaps a federated distributed model may be an opportunity that we need to consider.  This would support a multi-state data model that would allow stakeholders to realize the real potential benefits of APCDs, while minimizing the burden on data submitters and also increasing sustainability with the states and the researchers. 

I liken it to the library.  I can either own all the books in the library, or I can own the way the library is organized.  The have the library be shared, and then we are all paying for those shared services. That is exactly the way we are looking at it.

There may be different levels of sophistication with some of the distributive models. But if you do look at the CMS model with the edge servers, it has come a long way since it has been implemented.  We believe that with your leadership, that may be something you may want to look at.

I also wanted to bring up, because I am working on a number of initiatives in various states, that Connecticut is actually piloting a federated model.  Different people had asked about that before.  It is not working yet.  However, they are using a SATO health technology.  It might be something that folks would like to consider and look at.  That particular technology brings together an analytics platform, the federated model as well as the ability to use semantic mapping in order to really get into a clinical record.

One of the reasons why we are recommending that we look at a federated model is because as we look at the uses of the APCDs, it really doesn’t stand on its own.  As other people have mentioned already today, claims data is just one piece of the puzzle.  We need really clinical data.  We need lab.  We need test data.  We need lifestyle data.  We want socioeconomic data.  All of these things come together and are really necessary as we are evaluating the quality and the services and even the patterns of behavior for consumers.

Are there certain types of behaviors which, as our research company has noted, as people are moving from the commercial market into Medicare?  All of those things require this additional level of data in order to properly be able to recommend changes to behavior, which is really where we all want to get to.

Some of the things that we have noted with APCDs, which have also been brought up today, is that many of the states have spent a lot of money, as we have, gathering data, aggregating data, with very little time or money really able to be spent up to this point in uses of the data or really learning from the data.  That does take time.  It is a relatively new concept that we have put in the last four or five years.  Really, what we would like to see and support is more uses of the data and value of the data, both to employers, to patients, as well as to payers, and help that model reduce costs.

One of the things that we noted is most APCDs or claims database entities, made claims data available to third parties for research.  It is significantly limited, as has been stated already, because each APCD is set up differently.  Then the onus is on the researcher if they want multi-state or multi-geographic data to then normalize that data, which really they are the entities with most likely grant money.  They are going to spend grant money over and over and over again in order to normalize this data and bring it together. It just doesn’t make any sense.

Currently, claims data-based entities don’t use a standard data model.  Whether you use a federated approach or a centralized approach, that is really the essence of the framework.  You need to have a standard data model that everyone is using in order to further the efforts in all of these avenues.  Similar to what happened with the genome research, that started with a common data model.  Maybe that is where we need to move back to. 

We have also seen the opportunity for greater transparency and engagement when the APCD data is used.  Each state handles their data use and their transparency of how the data is used differently.  Some states are creating a lot of reports and a lot of cost model data.  They are very open and very transparent.  Others are not.

If we are all in the business of making claims data transparency more available, then how data is used to create a report needs to be also transparent.  Otherwise, collaboration by providers, hospital groups and payers is not going to be there.  You are going to spend all your time explaining how you came up with the data, rather than using the data for the end result, which is to change the behavior.

Skipping around because I know time is an issue, and I am going to save some of my material for later with standards.  Also another thing that we wanted to really talk about was the changing landscape.  Anthem is involved in a number of measures that are looking at variable payment models, both on the federal and the state basis.  In producing data models in order to adequately provide information back to providers and employer groups and all of our stakeholders, these transformational reforms require different data. 

We have found, even with our own models, claims data alone doesn’t tell the picture.  We can’t report accurately on quality back to our provider groups just with claims data.  We need to integrate clinical data.  In order to integrate clinical data, are we going to get it from our providers.  Are we going to get it from aggregators?

One of the things that we are finding is all the aggregators want claims data in exchange for the clinical data.  So now not only are we submitting claims data to APCDs, but all the HIEs.  How many places are we going to replicate this claims data?  It doesn’t really make any sense when it all is in a different framework, a different format.  The uses of the data, we don’t even know how they are going to be using the data, and they can’t tell us because they don’t know yet.  It is one of those things where they said we know we need it in order to measure quality.  We don’t know what data we need.  We are trying to figure out what the quality measures are.  As we learn and use the data, we will be able to hone it down in terms of what that data means to us.

As this expands, that even more in my mind represents the need for a federated distributed model.  We are going to be going way beyond claims in each state.  You are going to have to hook up all of your providers.  Is it going to be APCD through the HIE back to the payer model?  How is that model going to work? 

It does need, I think, some interaction by the ONC, which was mentioned before, which has really kind of left this out.  We continue to make comments on that area, as well, as AHIP had mentioned, and I think a few others.  But to date, it has really been outside of the roadmap that they have put into place.

To sum all of that up, we want to thank you for the opportunity to speak.  I will save all the rest of my stuff for the standards.

MS. LOVE: Now we are going to shift gears to the next panel.  I think there is a continuum, so it is not a hard break.  We are going to have two more speakers.

MR. BRANNEN: I am Tyler Brannen.  I am the New Hampshire Insurance Department health policy analyst.  I have been in the department for about 10 years.  I have spent the first five years as the department’s health care statistician.  I think at this point, I probably still use our claims database more than anybody else in the state.

I do have slides.  I wanted to use slides, though, because rather than just tell you what we have been doing, I wanted to show you what we have been doing in New Hampshire.  The first thing I want to do, though, is show you our missions statement.  It probably looks like a lot of other missions statements.  It is pretty general.  But there is one specific part that I wanted to bring your attention to.  That is as it relates to competitive insurance markets.

Our perspective is a little bit different.  I was asked to participate today by the National Association of Insurance Commissioners. My position does not necessarily represent the position of the NAIC.  But New Hampshire has done a lot of work in this area.  That was recognized. I was probably known to be the best person to speak to some of the issues, particularly from a regulator perspective.

We obviously enforced insurance laws. At some level, we are a consumer advocate, as well.  Sometimes we are the only consumer advocate.  It is a positon I am fairly used to.  A bit part of my responsibility these days is actually supporting the legislature.  I do that a lot by using our claims database, and the studies and information that have come from it.

In 2003, New Hampshire passed a law to create our all-payer claims database.  We are not a new state with these data.  We have a lot of experience with them.  We do have some things that are bit unique.  It is an insurance law, which at some level makes intuitive sense. Being the insurance department, we can enforce that law and make sure that the data submitters actually follow through.

New Hampshire also licenses third-party administrators.  They have been subject to the law, too.  There is reference a bit earlier, the Gobeille versus Liberty Mutual case, which has raised a number of concerns throughout the country and including New Hampshire.  We have already actually made steps to address some of the confusion around that. I will get to that in a moment.

The law that actually creates the database is a bit general.  The insurance department puts together the data collection rules.  We are required to work with our state DHS to work out some of the details, including their responsibility for any release of the detailed data.  That way patient privacy can be ensured.  Indeed, New Hampshire’s concerns around patient privacy are probably at the top and near the top nationwide.

The good news is that one of the prime legislatures that was involved in getting this legislation passed is a conservative republican, a self-described privacy nut, but somebody who feels as though one of the best solutions to dealing with our health care cost question is to have better informed consumers that start to treat the health care delivery system like a competitive market.

These are some of the examples of how we have used the data.  The following slides will give you a more illustrative example.  We have been making health care prices public actually since 2006.  Our health cost website went up in 2007.  But I started putting up reports at presentations as early as October 2006, which is about three months after we got the data.

During this entire process of working with the data, reporting on it, building our health costs website, I think it is worth noting that not everybody has supported our efforts in this area.  Insurance companies weren’t crazy about it at all at first.  To quote the president of one of the largest insurers in New Hampshire, yes, I think we were asleep at the wheel when this legislation went through. Somebody else said, yes, we knew it was coming, but I thought we would be dead and gone by the time it happened.  We are really surprised you guys moved as quickly as you did.

We have looked at a study of hospital cost-shifting, which is a theory that hospital prices are so high because the government payers don’t pay enough.  Medicare, Medicaid, and obviously the uninsured don’t pay as much as commercial insurers.  We wanted to test the theory and see if the most expensive hospitals were the ones in New Hampshire that had the greatest share of Medicaid, Medicare and the uninsured.  That was not true.

We have a full study on this, but we did have some interesting findings, such as things like occupancy levels seen to drive prices up.  The amount of Medicaid patients actually seem to drive prices down in some of those organizations.

We have done an analysis of our market looking at the discounts they get with health care providers.  I have got a slide on that.  Clearly, if a health insurance company can’t get good contracts with health care providers, they are not going to be competitive. 

We have looked at the impact of patient age on health care costs and premiums.  If you are familiar with the AVCA, you know there is a rating factor of a maximum of three to one.  That means insurers in the individual and small group market can only charge three times as much for premiums for the older population.  It sounds like a lot, but if you look at the underlying claims experience, it is more like six or seven times. We know that.

We also know that in five years, you can see those claims costs go up about 20 percent.  Age is a massive driver in New Hampshire and I am sure everywhere else when it comes to health care costs and premiums. 

Just recently, we have been looking at the cost associated with the opiate substance use population.  We have done analysis both at the fully-insured and the self-insured levels.  This is what our market looks like.  It probably looks like most insurance markets to the extent it is dominated by two or three carriers with most of the volume in membership.

One of the other things I would like to bring your attention to is how their emphasis, their focus on a particular patient population differs.  You can see Cigna is very focused on that self-insured population.  For those of you who don’t know, about half of us walking around with commercial insurance cards, including me with Anthem on it, are actually under self-funded policies. The state of New Hampshire retains the risk for those claims.  If I incur a million dollar claim, it gets billed back to the state of New Hampshire.  Hopefully the state of New Hampshire has some sort of reinsurance. But the point is Anthem is not on the hook for those particular claims. 

There are a number of advantages for large employers to do this.  One of them is they are largely exempt from state regulation because they are an employer, and they aren’t actually purchasing insurance.  The health insurer, the TPAs as we refer to them, are usually an insurance company in New Hampshire, your Anthem, your Cigna, your Harvard Pilgrim.

These are the same companies that are selling fully-insured products, processing claims, largely the same way, administering reports to all of their clients in the same way or perhaps differences depending on client preferences, developing disease management programs which have additional reporting to different clients in different populations.  Of course, these insurers that are often creating risk arrangements with different health care providers. Of course, there are probably a couple of thousand reports created for those populations, as well.

This is a screen shot of our New Hampshire Health Cost website.  I would like to spend half an hour or more just talking about our website.  It has certainly been a source of pride to the New Hampshire Insurance Department. 

What we have been doing for now almost ten years is posting bundled prices for consumers.  These bundled prices are actually based on the allowed amounts, the amounts paid to the health care providers.  They include multiple services.  In this case, if it is knee surgery, includes the surgeon, includes the hospital or surgery center, hopefully includes an anesthesiologist.  It may include x-ray or lab or whatever else is routinely done.

The point is we knew the consumer would know how many bills they are going to get from what different providers. This was set up for consumers.  We wanted to make it as easy as possible for them to use.

One of the things we are dependent on, though, in doing a website like this because we report at the specific provider level, by the insurer, by even the line of business, whether or not they are an HMO so you can get an accurate estimate, we need enough data to be able to produce a reliable estimate. We are not just calculating averages or medians.  We are actually exercising a certain level of judgment as to whether or not we can produce a rate that is going to be reliable enough for when people go to the next time to that type of service at that health care provider.  Including the self-funded data in that analysis is important for us to do to be able to produce the number of rates and number of providers and number of services on a website like this.

This is a shot from a report we produced a few years ago in 2011.  It was actually a second of two reports, where we were looking at the carrier discounts. I mentioned those a minute ago.  The blue bar is approximately for market share.  It basically tells you how big the insurer is in New Hampshire. 

The red bar is the average discount they get with health care providers across the system.  Now remember, the ACA requires that either 80 or 85 percent as a minimum of the dollars incurred for a premium go out in terms of medical claims costs.  That means relatively small differences as a percentage point in how they pay health care providers means a massive impact on that carrier’s bottom line.

I would go so far as to say if it is a 3 or 4 percentage point difference, they are not going to be competitive.  In the past, they have been able to competitive because they have insured healthier populations.  That is how they have sort of been able to stay in our markets with a limited population.  But as a larger carrier, no, you won’t be. 

Indeed, the fourth-ranked carrier there, MVP, they have since exited the New Hampshire market.  The primary reason given was because they couldn’t get good deals with hospitals, at least not as good enough as Anthem, as Cigna or major carriers.

This was an analysis done as a part of our opiate substance use analysis.  But the first step was to take a look at the difference in the occurred claims cost by carrier, by whether or not it was self-insured populations or the fully insured populations.  Now you are all experts in that area.    You can note that there is a consistent trend in every case that the dollars paid out, which would obviously reflect sort of the premiums or the administrative cost to an extent to the employers are lower for the fully insured.

That makes some intuitive sense.  The insurance company is now spending their money.  They may do it in a more intelligent way.  They may use disease management programs, wellness programs, other sort of utilization review criteria to make sure those dollars are managed most efficiently.  Alternatively, they may be selling those kinds of programs to an employer who is self-funded. They may or may not be as effective.

A caveat, I mean, this isn’t adjusted for the age of the population.  It is not adjusted for how much they retain administrative cost, which tend to be lower for the self-funded. There is a lot of things you can do on top of this.

The point is there is a bit of a trend, which you see a little bit more when you just look at the specific population receiving substance use treatment.  Ironically, the percentage of the population among each of the carriers receiving this type of treatment was almost exactly the same, about half of 1 percent.  You can see that the difference, though, between the self-insured population and the fully insured is bigger than it was when we just looked at the claims dollars in aggregate.

You can also see that the exception is Anthem.  Anthem also has the lowest costs among all the different carriers.  We did this analysis because, at the same time, we are doing a market conduct exam to see if the insurance companies are following all of our state laws about how to process claims, whether or not they are denying things inappropriately, whether or not they are creating access barriers for treatment for these populations.  We haven’t completed the market conduct exam, but this obviously can lend insight.  We are using the all-payer claims database to help us regulate the insurance market in perhaps somewhat unique ways.

This is actually taken from our law.  The main point here is that these data and the database that was created in New Hampshire, it is created as a public resource. The insurance department has used it a lot.  DHHS and other entities have used it, as well.  One of the main motivations when we started this project was really to better understand health care costs and utilization in this state. 

At that time, people had much lower deductibles. But we didn’t have any expectations as sort of high-level as bringing down health care costs.  We just felt as deductibles are going up, people should know what it is going to cost.  Of course, this particular legislature felt as though this was going to be one of the answers to our health care cost problems.  Transparency of health care prices in New Hampshire and probably everywhere has been receiving a lot of support from both sides of the aisle.  It is one of those areas where we can get support regardless of the politics for the most part.

One of the things I do want to mention is just that insurance is regulated at the state level.  To the extent we think about what is happening at the state level, it is one of those areas that we want to do things as well as we can.  Our claims database allows us to effectively regulate our markets, understand what competitive forces are existing, or, in some cases, even analyze and review mega mergers among insurance companies.

It does inspire, we feel, competition among health care providers and insurance companies.  When we put up that health cost website, a lot of people didn’t support it, carriers, providers.  But they got used to it. They felt as though we were doing it well enough that the information was of use to them. 

A health care provider is the most expensive out there. In fact, they tried to sell themselves in some cases as a lower cost provider. Look at this website.  You can see if you come to use, the insurance companies, one or two of them that objected to us putting up these prices said, we thought we had the best deals.  It is now clear we don’t.  I already mentioned sort of the legislative support for our work.

New Hampshire did file an amicus brief in the Gobeille versus Liberty Mutual court case, as did the NAIC.  Just on the basics, I am sure like everywhere else, a lot of our members are covered under these self-funded policies, more than half. To the extent the ERISA and the decision applies to a particular population, it doesn’t apply to me and other state employees. It doesn’t apply to some of the government entities.  We are still talking about 30 percent of the data that it matters as it relates to.

I mentioned the carriers that we actually have as CPAs. We do feel as though if we lost 30 percent of the data, that is going to pose some problems for our health cost website, as well as just understanding our markets. The permeability between the self-funded and fully insured is pretty standard.  You can go back and forth.  Most of the time, the employees don’t know.  Of course, if you are using the same insurance company, your card for the most part looks the same. 

The court’s decision did not specifically address the preemption of insurance regulation.  There is sort of a savings clause as it relates to insurance regulation.  I wouldn’t be the best person to go into a lot of depth around this.  It didn’t address it. 

New Hampshire moved very quickly after the court decision to pass Senate Bill 431.  This was going to clarify the reporting requirements in New Hampshire.  In fact, I think we got the hearing and the first testimony within a week.  But nevertheless, we got it past the Senate, passed out of the House.  It was concurred and signed by the governor on Monday. 

We now have certain clarification that the insurance companies and the TPAs still need to submit the data for the fully insured. They still need to submit the data for the non-ERISA self-funded.  They need to provide a sort of option for employers to submit their data to the state.

We realized we wanted to at least provide these employers with some information about what we are doing and why it is of value to them.  There was actually one point where we suspended the health cost website due to issues with our consolidator.  I heard from all of these employers saying, I am using your data.  When are you going to have it back up?  We said okay.  We had to answer those questions.

But that wasn’t our primary audience.  It never was.  We have always tried to focus on consumers, patients as consumers. Nevertheless, we do feel as though as we are going to be a certain education component to tell people who these data are submitted to us and that patient privacy is a priority of New Hampshire. But look, we are using these data a lot to help you to be better informed about how to control your health care costs. 

I have actually personally heard from somebody who wanted to know what kind of recourse they had against our larger insurer in the state.  They felt as though they should have known that going just down the street from their local hospital for the four CTs over a year, x-rays, whatever they were, would have cost them a fraction of going to the hospital.  They had a $10,000 deductible.  They felt they had been misled, and that they should have had that information ahead of time.  So very strong feelings for some people who are facing these high healthcare costs.

One of the biggest criticisms of our health cost website and the initiatives in New Hampshire is that not enough people know about them.  I think people don’t realize though the small scale of what we did.  It was largely me doing the programming, using the data, and then sending it over to our web developer which we paid $15,000 to set up the framework for the website and then loading the data.  I mean, this was not a massive initiative.  It was not a million dollar initiative in New Hampshire.  It was really a very small-scale effort.

We have received those grant funds. We have made investments in the website.  We have done a lot to promote it. Still, we will continue to do that.  It is always going to be a work in process.

That was a lot of information really quickly. That is my contact info.  I am happy to follow up with people at any time after this, as well.  It is certainly something that has been important to New Hampshire.  I just kind of want to remind you that there are always going to be challenges with these types of projects, with the APCDs and developing some national standards and working through the details.

I think it is important to remember not everybody is on the boat that we should be doing.  It is important enough. It was important enough to the people in New Hampshire just because they needed to understand cost better.  You have heard about all of the other advantages that could potentially come from these claims data.  But the cost issue, what we pay for health insurance premiums and what we are paying for health care services is really important to a lot of people.  I do have the gift for stating the obvious sometimes.  Thank you.

MS. LOVE: Thank you, Tyler.  Thank you, the whole panel, for being flexible.  We have one more.  Then we will open it up to the committee.

MS. KAHN: Thank you.  I realize I am standing in between you and lunch.  I appreciate the chair’s willingness to let me squeeze in here before you do your Q&A and take the break. 

I am Jessica Kahn.  I am the director of data and systems for Medicaid and the Children’s Health Insurance Program here at the Centers for Medicare and Medicaid Services.  I am speaking today from the federal perspective. 

But before I get into my prepared remarks, I just want to pause and say how amazing I have found the testimony so far this morning.  I have really been impressed with particularly what is going on in New Hampshire and some of other places and entities that have talked about how they are using the data as opposed to just collecting it, which were points that were really well made.  Just kudos to my colleagues who are speaking to the committee today.

From my purview, what I can share with you, obviously at the federal level, we are not as deeply involved in what you will hear from the states themselves about how they are Medicaid’s role in APCDs and how they are using the data.  But I can tell you a little bit more about how we fund it, so that you can understand what kind of federal dollars are available and in what circumstances.

Then what I have observed to be some of the limitations or barriers, some of which I have heard already addressed today.  I will just touch on them very briefly in the interest of time.

To be clear from a CMS perspective, state Medicaid agency participation and all-payer claims databases is not required. The state themselves might make it required for Medicaid to participate.  But from a federal level, that is not something that we would require.

The states are choosing, and they are sharing their Medicaid claims and encounter data if they determine that in doing so would address their business needs, such as for things related to what we have heard about today, planning for cost, efficiency, quality of care, system utilization and so forth.

In some cases where you see the state government itself is hosting the APCD, that dramatically increases the likelihood of Medicaid participation like Utah or Kansas and Tennessee among many others.  Not just because it is perhaps more technically easy within the state apparatus, but also because they have the policy levers to include Medicaid and also address some of Medicaid’s business needs in doing so.

I am sure everyone is already aware of the APCD council site.  But there are some really good examples on that site of how states have used the data to pinpoint key issues for Medicaid.  I encourage everyone to look at those case studies if you have not.

Now let’s talk money.  Federal matching funds are available for some of the costs associated with Medicaid agency participation in an all-payer claims database.  We have a number of different match rates that can be applied in different scenarios.  I am going to give you the high points here because I think it is important when we are talking about APCDs to understand what we can and can’t pay for.  Although I also want to note that I and my team have concerns about the concept of needing to build 51 of them.  There was a previous speaker who talked about broader than states and perhaps regional APCDs.

For me, that is quite appealing, the concept of having to contribute millions of dollars federal share for the infrastructure to be there in every state and with the lack of interoperability both semantic interoperability and technical interoperability that again other speakers mentioned makes that duplication that much less desirable. 

With that said, here are the basics.  If a state does not have an all-payer claims database, but would like to build one, such as what is happening now in New York State, the Medicaid program may be eligible for 90 percent matching funds for their share of the cost of that bill, so not paying for it on their own, but with all the other entities that would be using it.

Assuming that they can justify to CMS that all-payer claims database is going to meet Medicaid’s functional business requirements, that they would have otherwise had to have meet through a different technology solution.  In other words, they can help build it if it is going to serve some purpose that they would have to have built something some way to be able to do their program work anyway.

If that was the way they were going to receive encounter data through their claims or store their data, some other business functional need that this is the path that they are going to use to get there, then they can help with 90 percent of their share of the costs.  Their share generally cost allocation could be by percentage of claims that are attributable to Medicaid.  But there are other methodologies that we would consider.

In that instance where Medicaid is using it to meet some of its functional needs that it has to do anyway, the state could, after it is built with the 90 percent Medicaid match, then the state could also receive 75 cents on the dollar match for Medicaid share of the ongoing maintenance and operations cost of the all-payer claims database.  By that, I should be clear we mean the technical and systems cost, not necessarily staff, rent, the indirect cost.  But the things that are necessary in order to be able to, the data warehouse, the data analytics, the interfaces, those pieces.  That is kind of substantial and is ongoing.  There is not a cap.  Those funds to not expire. 

Another example is if an all-payer claims database already exists, then Medicaid, if they decide they want to participate, can receive that 90 percent match again to build their interface between their own state’s claims and encounter data warehouse and the APCD.  If there is one that is external, and they want to have a pipe to it to share their claims, we can pay for 90 percent of the cost for that interface. The state would have to match it 10 percent.  That is just for that technical interface.

But the cost for the Medicaid agency’s ongoing participation in that APCD, again this is not where they are doing it because it is meeting some immediate required functional need, but because they think it would be the right thing to do for many other reasons.  I am differentiating between required and good to have. Then the match is at 50/50 for those ongoing maintenance and operations cost.  That is because, as I said in that sense, the desired interface is something that the state wants to be a part of.  But it is not serving a purpose that without which they would have to build the duplicative infrastructure within their enterprise.

If this sounds a little complicated, that is because it is.  Generally what we do is work with each state on a one-on-one basis to say what are your goals, what are you trying to achieve, tell us what this looks like.  Then we help them figure out the right match rate for the right phases of what they are doing.

What this highlights, as I am saying this, is that the amount that we can contribute federally for Medicaid participation is pretty variable depending upon how the state uses it.  Just to point to New Hampshire, since they are on the panel and talked about the robust APCD that they have, we have a waiver with the state of New Hampshire that allows New Hampshire Medicaid to pay premium for Medicaid individuals to be in managed care plans.  We call it premium assistance.

In Tyler’s role, to call him out, what he talked about when he was talking about reviewing the quality of care and the adherence to state policy guidelines of those plans, it would be in the Medicaid program’s interest to make sure that APCD has all of the right claims because, in effect, those are the commercial plans that they are doing premium assistance for Medicaid enrollees to be able to participate in.  It is a really important connection and reliance that they would now have on the APCD and on their insurance commissioner’s partnership in overseeing the quality of care and the access to care for those Medicaid beneficiaries.  You can see why we work with each state on a one-on-one basis.

Some of the challenges that we hear to Medicaid’s use and participation in APCD have to do with encounter and claims data quality.  Not to say necessarily that encounter and claims data quality is worse in the Medicaid scenario than it is in other forms of health care commercial or otherwise.  But it is something, that plus the lag times, which make it a challenge for the Medicaid agency to rely on it to meet business needs.

If they are already getting a lag and getting the data from the clearinghouse or from the managed care plan or directly from the provider for fee for service, by the time they pass it to the APCD, it could be quite a while before that data can be useful to them or to others. Obviously the value of any data at APCD is diminished if the data is stale or represents disparate data models or definitions or gaps.

The CMS has been really working with state Medicaid agencies on this idea of improving the quality of their claims and encounter data.  We have historically received claims and encounter data from all of the states here at the federal level for a data set called the Medicaid Statistical Information System. 

The challenge is there wasn’t a shared data dictionary. There wasn’t a shared data semantics about how to code certain kinds of waivers or certain kinds of ways that people participate in the Medicaid program eligibility groups consistently.  So there hasn’t been historically a lot of use of the data.

What we have done is rebooted that, created a standard data dictionary, made some very clear guidelines on coding and mapping.  We have a federal system that is live that the data is screened through business rules as it comes in, so that we can work with states to improve the quality of the data they get for claims and encounters.  It does include many other types of data, as well, such as beneficiaries and providers and others.

I think that this effort that we are doing at the federal level to help states with their claims encounter data as a secondary benefit will improve the quality and timeliness of that data when it is shared in the states that have an all-payer claims database with Medicaid participation. 

There are a couple of other areas that may limit the value of APCDs for the Medicaid program.  The first is, and again, I did hear this mentioned by some of the previous speakers.  That is the idea of layering clinical data and claims data together to have a more full view of not just utilization and cost, but also impact of Medicaid coverage.  We have a lot of states that have been investing in this for a while.

Where the APCD is not linked to a health information exchange or is not able to layer in clinical data, the states might find that it can’t meet all of their needs.  It limits the value to them and what they can get out of it. It can cause some operational inefficiencies and obstacles with person matching if they are trying to see, here is Jessica Kahn’s claims in the APCD.  She cycles in and out of Medicaid and commercial payers.  But the clinical outcome data that is in the HIE, there she is under Jessica Kahn.  It is a question of matching.

I know there is some discussion of standards later.  I am just making a nod to that, that sometimes the limitation of what data is in the APCD and the need to link it to other datasets can be challenging.  Where it is not already occurring, expanding the vision of APCDs to include functional linkages to health information exchange with those proper privacy protection might engender more Medicaid participation or more robust Medicaid participation.

The second one, I was glad to hear mentioned, that is that Medicaid, in trying to drive better health and wellness, is very tightly coupled with human services programs.  More than 40 of the 50 states, plus the district, do have integrated eligibility with SNAP and TANF, which are food stamps and payments for people who need temporary financial assistance.  That data, understanding not just where people have claims and were able to get services, but was transportation necessary.  Was childcare a part of that?  Do they have stable housing?  All of those pieces together are what Medicaid really needs to understand in order to make sure that access is truly there.

I heard that touched upon, and so I am not going to go into much more detail about that.  One thing I do want to raise is where this comes up most pressing for us is in the definition of a provider and therefore also in the definition of a claim.  If you are thinking about health care including social determinants, and a provider could be a foster care parent, a provider could be a parent caretaker relative for a Medicaid individual who wants to stay in their home with their disabilities needing long-term care in the home.  Providers and quote services are much more broadly defined in the world of health and human services combined.

I think if we want to truly understand health care costs and the impact that those costs have on utilization, the impact of costs and utilization finally on health at the individual and the population level, then we need to be careful when we are constructing a strategy that is fairly narrow defined to things that have an ICD-10 or a HCPC or a billing code. 

So with that, understanding where we are timing wise, I will pause and just say again that we are very open to states’ proposals and to negotiate the availability of federal matching funds on a state-by-state basis, though I remain very interested in where we could look at efforts to leverage APCDs beyond a single state and some regional work, so that we can have some multi-states.  With that, I thank you.

DR. SUAREZ: Thank you very much.  We are going to proceed with questions from the committee.  We are going to move a little bit of the schedule.  We are going to take a lunch break at 1:00.  It is 12:30, so we will have 30 minutes for Q&A.  We will take 45 minutes for lunch and start at 1:45 again.  We will proceed with the remaining panel 2 at 1:45.  They graciously have agreed to be moved to after lunch.  Any questions from the members of the committee?

MS. GOSS: Thank you.  Ben, you made a comment earlier about eligibility file structure.  I wasn’t sure I was quite tracking with you when you were talking about that, what kind of eligibility file you were talking about.  Were you talking about an eligibility for coverage file?  Were you talking about an enrollment file for insurance?

DR. STEFFEN: So, I was talking about the eligibility file that I was familiar with would track an NCX12 transaction.  I don’t know, 270, 271 form.  That is what I think most states are using with some augmentation as the sort of precursor to their eligibility file.

MS. GOSS: Jessica, I was curious about your statements related to the definition of a provider.  I understand how that trips into the challenges with the definition of covered entity under HIPAA.  One of the recommendations that this full committee has sent to the secretary of HHS, Secretary Burwell, is to consider the expansion of the definition of covered entities.  I am not sure, one, if you are aware of that.  Two, it seemed to me that there may be an opportunity for you to do some collaboration within HHS related to sort of their thought process around that. If there is a glide path that would give us the better ability to get data flowing using standard transactions with a larger set of players like vendors, employers, et cetera.

MS. KAHN: No.  I wasn’t aware of that recommendation, though I will certainly pursue it. I have been involved in many of the cross-departmental efforts discussing provider definitions and provider directories in general because I think that is where we have some overlap here between the two efforts.

Again, I think that is a possible glide path, yes, with looking at it from a HIPAA perspective and from the standard transactional perspective.  Some of the challenges that we are going to have are when we are dealing with providers who, for example, don’t have a national provider identifier and may not, based on certain definitions of providers. There is a lot of weeds there, but I think you are right that would be one opportunity to at least broaden it.

MS. GOSS: Some of the work that I did on MLTSS really seems to me that there is an intersection point of probably a lot of pain downstream if we don’t start thinking about this now.  Thank you for any internal coordination you can do on that. 

I also wanted to see if Doris is still on the phone.  You made a comment earlier about some work that you are doing that is producing some data file formats or functionality that other Medicaid can use.  I didn’t quite track with all of that, and wondered if you could just clarify that remark, so that we understood what kind of resource, understanding the tools developed with federal funds are usually very easily available to other partners in the federal space or public.  Could you talk a little bit more about what that functionality is and how it could be leveraged?

DR. LOTZ: No, because I am not the technical person the phone.  You are correct, though, that whenever it gets exportable to someone else, we are happy to share it broadly with other Medicaid programs.  I can’t speak at a level of detail that you are probably looking for as to how the CMS and its vendor, SEI, are going to clean up the identifiers, so that we can have some confidence that we are looking at providers in a way that allows us to understand the data.

We are looking at members in a way that we can understand the data across those datasets. We can also look at reducing those duplicate claims.  At the end of the day, I am still a physician first. I rely on other people to operationalize this.  But I would be happy in some delayed way to follow up.  It should be something worth sharing by the end of the summer.

MS. HINES: Doris, you mentioned having some kind of interface showing social determinants data with your APCD.  Could you say a little bit more or provide us a link because that sounds very intriguing. I don’t know of any other example of that.

DR. LOTZ: Again, that is something that we are just beginning to develop.  Well, not beginning to develop, but we have been developing it for about a year.  But I don’t have a link ready to go until probably again the end of the summer.  It is going to be a busy summer.

I could certainly forward an overview of the project. We have created a nice PDF with some screen shots of what it will eventually look like.  But it is not operational yet.

MS. HINES: Any information you have, lots of nodding yesses in the room, thank you.

DR. LOTZ: I do that through Joe Porter, who is someone that I work closely with. 

DR. COHEN: Is that going to be linked at the ZIP Code level, the census track level? What level are these overlays going to be at?  I am sorry for interrupting.

DR. LOTZ: It is going to at the ZIP Code level and in some places even smaller.  I think that there is some census tracks that are actually smaller than the ZIP Codes that encompass them.  We will have it down to both levels.  You can choose which one you want to use.

DR. SUAREZ: Linda.

MS. KLOSS: Thank you all.  Let me probe on a question related to the discussion that several of you raised about a common data model.  I just want to understand if you believe that it could.  I mean, I am assuming that when you are talking about a common data model, it is not a common data model that gets imposed on the plans of those who are processing the claims.  But it is a common data model that exists for this purpose.

Therefore, is it a common data model, or is it a model to normalize data to a common analytic database?  I just want to make that distinct.  It seems like the latter is certainly a much easier lift than going back and trying to retrofit all of the systems.

When we think about standards, we are thinking about a standard that could normalize, so that it could be used whether it is state-run analytic database or it is a federated or regional.  It just seems to me that in our pluralistic world, it is not going to be one or the other.  When we look at this, we need something that is flexible enough to accommodate models. 

MS. INSKEEP: I think that what we are all trying to get at are the costs of health care, the rising costs of health care.  What are consumers paying for health care?  What does it really cost?  If you have the insurers and others submitting data with all these disparate ways of submitting, I am talking about field order, the fields included in the file.  I know this is excruciating.  But even just the transmittal process, how many times some of these states have you resubmit the fact that their QA is years in arrears.  Even looking at the data, that adds cost to health care. 

I think our opportunity is if the data were collected within parameters, submission parameters, field order parameters, the number of files parameters, and that health plans who are the primary.  We kind of have that burden of producing, if you will.  If we could really hone in on those specifications, I really think that not only would the states be informed more efficiency, but it would also help of the goal of decreasing.

MS. KLOSS: We are not talking about imposing that on the plan.  The plan produces data in that format and maintains its own dataset however it wants to according to the standards that exist.

DR. SUAREZ: We will be talking about the standards in the afternoon. 

MS. INSKEEP: The submission standards not necessarily of claims data warehouses.

MS. GOSS: In this case, you are actually talking about standards from a data element, data structure, file transmission and business processes around that functionality. 

DR. RIPPEN: I think there was another nuance depending on different people just talking about it, which was that the common data model would be important for federated, which is different.

MS. KLOSS: We can’t develop standards for each model.  How you shimmy it is not the issue. 

MS. GOSS:  It is what you are shimmying and what the rules are around it.

MS. TURNEY:  Even with a federated model, I just want to add to that, at least the one that I am familiar with that they are using in Connecticut, basically you don’t have to change your underlying data warehouse.  What you would then do is they defined a common data model for the virtual space. Then you create an index that links to their data models, so that what they are looking for it knows where to find.  Then basically, that is how you can expand out from there that particular model that they are using in Connecticut is relatively low cost.

When you are adding a new hospital or a new provider, you are talking about 20 hours to 50 hours of somebody’s time versus a lot of time and money if they were going to create data extracts, have to format them a certain way, have to have them aligned a certain way.  Now that does not yet speak to the second part of it, which is the data normalization because your expectation is if you are looking for maternity, you don’t want to see maternity diagnosis for a two-year old.

At the end of the day, there does still have to be quality checks and threshold checks that also should be common.  I will tell you that with most of the APCDs, we were one of the first payers to ask for that data.  In the beginning, it was like, well, what do you mean?  I said, well, if we are going to submit data, I want to know it is valid before you get it, not three months, six months and sometimes it was a year later. That is very important to have your quality standards, your threshold standards and your data model up front regardless of whether you are going to centralize or have decentralized.

MS. KLOSS: That is what I meant.  That could be a common layer.

MS. TURNEY: Exactly.

DR. RIPPEN: I have a question from the perspective of the employers and the self-insured employers and kind of this notion that there might be an employer-based data warehouse.  I guess I just wanted to get some feedback on I know there is a lot of sensitivity from employees and employers and the access to clinical data.  It has been around for quite some time.

Is there a concern or, with regards to that as far as the employers having potential access to perceived access?

MS. BROOKS: Yes, there is a concern when you implement, ask for biometric screenings and health assessments. Them knowing that there is this aggregated database, I am sure will pose some level of concern.  We haven’t heard directly from employees to date.  But the reality is that the employer is paying the majority.  They want to provide the employees the most optimal offerings to benefit them.

That is something that employers will have to navigate for sure because it is increasingly becoming challenging for them to be the payer, but also be the consumer advocate without having to address those concerns.  But we haven’t heard anything directly yet from employees regarding that.

DR. RIPPEN: There other thing is especially once you start consolidating, the health care industry, there has always been concerns of cherry-picking. If you are a sick employee, and all of your employers are sharing data about health and cost, again sometimes it is more perception.  Again, as you think about the strategies, it is something to consider.

MS. BROOKS: Employers are very sensitive to that. The employers aren’t seeing the identified data.  They are getting it de-identified.  From the aggregate level, as a coalition even, we don’t see any identified data.  They are very sensitive to that. 

They do rely on their vendors and suppliers to see that identified data and to make the recommendations and create interventions.  I rarely hear of an employer who actually knows or gets into that level of detail for their own risk management.  That is something that is a common practice today already to not get identified data.

MS. THORNTON: This is Kristy.  I would echo that in terms of the de-identified data.  I can comment also from the CHPE perspective of having self-insured claims included in there.  No employers have any access to any identified data.  Third parties again get de-identified data.

It certainly is a concern.  But I think there are some mechanisms that are already in place for the fully insured, claims that can really address this issue.

DR. COHEN: I would like any of the panelists who wish to sort of address an area that is again my concern.  Most of you spoke about the importance of the APCDs from a policy perspective with regard to health care costs and health care quality and consumer education.  Actually, Eric even mentioned that this is a claims database and might be less valuable for my interpretation, less valuable for other public health applications.

My experience is initially hospital discharge databases were developed for rate-setting purposes.  They became one of the fundamental sources of information for measuring morbidity and populations because we had no other database.  Vital statistics were originally developed as systems of registration of births and deaths. They have become a core surveillance system for public health risks and outcomes. 

I wanted to get your experiences and perspectives on the potential applications of APCD for population health purposes in your states.

DR. STEFFEN: We are seeing interest on the part of hospitals to use the APCD for their ACR-required community health assessments.  We have one large system that plans to use it for three of their hospitals. 

We are also in discussion, but have not moved forward, with the state hospital association on making our data from the APCD available for more broadly for that same purpose.  I think on that domain, I think there is a lot of potential.  In my judgment, the first community health needs assessments were sort of amateuristic, at least in Maryland.  But now, I think hospitals, as the new model takes some traction, are paying a lot more attention to it as they realize managing the health of their communities is going to be something that we are going to take quite seriously.

The second issue in terms of using the APCDs for sort of assisting clinicians is something that has been debated considerably in Maryland.  It is my feeling, at least in terms of the reporting periodicity that we have, which is quarterly, that it is probably generally not that useful.

As I mentioned, we are working with some of our administrative networks to test the idea of expanding our encounter and notification system through the HIE to pipe certain information to a pilot of practices.  I think there is thinking about how claims can be used for clinical interventions.  We are not there yet.  I think on community health need assessments, there is a lot of potential.  We are starting to see some traction.

MS. HARRINGTON: Maine Health Data Organization doesn’t only just collect claims data, but we also collect the hospital discharge data.  What we are seeing over the last couple of years is users that have historically used the hospital data like the Maine CDC, like our hospitals that are doing the population health work, they are now looking for the claims data.

What they are interested in is integrating the two datasets. That is something that I actually just had a retreat with my board of directors a couple of weeks ago.  That is one of the things that we talked about is looking at all data sources.

We also get questions about why aren’t we integrating vital statistics, worker’s comp?  Can the Maine Health Data Organization become the state’s data center, where all of these things are integrated including some definition of clinical data, whether we actually integrate it or I think, as was stated by another speaker, working with your health information exchange to bring those datasets together when you have a use case that all agree is an acceptable use of the integrated data.

I think it is something.  In fact, on page six of my testimony, there is a list of the strategic priorities that my board has come up with.  Number six is seek out opportunities to collaborate and advance the use of health data.  It is much broader. The way people have used health data five years ago is very different than how we are using it today and how we are going to use it five years from now.  We are not even prepared for it.

It really is trying to figure out where are we today and keep a nimble structure, so that as the needs change, and we can begin to start integrating this stuff, it seems like it is easier said than done.  But I think we are there. This will be a priority for us over the next 18 to 24 months.

MS. LOVE: To follow up, the question was what are the main barriers to linking that clinical and claims data.  The first one is, if you don’t get identifiers in the dataset, that would be the first barrier.

MS. HARRINGTON: Quickly, in Maine, because as I mentioned earlier, we now do have a law that says we can get identifiers.  Years ago, we did a pilot study with our health information exchange to see what was the match rate of taking a de-identified with an identified through probabilistic matching.  It was pretty high, but now that we have an identifiable dataset, they have an identifiable dataset, the linkage will be a little bit better. We don’t all get consistent. There are all challenges with that, as you all know. 

I would say the biggest barrier with linkage, and I hope this is not offensive to anybody, but it is politics.  We have been talking about this for years.  Technically, it is possible.  Anything is possible.  That is not the problem.  It is the politics of doing it.  I think we need to lead by doing, and we need to push through this.  But that is our experience in Maine.

MS. TURNEY: I want to add to that because we have actually been working with some of the HIEs to bring data in.  One of the things that we were really surprised by is that when we were getting basic data like height and weight, none of the units of measure were coming across.  Some of their own data were pounds, others were grams, others were inches.  It was all different. Yet, they didn’t know it until we told them.

Some of the types of ways that we are utilizing the data need to be kind of built into the data model.  They have been collecting data in this particular HIE for at least three years, but not a lot of uses other than looking at an individual patient.  I think as we look more at utilizing the data in aggregate, and also one of the kind of issues with using HIE data specifically is you don’t often get historical data.  You may have historical claims data, but you may not have more limited historical clinical data. 

Then of course, trying to get out of the clinical data, the textual data that is basically the most rich, probably the clinical notes.  I would say those are all things that are going to pose equal challenges as we move forward. Even though there is EHR standards, every place we have dealt with has a different version of Epic, a different version of the software.  They don’t talk together.  They are not common.  Things have changed that even those vendors who have versions don’t keep track of.

MR. BARRETTE: To the original question about how to use claims data for public health type issues, it is possible to do those types of studies.  But it is important to remember the limitations that come with them.  In one sense, it is still good.  It is great to try and work to integrate the clinical data.  But with that comes a lot of challenges and potentially a lot of data problems.

On the other hand, you can do a study of heart disease prevalence from claims.  But you need to remember that there will be people that are undiagnosed.  There are people who are not using services, so they are not going to show up in the claims.  Then you need to sort of adjust how you generalize your results from there.

Claims are useful in a broader sense than just cost issues. Understanding the limitations, which comes from understanding the data, is really the key.

DR. PHILLIPS: Briefly to Bruce’s question, I actually use Maryland in a book chapter on integration of primary care and public health.  They have actually developed something called the State Health Improvement Process, where they are using population health data and claims data in a mapping platform to help create local health care improvement coalitions.  It is not only forming CHNA, but it is informing how they actually address problems.

DR. SUAREZ: Vickie, you have the last question.

DR. MAYS: This is for Doris, is Doris still on line? 

DR. SUAREZ: All right.  Thank you so very much again for your graciousness in adjusting our schedule for your testimony.  We are going to go for lunch break until 1:45. 

(Recess for lunch.)

Afternoon Session

Agenda Item: PART 2: Federal-State Issues

Panel 2: Federal Panel

DR. SUAREZ: We are setting up to go ahead and get started again.  Thanks so much to our three panelists for your graciousness and flexibility in allowing us to move this portion of the discussion to after lunch.

I know we have also some additional constraints so we will start with Deb, and then we will go to Trish and Michael if that’s okay.

MS. LOVE: I wanted to say that this is an important panel in that as we heard this morning that some of the state initiatives, there is a lot of overlap now with federal agencies as the APCD start developing in states.  So SAMHSA was invited to be here but they are in a rule-making process so they may be listening in or they indicated that.

Then we do refer you to a testimony that OPM did submit in response to this panel, but they will not be here to speak so I just wanted to call that out and make people aware of that.

MS. SCHIEL: Thank you.  My name is Deb Schiel from the Center of Health Information and Analysis, and I’m just going to touch on some of the issues related to 42 CFR part 2 if the proposed rule in fact get adopted and some of the challenges that will be related to that.  Number one, I think it is more challenging than the Gobeille decision.  I’m not quite sure how you pronounce that; I’ve heard it pronounced both ways.  Exclusions of substance abuse disorder claims would certainly prevent the state agencies and researchers to be able to study opioid crisis in Massachusetts.

CHIA is currently working with our sister agency, the Department of Public Health, right now to link the DPH databases with the APCD specifically to study substance abuse.  For instance, currently they’ve had some great success linking death data to the APCD data.  So we have some early successes with that and this would create a robust dataset that we can actually follow a patient from maybe early uses of opioids through some outcome measures.

Second, if the identification methodology varies by payer, depending on their interpretation of the rule, this could actually create or destabilize the standardization that we have already achieved in the APCD.  For instance, some payers may remove a single claim line that’s associated with the SUD claim.  Some may delete or exclude all claim lines.  Some may exclude all patients that are associated with the SUD claim.

Just a reminder, too, that SUD diagnoses and procedures are all over, can affect many service categories.  For instance, you can have a mom who delivers and on their tenth level diagnosis could have an SUD diagnosis.  So it doesn’t clearly just affect substance abuse treatment and diagnosis.  It can be kind of scattered throughout your dataset.

So CHIA is hoping that SAMHSA will consider the APCD Council’s recommendations and we hope that this will be — that they will seriously consider it and include state agencies as lawful holders of the data under their final rule.

Thank you.

MS. LOVE: Thank you, Deb.  We will probably have some questions and hopefully before you leave.  We’ll go to Trish.

MS. MACTAGGART: Hi, I am Patricia MacTaggart.  I’m with the Office of Care Transformation within the Office of National Coordinator.  One of the things that we were asked to talk about today was really the opportunity to work with the states and their stakeholders in developing and utilizing and improving their databases, including the APCDs, and I want to reiterate what I heard you say this morning of how impressive the work has been by states to date.

What we’ve learned, we have learned just working with the states.  So one of our key roles in the Office of Transformation within ONC is to be working with state innovation grant states who you’ve had some of them testify today, because they are indeed the leaders in some of these areas, on what their issues are in moving forward.  I did hear one of your speakers make reference to this, and it is true that we have heard long and hard from everybody that this is about, for payment reform and service delivery transformation, the absolutely critical need to be able to aggregate claims data and non-claims clinical data.

I say it that way because there is clinical data also on claims, but you cannot aggregate something when the data source is not standard and available to everybody in moving forward.  So we have been working with states and different states are accomplishing this different ways.  Some are leveraging their current APCDs, and many states, as Jessica said, are starting with their Medicaid claims data and building into an all payer system as a way to work through some of their operational issues.

Some of them are leveraging their Medicare qualified entities.  Others are looking at other state regional efforts and others are forming something new, some being the APCDs and other ways of doing it.  But no matter what mechanism they are looking at addressing, there are some issues that have come up in no matter what option they are choosing to select.

Some of the considerations fall into some basic categories.  I’m just going to highlight a few of these.  The purpose of the APCD or the data repository — is it for public reporting, rate setting, or actually total cost of care?  The more the data is being used for actual payment, not just for quality reporting, the more significance everybody is placing on the quality of the data going in because it’s used as a real impact on the financial line.

The other issue, and Deb just alluded to it, it’s what data is included in the database.  Clearly most are dealing with the medical, the pharmacy, and some of the dental claims, but the substance abuse use data is missing from a lot of the datasets due to at least perceived issues with 42 CFR and real issues with 42 CFR.

Then last but not least, that we hear a lot about the issues of what can be leveraged on the policy in legal side in order for payers to submit information.  You’ve obviously heard earlier the mandatory versus voluntary, but there’s also a lot of discussion about what can be leveraged through contractual relationships.  What is the purchaser’s role in writing things into contracts that will allow or require data to be moved into a uniform database?

Then last but not least on the business operations side, there has been a lot of debate about de-identified and identified information, and it really breaks down into two things.  There is whether it’s de-identifiable or identifiable data going into the All-Payer Claims Data system, which is very different than the data being released from the database being de-identified, and sometimes those are not clearly separated and delineated.

A couple more areas that I’d like to point out have to do with, and you’ve heard lots of talk about this today, the data quality and the data source, the whole issue of mapping as there are the ability for the provider to send the information to the health plan or the ACO and then onto the all-payer data system.  There is the issue of the mapping and the quality of data that comes out of the mapping.

On the data extraction and transport, transformation, and aggregation, obviously you heard earlier the role of the health information exchange, but it’s really some of the core mechanisms that deal with other things that you heard in other testimonies, identity management, attribution, that affect the quality of the data going in and out, and that capability needs to be there as transported from the individual data source onto the all-payer system.

Then last but not least, it’s the reporting and the consumer and provider tools, because the value of the data, as we’ve heard from multiple of you, is how it gets transported out in a user-friendly way that people can use.

This slide I’m not going to spend any time on, but it really is health IT in the context of the all-payer or alternative payment model framework that has been adapted from the Health Care Payment Learning and Action Network and I think the importance of this is why we are doing this.

Again, the emphasis right now on moving forward from every state that I’ve worked with and whether it’s at the provider or purchaser level, is we are talking about alternative payment models that providers cannot operate effectively and without additional information.  Purchasers and regulatory need that additional information as well.  That environmental change has really brought some other things to the highlight, and I’m just going to highlight three of these.

One is the lack of interoperability and the lack of standardization for the claims and encounter data and the clinical data, which is clearly an impediment to the data synthesis.  There are also multiple approaches, there is no one approach that is taken for the standardization and the data aggregation across the alternative payment models.

But the last one is probably one of the biggest challenges, which is that organizations are treating identified data as proprietary assets and how that affects what’s in the databases is also critically important.

In the near term, I think we all acknowledge that existing multi-payer claims and clinical data aggregators need to be leveraged and again, the importance of the individual pieces to combine in aggregation, the all-payer claims data systems, multi-stake holder HIOs, private HIOs, qualified entities, qualified clinical data registries, all of those need to be able to work together and it really highlights to one of the clear pieces of no matter which marketplace or regional approach we have for aggregating the data, the ability to have it standardized.  And as I said earlier, it’s not just the data elements and the formats, it’s the ability to do it efficient and effectively that is critically important.

Our role is to help work with states to enable the multisource data kind of aggregations.  We are looking towards, as states are, to the multi-payer claims and clinical data aggregation.  This is not just public or private, it’s how we work together to make this work.

I think the last thing that I really want to emphasize is again how all of us work together to use policy levers that are available to us to move forward to facilitate both payer and provider participation in the qualified data aggregators.

DR. SUAREZ: Thank you so much.  Mike, your turn.

MR. LUNDBERG: Well, first of all I want to thank you.  I got my own microphone.  Usually people just take it away from me.

(Laughter.)

I will try not to abuse the privilege.  So my name is Michael Lundberg and I am the executive director of Virginia Health Information, a non-profit 501(c)(3) organization located in Richmond, Virginia.  I’d like to share some thoughts today about Virginia, the all-payer claims database that we established in 2012 and why they are so important to consumers, policymakers, and others.

Let’s just step back a little bit.  In 1993, Virginia became the 38th state to establish a hospital patient level data system.  It was based on the UB-82, which some of you may realize, that was an early standard of individual patient’s hospital care.  These data were eagerly anticipated to help businesses and consumers make better, more informed healthcare purchasing decisions and to support public health and policy.

As healthcare costs continue their upward spiral, employers and payers were looking for ways to control costs while many assumed that the quality care was good wherever it was provided.  Hospital discharge data did not disappoint those yearning to get a better idea of what care was provided to whom and as a general idea of the outcomes.

These administrative claims data are still very valuable for describing the care received in hospitals.  Some states have also established outpatient surgical care databases to further fill in the gaps about what is provided versus what is reported.  While many have recognized the importance of understanding the care paths taken to all care provided to an individual, it has only come out in the more recent years.  It’s not that these data were not collected.  They were simply not widely available to any but the largest employers and government programs and to supported research.

Today, we are moving down the triple aims winding road for better health, better care, and lower cost.  One approach of the journey is value-based purchasing.  Every foot of this road depends on comprehensive information on the use of services, their costs and quality for inpatient and outpatient care.  This information must be complete in order to be used effectively for population health evaluation, improvement, and monitoring.

Like many health departments, the Virginia Department of Health, for which they contract with us to do this and some other databases, has a key role in understanding the health of Virginians, conducting programs to improve access to care, and reducing disparities while being a good steward to funds needed to accomplish these goals.

Comprehensive APCDs are an important tool.  Health systems and payers need comprehensive information and ACOs.  Consumers and businesses need information on cost and quality given the rising cost in most cost-sharing consumers.  Virginia’s is a voluntary APCD, and it contains comprehensive information from all the major payers.  It establishes information on over 2 million fully insured and employer, self-employed, and individuals including state employees.

Medicaid beneficiaries are also included and number about 1.3 million Virginians.  These public and private programs can provide important insights into the health of Virginians and lead the way to continue the innovation in how healthcare is provided, to support better health, better care, and lower costs.

What VHI currently lacks is information on other government-sponsored programs, including about 2.6 million Virginians with insurance coverage through Medicare, Tricare, and FEHB.  These three are very important because of the demographics of the population of Virginia, the fact that there’s a lot of military folks and state and federal employees around here.  So it’s a critical issue.

We are very pleased at the effort of CMS to potentially expand our ability to use Medicare data through the Qualified Entity program, and we’re pleased to say we were just recently conditionally approved for meeting the minimum requirements as a QE.  So we will proceed full speed ahead through the different efforts that are necessary in order to secure the information and utilize it for public reporting and other uses.

Comprehensive information is important to know about the health of older Virginians, to help them stay healthy.  Similarly, we are pleased that we may also be on a path to acquiring information including FEHB program information and look forward to working with them to responsibly access these data in a clear and responsible manner.  We also believe information for APCs can also benefit government programs and we do provide information back to those submitting claims data.

We do this through a variety of analytic tools that are established through our vendor and refined through our use and our user’s needs.  We understand that across the country, payers are asked or required to provide APC data in a variety of formats.  Clearly, this is burdensome.  While many states have very similar requirements for data submission, we do support a standard set of information in the same format across states.  We should not assume that payers have unlimited resources to make varying requirements across states.

APCDs are an important tool for Virginia and other states to help consumers, to providers, based on cost and quality.  In Virginia, our health department is using the APCD to understand the burden of diseases such as diabetes and heart disease.  APCD information is also actively being used to support efforts to address the horrific increase in deaths from opioids and addiction and the many issues related to behavioral health access and treatment.

I’ll say this in aside, in Virginia, there are more people who died last year from opioid overdose than there were automobile accidents.  So it is a growing issue.   I know many of you dealt with it also.

While resources to operate APCs are also always a concern and vary by the state, complete information from all government health insurance programs in a standardized fashion, I might add, will greatly increase the value of APCDs to all stakeholders.  We support the ongoing dialogue to increase the utility of APCDs while addressing the concerns that burden submission, and the challenges we face and receive.

Thank you.

MS. LOVE: Thank you, Michael.  So, Q&A.  Have we worn everybody out?

DR. SUAREZ: I’ll start with a couple questions and we can go from there.  Thank you again.  Great testimony.

I think my first question is more generic about quality measurement through claims-based databases and it is really about how we as providers and payers and others are more and more required to report all sorts of quality measures to different programs and systems and certifiers and accreditors. 

Now we are seeing some of the quality measurement activities converging, at least in some of the federal programs, through MIPS and those kinds of initiatives, but when it comes to deriving quality assessments, quality measurements, using claims-based databases at state level or subnational levels, it creates certainly the possibility that we will have different types of measures compared to the ones we report or the ones that are reported in other places.

So how much of that inconsistency or difference do you see and what can be really done about it?  In other words, if a system is reporting to NCQA, the Joint Commission to CMS, to Medicaid and then to the — and then the state is aggregating data out of the claims and using that to generate quality reports, for example, how much of the inconsistency across all those quality measurement and activities are you seeing are happening and what can be done?

MS. SCHIEL: Again, I’m not the person that is doing that project, but we are working with the health policy commission in Massachusetts for actually accrediting ACOs and its longer term project, but we are trying to see how many measures we can actually produce off the APCD that are administrative, and they are the basic HEDIS measures, nothing fancy right away, but our initial assessments look like it could be actually for a good set of measures, doable for the APCD.

MS. MACTAGGART: I don’t know that I would use the word inconsistency.  I think we’re on a trajectory and I’ve heard this from all the states and as you said, on a national basis, we are trying to align quality measures, but part of — if you go back to HEDIS, we did HEDIS measures and I was one of the original cochairs so I can say this, based on what we were able to get.

My comment earlier about working on clinical data from claims and clinical data from other sources is where we are going to.  The goal is to be getting the best data and doing it the most efficient way of doing it.  So if you can take claims data and add some lab clinical data to it, you might be able to do measures that we couldn’t do before. 

So if you think of inconsistency as a trajectory, yeah, we are transitioning from the way we’ve currently done them and trying to — I think the one consistent thing is we are all trying to move away from strictly medical chart audits to find the most efficient way.  But I don’t know necessarily that I think that is an inconsistency.  Some are on a faster trajectory line because of where their state is and their providers are with electronic health records or the ability to exchange clinical data from registries, and I think everybody wants to go there.  The question is the timing for doing that.

So I said, one of my first comments was and I continue to say this, this is not about claims or non-claims data.  It is about how we use both of them the most efficient and effective way to get to the quality measures we want to be doing, and as states have already started to move to social determinants, those other social factors that — and I call them services that impact health — they are not in our traditional data sources.

So we are going to need to aggregate data from multiple data sources.  So the need to have standardized definitions and formats is going to be even more critical so we don’t lose the value of the information in the process of getting there.

MR. LUNDBERG: I like Patricia’s comment about trajectory.  Walter and I go back a long, long ways, when we were both working on different aspects of this.  When we first started developing the information on hospital outcomes, there were no standards.  There was no NCQA.  There was no NQF.  They didn’t exist.

There were a lot of comments about varying folks who were foolish enough to stick their necks out and do these types of reports, did so at their own risk, and we decided early on that if we were developing things on our own at that time, we had to do it with full stakeholder involvement.  We had to have them help us with the concept, with the use case, with the methods, and with how we would display the information as well as put it out.

We still follow with that, but we don’t have standards.  Well, what we’ve adopted in recent years is using information that’s either NQF or NCQA endorsed measures.  So on top of that, then we use a vendor now for the APCs specifically that is NCQA-certified to produce a number of the HEDIS measures.  So Walter, that’s going to help address the issue as far as how it’s interpreted.

I do have concerns sometimes when we see lots of different all-payer readmission rates and all kinds of other things where people are one-offs or people develop proprietary things.  I think the best thing we can do is insist that they open up and make those things available as much as possible.  It’s the only way you can have people, first of all, critique it to make sure it’s helpful, and then also be in a position where they can replicate that information for other things.

So to me it’s all about standards.  It’s all about rigor.  It’s all about making the best with what you have.  That doesn’t mean you’ve addressed any data quality issues.  I think that a lot of other speakers today have talked a lot about normalizing the information, the importance of that, and I think that’s another critical part.

MS. LOVE: I wanted to draw the parallel that hospital discharge databases evolved and as did the measures.  I see APCDs early in the evolutionary cycle, but the business case where standardizing is coming along, we’ll hear in the next session, but the business case for standard measures and how many states now are able to generate a total cost of care measure that is a similar methodology from the similar data?

As they do that, they learn about their data and how it may not or may stack up with other states and benchmarking, but they learn a lot about what is under the hood of their data.  But it is a road and the roadmap was used several times today.  So it is a journey, a roadmap, of maturity.  The APCDs are just in their early juvenile phases so I just wanted to point that parallel out.

DR. SUAREZ: Just one more question before I — yes, I think that is a great point and I think the other one is as we look into the roadmap into the future and we’ve been thinking about the past, the present, the future, I’ve been arguing a number of times how into the future, if we’re really going to depart from the traditional fee for service approaches for payment where a payment depends on really a billing process where services are billed and there is all that intrinsic interest in increasing the services, so they are increasing the bill so it increases the payment and all those things, we’re really trying to depart from that and CMS is taking directions around that.  Every payer is trying to look at that as well.  So we are going to go to an alternative payment model or some payment for outcomes or payment for performance.

How do you see the role of the claim, which in that case would probably be very different?  We would not necessarily have service-based claims or some sort of a fee for service claim.  How do you see the claim process evolving into that new payment model and the transformation of the dependence on claims for analysis to the dependence on something much larger that combines, perhaps, some of the encounter level data that is contained in the claim, maybe it’s named differently, plus some additional clinical data?

So it’s really building up into the roadmap that process of, today we depend on claims; tomorrow the idea is we would depend on something different because we are not going to be doing fee for service billing.  So how do you see that transition moving into the future?  Or do you see that transition?

MS. SCHIEL: It is going to happen and we will have to keep up with it.  Right now, we are seeing it under the claims and for each payer it’s very different.  Sometimes you will see that they are actually zeroing out the dollar amount and it’s under one record.  One claim line is how they’re doing a bundling.

So it varies by payer, so you have to be really tricky about making sure if you’re going to be doing utilization measure, for instance, you’re grabbing all those claim lines so you are getting all the days for an inpatient admission and I think groupers will help.

But it is, I think, being very nimble and continually talking to the payers.  We do do surveys with them.  When we validate the data, we do have these questions for them.  How are you doing global payments?  How are you doing — what changes to the claims are coming in for alternative payment models?

And they are still submitting the claims and they will have explanations for it.  A lot of times it is outside the claims right now, so it may take supplemental data for a little while to be able to catch up, but it is something that we have to kind of keep track of and track with the payers.

MS. MACTAGGART: I think one of the things Deb said is very important.  We are talking about alternative payment models and they are here now.  So it is not like we haven’t been doing prenatal care global payments for a while.  We have found a way to track data.  Some of it will be — it’s not about claims data or not claims data.  It is what we can pick from the data sources that we have and have them standardized enough that we can aggregate them in a way that produces real actionable information, not just more data.

And we are going to have to look at things differently but even our measures will be different as we do episodes of care and more whole person care.  We are setting whole cost of care, as you said, we weren’t even able to set those without being able to look at data across multi-payers.  But now, as we operationalize those, our ability to look at things differently is going to require us to step back and not just say there is one answer.

Again, I go back to trajectory.  We are going to be more dependent on one or the other over time and less dependent on the other as we project, but that’s not a one-year or three-year.  That’s a five- to ten-year kind of process.

MR. LUNDBERG: Slowly, that would be my thought as Patricia alluded to.  There will be a time where it becomes a greater and greater concern.  At the moment now we have some payers that for most outpatient care, they are mapping encounters to the claims, but there is no payment information in there.

We have a standardized proxy that then we compute from everything else to put in there.  It’s not good for comparing the prices of the individual payer, but it is a better idea of global, of what you would get.  Virginia is also in a position where we cannot release actual payment analysis from a payer to a provider, but we can do things in general as far as the burden and cost of care.  So we would utilize those things.

I believe there are already groups within APCD Council that are talking about modifications to be able to allow for this type of payment and have bundling and other payment amounts.  I think they will be slow in development.

MS. GOSS: Sort of a nice segue into what I wanted to ask, Walter.  Thank you.  I noted that on the agenda that the three of you are not slated for the next panel, so I am going to ask a question related to the next panel.  So it is related to your perception of the type of standards we need and how you are using that term.  We heard earlier the thought process. 

Some of the discussion talked about it’s not only how we exchange the data, federated or decentralized versus centralized, but also the structure of the data, the business processes around data use agreements or even the — a comment before, earlier, about how you do global payments, those kinds of things.  So from your perspective, what kind of global standardization would you like and what are those categories of things we need to be thinking about?

MS. SCHIEL: That is a great question.  I had worked on the mapper under an EHR incentive program where we had 13 states develop our EHR and sent the payment systems so payers could actually apply for the system, so 13 states.

Jess Khan was a big proponent of it and hugely helpful in it and we saved tons of money.  We sat at the table every week or over the phone and we really got to leverage everyone’s needs because again, the final rule still has interpretation and everybody has local needs, but we wound up building a system with HP setting it up and being — starting out sending incentive payments out to providers very quickly using that model.  So I am a huge proponent of that.  I think you can — people have different questions or different phases of development and that actually accelerates everyone’s learning curve on that.

I also feel that there are going to be some issues like Medicaid or Mass Health File or eligibility file, which was spoken about this morning, we had to tweak because it did not meet the eligibility requirements of the commercial population.  It’s just so unique that the only way of doing it was we developed an enhanced eligibility file that we’d be happy to share if it fits anyone else’s needs to be able to handle those issues and be able to report effectively off of Medicaid populations for their enrollment.

So I am a huge proponent.  I think other than a few weird little fields here and there that may be particular to a state or to a region, I think a standard approach is — I think you can get 99 percent there for most APCDs and I think especially big payers would appreciate it, that they aren’t doing it.

I think we are participating in the DOL for the Gobeille standardization dataset and I was thrilled because there aren’t data fields that you have to worry about that you would never see anyway really focusing on the solid ones.  So I am happy as an analyst.

MS. GOSS: So as a standard, you are really thinking about the data structure, the data content that gets exchanged.  That’s a key issue for you.

MS. MACTAGGART: I would say it is data content, it’s data structure, it’s data transport, it’s data extraction.  I think you get the efficiencies by standardizing those elements.

The other piece that you said that I would really like to build off because I think it made a really good point for doing it, and the way I would say it is that we are not talking about even in quality metrics that everybody has to do all of the quality metrics, but there is a core set that if you’re going to use, you should use them in a standardized way for doing it which does not mandate that every commercial is going to do every one that Medicaid is going to do.  They may have it more enhanced and bigger, but for those that Medicaid and Medicare and the multi-payers are doing, those are the elements of standards.

MS. GOSS: Setting a floor, a base requirement —

MS. MACTAGGART: I don’t even know if it’s the word floor because that assumes that everybody is going to do these four or whatever the floor is for doing it.  My point is more of where you are going to do them, whether you’re going to do three or if she’s going to do six, you don’t need to get necessarily for standardization agreement that you are both going to come to five, as much as for those three that you’re both going to be doing out of the six, those three are done the same.

I think the other thing that came up multiple times this morning, because you asked the other question, which is what are those kind of foundational things?  Rules of engagement/governance is core.  You can’t do that without some core functionalities and foundational things.  I think the other one is sustainability that comes through.

Obviously, this costs money, so financing is part of that sustainability.  Business operations is clearly important at multiple levels but I consider those as almost foundational.  Then within, above those, or in addition to those are those things that you do standardize and it’s things like identity management; it’s things like being able to attribute that patient to a provider which becomes really critical not only for quality metrics and for payment, but for service delivery.

And again, the goal is to get this data so we’re collecting the data once and using it in the way that you need to do it.  So for new alternative payment models, for example, we are trying to avoid hospital readmits.  Part of that is a quality metrics, but part of that is to make sure that the provider for a hip replacement when the hospital discharges, it’s not only the primary care doctor that is related to that patient gets that information, but physical therapy or the rehab center that they need to go to and in some cases, home-delivered meals.

So it’s again thinking of these data elements and standardizing them so when we’re using them for an all-payer claims data system, we’re not getting them from a different set of standards the way we are for other purposes.  Hopefully that is helpful.

MR. LUNDBERG: I think it is an exciting conversation especially if we broaden these beyond the APCDs but I’m actually going to limit my points to a comment about APCDs.  So we are a big proponent of standards especially with medical, the medical professional, the eligibility and the provider claims.  When we first started out, we were actually presented with a record layout by the payers that they had developed.

We then worked and tweaked that so if they had a standard they were all willing to adopt, we blew it from the very beginning in that we added some elements, but primarily just payment information in some of the other things.  But we are a big proponent of that.  I’ve been listening to our friends of the payers for a number of years about this, and we continue to say that we are willing to do this.  I think it’s really time to get off the ball and move with this, and I’m excited about the potential to that.

I think transport is another very good thing, secure transport, so everybody understands the security associated with data when it’s submitted, when it’s received, when it’s at rest, and when it’s stored.

I think EDIS is a really interesting thing and I think standardizing that is something that NAHDO came up with ideas on standardizing EDIS years ago and they were very instructive because we do sit in our holes sometimes and think that we’ve got everything taken care of because we did it that way for 15 years.  Well, the world has changed and things have changed.

So that would be very good.  Some challenges with that are some APCDs are mandatory and they have teeth if there are problems with data not coming out of processing verified.  Some are voluntary.  If they don’t have, they can point the information about to the payers, and the payers I believe in good faith are trying to correct data when it’s within their realm.  But the point was made very well today that they are in the business and many often are paying the claims and they are punished through state laws if they don’t get the claims paid in a certain time.  So that has to be considered when you do that.  There could be one state with incredibly high debits that just won’t work in other states.  So that dialogue I think could take some time, but we certainly would be willing to start that pretty much now.

Output, again, we really touched on that a little bit about standardizing the output that you have.  I also agree with the fact that you wouldn’t necessarily say that every group has to do these 30 measures, but if they’re doing those measures, they should do them in a different way.

The other consideration is if you look at what the NCQA has done for years with their measure, some of these were based on administrative data and involved extraction.  So you’d have to note the difference whether you have the ability to bring in clinical information or not.  There could be a measure version A and B and that could potentially help stimulate movement more toward having clinical measures in there because I believe they will be added.

DR. COHEN: That was a good segue into the comment I wanted to reinforce that hopefully we’ll get into a little more after the next session, but since as you mentioned, these folks may not be here to discuss standards.  Denise, Walter, and Mike all talked about using the hospital discharge case mix model for the development of APCD standards. 

There is a much more fundamental and I think robust example which could be easily applied, and that’s vital statistics because it is a federated system that has worked out issues around standards for data collection, variable definition, quality, transmission, and storage all through negotiation between the feds and the states and NAPHSIS, the organization that represents vital statistics.

There is the VSCP, the Vital Statistics Cooperative Program, that sort of governs this interaction.  They’ve created the actual variable definitions.  NAPHSIS has developed an interjurisdictional exchange protocol that allows transmission of data from one organization to another, whether it’s states or the feds or other organizations.  They have developed something called STEVE, which is the State Territorial Exchange of Vital Events, which is essentially a federated hub that gives permission for different folks to use the data.

As we move forward thinking about recommendations and solutions, because everybody agrees we need to move forward in standards, I really recommend that we thoroughly investigate the potential use of the vitals model as a really explicit example of how a lot of these issues have been successfully addressed.

MS. LOVE: That has been my vision for years and we even proposed STEPHIE which is the hospital exchange for public health interchange and states do work through the NAHDO network to share best practices and standards, et cetera.  The problem is we don’t have a federal partner or any funding to make that happen but I think the states would — I think your idea is a great one and if there’s a way to make it happen, I’d love to explore that.

MS. GOSS: And it really builds off the idea that I think we heard from the activity in Massachusetts where you’re extracting out of the APCD the information that you need for the risk adjustments.  You are doing that reporting on behalf of the — it’s a value return for them, and the conversation has been really clear there’s value across a multitude of stakeholders and from a sustainability perspective, we’ve all got to come together and help build this brave new infrastructure across the country that will let us get at this and we can leverage what’s already there and build on it, it’s going to go a long way to helping with educating people in the health claim process.

We have enough challenges with HIPAA, minimum necessary education and understanding based on yesterday’s hearing, we’re still trying to catch people up on using all the robust HIPAA transaction sets.  We have added HITECH.  We’re still going and maturing in those standards and use of them and how they’re deployed in actual products.  Now we want to add more on top of it on how to start to really harness this uber data to give us better insight on how to control our costs and get better outcomes and we are all sitting around here.  We’ve been thinking about it but we’re a microcosm of the rest of the world that’s really just trying to make business happen on a day-to-day basis.

MS. KLOSS: It has occurred to me as I’ve listened today that maybe what we’re calling this is wrong, all-player claims database, when we know, as you say, Patricia, that this is a trajectory and we may be at this stage now but it’s not going to be sufficient.  In some ways, it’s a public good or public use database strategy that may be beyond how we’ve viewed it but again, let’s not make the elephant even bigger.

(Laughter.)

In reality, we probably need to be future-focused on how this goes forward, not thinking about solving today’s problems only.

DR. RIPPEN: So I think that there are a lot of interesting questions and especially if we think about the transition from payment only to informed to clinical, and we have a lot of quality reporting and a lot of different demands.  We have PCORnet research infrastructure that has research data.  There is a lot of interest, a lot of different players, and again, going back to the focus because one is what’s the business of and what’s the structure, which is more a business question, as opposed to what are the guidance to address the challenges of interoperability or clinical data model?

Do you have any thoughts about prioritization?  Because the background is that I do know that even in research, there is a competition between different clinical models as one kind of aspect.  So if we are going to move on in a stepwise progression, what is the low-hanging fruit and what’s the glide path?

MS. SCHIEL: I think right now where Massachusetts is trying to tackle the opioid crisis, so I think working with our partners at DPH has been something we were really effective — our master patient index is very effective, and so we’ve had some early successes in linking those data.  We are hoping we’re going to get ambulance data, pharmacy monitoring data, and just link it and just be a very robust opioid tool to affect the opioid crisis.  So that’s where we are right now.

MS. MACTAGGART: I don’t know that I have a good answer to yours except to say what I see happening at states is the value proposition, which is your opioid, of where the focus is, where do I need action today.  If I am required to do a total cost of care in my payment model, that is my priority.

But going back to what you said, it really is getting that trust level there that requires both technology, which is the infrastructure, but also the data as you said.  It is the extraction, it’s the transport, it’s where it sits, it’s the data elements for doing it, and the more efficiencies we get there, the more trust we are going to have for the exchange and the use of that information.

Because I think to sum it all up, it’s really about taking the data that we currently have today, putting it in an efficient way and making it actual information.  That’s what people are trying to do but I don’t know that I can tell you which is going to be the biggest priority.

MR. LUNDBERG: Were you talking about organizational priorities or just those associated with interoperability?

DR. RIPPEN: Whichever you think is the driver because again, going back to some of the nuances, there is the trust and policy, then there’s the data and the prioritization.  So again, they are all kind of like spaghetti with the sauce.

MR. LUNDBERG: One certainly begets another as you go through.  When I say that we were established in 2012, we didn’t start getting information until 2014 and spent quite a while working through that.

So we are still so data-focused.  Our heads are down trying to do things like we’ve had healthcare pricing information, not by region, by procedures, that we got directly from the health insurance companies and we combined all those.  So it wasn’t the data, it was averages.  We are now working through a taskforce to develop that information just using APCD.  So that’s a priority.

Because of the decision at the Supreme Court, we are working very, very hard to return self-insured information back by working with the payers, trying to encourage them as they are going — if they’re dealing with an opt-in to ensure that they have information about the value to employers who participate.  So that’s a priority that we have.

Improved provider identification is a really important fashion.  We have about 120 users of the information through an online system we have and the provider information is typically one of the most difficult ones, and so we work very closely with those that are using it to understand what the nuances and to work with us to improve that.  So quite honestly, it’s only until we get all those things taken care of, particularly with the data quality and things that we’d be in a position that lead to better interoperability.

We have good NPIs.  The patient index is not perfect.  It’s probably within the realm of what most folks have, but you just ID one wrong person and that’s a bad thing.  So we try to cautious about that.

DR. SUAREZ: Thank you very much.  Thank you.  We are going to call in our next panel which is Jo and Bernie and Sheryl.

Just a little bit of setup, this is the panel we had reserved to discuss the issues related to standards and the use of standards for reporting.  So I think we’ll have a chance to go back some day, points and issues that people have identified throughout the day that relate specifically to standards.

So I think we are going to start with Jo, and do you have a presentation, Jo?  Just comments, okay.

Agenda Item: PART 3: STANDARDS: Background, emerging Issues, and Challenges

Panel 3: Overview of Reporting Standards and Previous Standards Efforts

MS. PORTER: Well, thank you.  My name is Jo Porter.  As I mentioned before, I am the co-chair of the APCD Council along with Denise.  I’m also the director of the Institute for Health Policy and Practice at the University of New Hampshire.

I’m going to take a little bit of time this afternoon to talk a little bit about the history of some of the standardization efforts that have been happening.  It was interesting as I was going through the process of creating this testimony of realizing that we’ve been talking about standards and doing work towards the standards of some way or another around APCDs for almost 10 years, which might sound a little surprising to some folks, but it’s been a while.

So I am going to talk a little bit about what that history looks like and where we are today and I think that sets us up for thinking about where we could be in the future and then I welcome questions and comments about that.

So in the early years of the APCD Council’s work, looking back particularly in the years of 2008 and 2009, state APCD development was growing pretty rapidly.  We were sort of adding a couple of states a year, if not more than that.  There was an expressed need for consistency in the APCD data collection efforts.

At the state level, that interest around harmonizing data collection was really to support collaboration more than anything else, I would say.  There was a lot of interest in sharing data and analysis tools, similarly to what we’ve talked about today.  So some recurring themes over the last eight years.  For data submitters, like my colleagues who are going to talk to you soon, the consistency was sought to better support efficiency in data submission especially as they were finding themselves needing to submit data in multiple states.

Back in 2008, the then director of the state of Maine’s Health Data Organization, Al Prysunka, who is in the room, which houses the Maine APCD data as you’ve heard from Karynlee earlier, actually reviewed the New England states’ data submissions and documented some common fields among those states really with an effort towards regional analysis and shared tools for common analytic needs.

Then in 2010 and 2011, the Council then compared data collection in six active states, the New England states plus a couple more that were collecting data, to determine if there may be a common set of data elements across the states.  At that point, the Council also convened a Technical Advisory Panel, as we called it, to build consensus towards process for the standardization of APCD data collection, talking with folks about what we should do and how.

That Council-appointed TAP, as we called it, included participants from Aetna, AHEP, ARQ, the American Dental Association, the American Medical Association, ASPE, CMS, Cigna, Harvard Pilgrim, Humana, Kaiser, Medco, the National Conference of State Legislators, some folks actually with representation with NCVHS, the National Governors Association, and United Healthcare.  The goal of that set of meetings was really to say, okay, here we are, this is where we find ourselves and where do we think we should go and what do we think we should do about it?

In that TAP and in then in the broader APCD community, there was a feeling that it would be valuable to identify a more formal standards mechanism to define the data submission standards and to work towards APCD data collection consistency.  This actually led to two efforts with formal standards bodies in 2010 through 2012, one with the National Council for Prescription Drug Programs, NCPDP, and Margaret can talk a lot more about the day-to-day, how that worked out, that focused on pharmacy claims data, and then another effort with ASC X12 for medical and dental claims, which Margaret could also talk about because she did both of those sets of meetings.

So that time from 2010 through 2012, the APCD Council joined the NCPDP work group one, the Post Adjudication Task Group, to draft what is now available as the Uniform Healthcare Payer Data Standard Implementation Guide Version 1.0.  That was approved in October of 2011 by the NCPDP Advisory Board and as the standard that was developed to support reporting requirements for pharmacy claims data submissions to state APCDs.

In 2012, the APCD Council approached ASC X12 as, again, the formal data standards maintenance organization to talk about the methods and approach for creating the standards for the medical and dental claims data submissions.  Historically, and we’ve heard a little bit about this already, state APCDs had referred to both the 837 and 835 as in the industry referent standards for medical claims data submission instead of having a single transaction or standard for post-adjudicated data reporting.

It was becoming apparent at that time that state APCDs needed a transaction for post-adjudicated data reporting and in meetings with ASC X12, it was discovered that in other conversations that they were having there seemed to be identified multiple business needs that required post-adjudicated data reporting.

ASC X12 stated in a release that launched this effort in August of 2011, and I quote, in addition to state reporting requirements, Medicare and Medicaid have defined business needs for similar claim/remittance data.  The information included in this business process is very similar to the information reported by plans to Medicare and state Medicaid programs for paid encounter reporting.

Therefore, ASC X12 convened a Post-Adjudicated Claims Data Reported Special Appointed Committee, known as the PACDR SAC, to define a standard to meet the data collection needs across multiple business cases including state APCD but also including Medicaid Encounter Reporting and Medicare Encounter Reporting.  That process, which happened over many months, over many calls, and many meetings, included reviewing the existing standards, the 837 and 835, as well as data elements from the common APCD core that the APCD Council developed as it reviewed the six states that I mentioned earlier.

In 2012, PACDR guides were published.  One PACDR guide around professional claims, another around institutional, and another under dental.  So I think it’s important for folks to understand that in addition to the reporting required state APCDs, the guides support many other business needs and the reporting of Post-Adjudicated Claims Data, including Medicaid encounter reporting.

So the guides support the reporting of the Proposed Core Set of Data Elements but as well, there are many data elements in the PACDR that are not typically included in any APCD activities.  That being said, the Council recommends using the Proposed Core in conjunction with the guides around state data collection rules to make sure that you maintain use of the industry standards.

One of the things that we’ve realized over time, though, is that there are also many state elements that are referenced in the PACDR.  However, as data needs have evolved, there are now items that are part of state data submission rules that are not in the PACDR guides and while ASC X12 supports a formal process for data maintenance, states are not required to submit data maintenance to ASC X12 and ASC X12 is not required to survey states for their data maintenance needs.

Last year, there was an NCVHS hearing in which we were invited to provide testimony.  That meeting focused on the hope to assess effectiveness and usage of adopted HIPAA healthcare transaction standards, code sets, and operating rules, and we testified a bit on the challenges of updating that standard, and what we said then, which holds true today, is the efforts of the authors of the ASC X12 guides and the state APCD representatives in the workgroups are all voluntary. 

States, many states, cannot afford to volunteer their staff for these types of efforts.  So their participation becomes limited.  In turn, key voices and knowledge of the business are not heard and incorporated in the development of the standards and guides.  While we do not necessarily have a proposed solution complete for this issue, it is critical for the committee to be aware of that.

In summary, we would say that supporting a process for updating and maintenance of the standards remains an opportunity for the committee to consider as you think about what will come out of this hearing.

To talk a little bit about where we are today, we heard earlier from Kristy Thornton about her work with the Pacific Business Group on Health and specifically about some work on the Center for Healthcare Transparency.  As part of the work for the CHT, the Center for Healthcare Transparency, they actually drafted what they were calling a model data submission manual to support some of the work that they do with regional data intermediaries, which have purposes that require administrative claims data that are similar to APCDs.

In an effort to standardize data collection for its regional data intermediaries, CHT drafted this model submission guide and incorporated references to NCPDP and PACDR guides as well as updates and additions to the core set of data elements that were developed by the APCD Council.  As an example, one of the data elements that is reflected in many current state APCDs or is proposed to be included is a HIOS plan ID.  In 2010 and 2011, those things were not considered important or not even available.  So that is an example of a data element that doesn’t exist in the current PACDR but it exists in many of the state submission guides now.

Given the historical experience with the development of the core set of data elements, the APCD Council worked with CHT in the latter process of its development of the guide and reviewed it at the end of 2015.

So where we exist today, the council and the states and the learning network.  We certainly find ourselves with a need, given the anticipated desire for more consistency, especially reporting self-insured claims data in light of the SCOTUS decision in Gobeille vs. Liberty Mutual, which we have heard a lot about, and reevaluating state variation in data collection.

So with permission from the CHT, the CHT model data submission manual is currently being reviewed by APCD states in meetings that are facilitated by the APCD Council to identify common data layout for APCD efforts that meet current needs of the states.  That common data layout continues to make references to the standards when they exist, but then also captures data elements that might not be concurrently available on the PACDR.

Once that data layout review period is done, hopefully in the next two weeks or so, we plan to share it more broadly in its draft form with the payers and other stakeholders in the community.

So in summary, I will just say that there are many demonstrated needs for APCD data at the local level.  We have heard that consistently today.  States have identified and responded to that need for information for many years and have done so in a way suited to meet the needs for health system improvement for their populations.  The APCD Council appreciates NCVHS’ willingness to better understand the history of standards development to date and the opportunities for continued work around standardization, and we look forward to future conversations to strengthen the data collection efforts at the state level.

MS. LOVE: Thank you, Jo, and I think, Sheryl, you have a time constraint.  So we will go right to you.

MS. TURNEY: Thank you again for letting me speak.  I am Sheryl Turney, again still senior director, All-Payer Claims Analytics at HealthCore.

Basically I will focus my comments on the standards.  One of the things that I mentioned as I spoke earlier was that there really is a need for greater alignment to APCD standards, and we believe that we can mitigate many of the challenges presented by the various people that talked today on APCD and our future efforts if we really focus on establishing, again, a technical data model that is common across the horizon, as well as a focus on data quality, data thresholds, and then the data content, which we spoke about earlier as well.

Because all of those things tend to be the things that as users of the data learn more about the data, tend to change over time, and one of the things that we — and I will use this one simple example — that we brought up dealing with a state early on was about providers.  You know, we had a lot of challenges inside of payers’ footprint, because providers were going through a big metamorphosis, and an individual provider in one of our states, and I’ll say Connecticut, basically went in a two-year time period from an individual to a part of a group to part of a different group to part of another group and then integrated into a hospital system.

So trying to traverse the NPIs that change as that provider moves through the system has to be something that lives, and it isn’t static and it is going to be constantly changing.  So standards will have to be able to address those types of examples.  I joked about it, but basically said if we had standardized on the checking account number, we would have totally gotten the right provider as we moved through, because they do always want to get paid.  But that’s aside.

But at the end of the day, determining what is the data element that is going to be common, because some things will always change, and other things will not, and essentially claims were used so that providers could get paid, and that was the essence of what is there, which is why it is traditionally financial.  But we are using those claims for a lot of different purposes now.  We are using them to understand the frequency of use that we have from our members and the services that they utilize, but again, not everything, not every claim, results in a payment, because now we are looking at outcomes.

So we are looking at different data from a payer perspective as well in order to create that standard, and currently whatever standards we put on the table really have to be able to report on outcomes in order to have our living health system really live and grow with us, and that currently doesn’t exist, and it’s hard, because that’s the data that is unclear. 

So claims have been easy, because that’s the data we have always known, but that’s not the data that is really going to be needed as much in the future when we are looking at changing our mechanism for changing behavior in terms of the way that healthcare is delivered, because we are moving away from that model.

So we are recommending — and I’m going to skip to the end — basically that we focus on what is the type of data that we need, which means we need to know what the questions are that we are going to be asking.  Many of the APCDs that were initially implemented knew they wanted to deal with reducing cost, increasing access to care, making sure of the quality, but they really didn’t know the use cases or the questions around any of those elements, and so in the first couple of years the APCDs were implemented, they were constantly changing, which is what makes the cost of sustaining them so expensive.

So really what we are saying is let’s work together to focus on what are the real questions we should be asking.  What are the behaviors that we need to have changed?  What is the learnings that we need to have in order to change the landscape so we are asking the right questions and collecting the data around those questions, and then doing it in a standard way across the landscape so we can bring together the clinical data, the claim data, because this is just the tip of the iceberg.

When I brought up before the HIEs, you are going to have a state APCD, as we do in Connecticut, with HIEs with the same data, with other third parties with the same data.  Then you have duplicate data all over the place, some of it identified, some of it de-identified.  How do you bring all that together to know what’s unique and what’s in common?

That becomes even more difficult when you’re dealing with all of these third party aggregators that we haven’t even talked about here.  So from our perspective, we are looking at it saying we need a way to identify data that maintains the privacy of the individual but also allows us to connect the data in a way that across the spectrum you can bring the data together.  We need to be able to define the certain data elements and the structure of those data elements and the meaning of those data elements, as well as how we are going to evaluate those data elements in order to have a living health system really work for us.

And I think with that, I’m going to be out.  I am sorry I can’t be available for questions, but you can always ask me later.

DR. SUAREZ: Thank you, Sheryl.

MS. INSKEEP: Hi, I am still Bernie Inskeep from United Healthcare, and I’m still happy to be here.  So thank you for giving me the opportunity to speak with you.

Talking about standards, emerging issues and challenges, reviewing current formats.  I think Jo really delved into kind of the history and how the formats came about.  A couple of observations on the formats.  While well-intentioned with the PACDR X12 and trying to determine a standard based on perceived needs, what we actually see from a health plan perspective is that state Medicaid areas are looking for 837s.  They are not looking PACDR 837s.  There is only one state trying that, and they are having a really high rate of rejections, because it’s technically very difficult.

The APCD core, which is a favorite of United Healthcare, really lends itself well to how payers store data.  Again, our data is in a flat file.  So the APCD core is designed to be in a flat file, which nicely corresponds with how health plans store their data.  So again, we get in this three dimensional object; it’s transformed into a flat file.  Think piece of paper.

So it is very helpful when we take that piece of paper and create another piece of paper to submit that onto a state.  Now our challenge is, of course, getting all of the pieces of paper to be the same, and that’s really what we are hoping for in order to meet everyone’s needs.

What concerns me from a perspective of ongoing challenges that we have had is that it appears that some states look at that core standard.  They kind of check off, yep, we have these fields, we have these fields, we have these fields, but their piece of paper, if you will, or their submission is just very different because of how they reconfigure it. 

So these things are all important, and I think that the good work that was done into the APCD core is really something that I have been asking repeatedly that we all coalesce around and really involve ourselves, because as we get together as a group, you know, if it starts with the states first and the researchers and they are thinking about the salad bar which is the 837 transaction and the 835 transaction and everything that could possibly be available, the early involvement of payers not to say no but to help design ways of getting to the goal in a way that is most reasonable for everyone would be very much appreciated, particularly by my company.

We also have some kind of emerging issues with our Medicare Advantage data.  We have some limited identifiers with Medicare Advantage.  One of them is the HICN number, which has — it’s the only really unique identifier, and there are regulations about if that can be transmitted and in what circumstances, and so we have some issues in some states that want the same identifier in all transactions, but yet they do not mask this identifier in the database.  So that causes some particular challenges. 

So it is not just what we are sending and how we are sending it.  It is also how it’s kind of captured on the other side, stored, and used.  Race, ethnicity, and language, that’s another challenge that is ongoing.  We had a gentleman that mentioned it earlier, that health plans need to do a better job with that, and we are not allowed to use that to transact a claim. 

We cannot incentivize anyone to provide us with that data, and we can’t penalize them either.  So we don’t have a carrot.  We don’t have a stick.  So we are in kind of an awkward position.  So we have been trying to work with states to try to look at other alternatives, but I just wanted to bring that up as another example.

Another kind of challenge that I think we should all consider as we talk about this whole seeking of standards is something called a claim version number.  Most states believe that health plans couldn’t possibly operate without having the most recent version of a claim that restates the entire claim so that there’s this really nice succinct picture of the claim.

Well, I am here to tell you that many of us have very aged systems that are workhorses.  They do a fantastic phenomenal job of taking enormous amounts of claims, processing them in a very timely manner, and accurate manner, and paying the claim, and in some of these systems that we have, they just don’t have a claim version number.  So it’s just kind of something idiosyncratic to be thinking about, because the differences are real, and so as we try to educate the states on how our processes work, then we have the states and their vendors dealing with all this variability and everyone’s unhappy.

And then there’s the approach.  So this kind of gets to kind of the policy question, are we looking at something centralized within a state?  Are we looking at a federated centralized model?  Are we looking at a decentralized model?  But each state approaches their APCD differently. 

So in each state, they have something called a threshold, and in some states, they have something called a de minimis threshold.  So if your entire company has this large amount of people and you have one submitter with a separate system and they have maybe 57, they think you should be in, and let me tell you, there’s no way to submit an empty file, and if you only have 57 reasonably healthy people, you are going to have months with zero claims, and it’s impossible to submit an empty file.

So the way that the states go about this and regulate our participation is sometimes really counter to how we operationally work.  So kind of one plug for considering a different model or considering this consistent model if all states were done the same way, with a large extract or a large extract periodically, then this whole question and this whole threshold issue would be so minimized.

So just talking about benefits, efficiencies, barriers, there are a lot of benefits to standardization.  From the state perspective, the implementation timeline would be shortened tremendously.  Most states believe they are going to have their APCD up and running.  They have been communicating.  You know, some states give us a couple of weeks.  Some states give us a couple of months.  Generally, it is between 6 months and a year.  It just takes that long to create the extractions, go through the testing, QA the files.

The regional initiatives would be much more practical and much more efficient, and so those that are interested here in regional initiatives, obviously the interoperability with the data formats would eliminate all that cleanup.  Any resubmissions, which for a company our size, submitting as many files that we do, we have coding issues that develop for unforeseen reasons, and those resubmissions would also be much more quickly accomplished.

Coding and file QA would be streamlined.  Data submission guide would be shortened.  So barriers.  Obviously the barriers are for the states.  When states talk about particularly some of the recent conversations that we have had with the Gobeille decision and the desire for ERISA self-funded data which is a highly manual process to go group by group and fraught with all of those potential issues that a manual process can bring, we have very openly asked states have you considered a standard versus all of this doing things the way just your particular state is doing them?  Have you thought about the bigger picture?

Generally, the answer — and I will quote one state, and the answer was, well, we are very committed to our standard.  So it’s really a barrier that — I mean, the states are used to what they have.  They are comfortable with what they have.  They like what they have.  So this whole notion of kind of moving them off of that, I think it’s going to really take some work.

Emerging reporting needs.  You know, we are seeing a lot of variability by states, but it seems as though what we have talked about really before this panel talks a lot about the standard reports, the HEDIS, all of these things that are driven by national standards that would be just the perfect place to start versus everyone building their own and having other idiosyncratic ways to do this.

In terms of a roadmap, my roadmap is very detailed, because I do detailed work, but really it’s in my testimony, but it really starts from creating a taskforce run by an independent group, so not necessarily someone advocating for states all the time or researchers all the time or health plans all the time, and so whether it is just whatever independent group, how we pay for that, I don’t know, but I know that I would certainly be willing to talk to people in my area to see what support we could get to help further this type of initiative.

But I think we need to talk about things like thresholds, variance requests, waiver processes.  I mean, we have states that when you don’t hit the threshold on a file, you have to submit your variance through the vendor and then if you can’t remedy that, because your data generally is what it is, then you have to fill out an additional form.  You have to send that to an automated — to a centralized mailbox.  Then they have a certain amount of time, and then they either approve or don’t approve the fact that you only have what you have.

So it gets to be a lot of administrative burden really on both sides, and there’s a lot of churn as it relates to perceived data quality based on thresholds, also exemptions.

After that, of course, ongoing meetings and then I think that everyone needs to be involved, not just carriers, not just NAHDO, not just APCD Council, not just the states, but we have quite a few vendors that are very, very vested in this.  We have consultants that are hired to work for the states that are very versed in these topics.  We have researchers, and we have got national databases.  HCCI was here, but there are other databases and others that are very knowledgeable in this, and it seems as though as an industry, if we could leverage some of that expertise, it might really help to inform the group, as we move forward.

That’s all I had.  So thank you very much.

DR. SUAREZ: Let’s start with the questions.

DR. COHEN: Thanks, it’s great.  It’s been a wonderful day.  Long but wonderful.  Clearly I think the expertise and knowledge and thought leaders for APCD reside in the plans and the states and the health benefit organizations who have been doing the heavy lifting for a while.  So what do you see as the role of the federal government in helping move standards development forward?  We are from the government.  We are here to help.

(Laughter.)

MS. INSKEEP: Our corporate says something a little like that.  From the federal perspective, I think that it would be helpful in terms of standing behind, maybe participating in, lobbying for, agreeing with, a standard.  Right now, the states are making their own decisions.  It reminds me a little bit of international law, right?  So you can tell all the countries what they should be doing, and the countries do what they do.

So but if we could all get together and create an incentive for participation in the standard, whether it is through the grant process — of course, I come from a for-profit business.  So we always think, well, where’s the money going?  So that is one suggestion how the federal government could help with the financing and tying it to the standard, but also as we go through this, if the standard can be adopted and the measures used both by the federal government and the states and we can get more efficient, then I think that that would really help the whole process along.

MS. PORTER: I would add a couple of things.  I agree that we want to — there is a lot of potential support in the development of that standard.  I think we want to be careful about what we say when we — what are we talking about when we talk about standard and do we want to always reengage — at what point do we want to reengage at the DMSO level?  At what point do we need to do some homework before that part is ready?  I think there is a role for clarifying and identifying the process for that.

So I think that there is development of the standard.  I think there is also maintenance of the standard in the longer term.  I think that that’s, as I was saying before, there was an extensive process that happened in 2010, in 2012, and while there are formal mechanisms to do maintenance, there isn’t a formal approach to that, nor is there funding for anybody to be part of that.

So then I think that the other place is that if we are successful in getting to a common data layout, I don’t know — it would be interesting to find out where the federal government would say they would be in being supportive of that common data layout, and what do we call that and how do we name it and how do we sort of, quote unquote, own that over time?

Once that common data layout is defined and owned, quite frankly I think the states may need some help in implementation.  For 14 states, that’s a pretty significant change in what they have been doing to what they would do to that.  So I think there is a real impetus for commonality.  We definitely hear that now.  But I think the implementation approach, the maintenance approach, and the development approach are still — there is still room for figuring out the best way to do all of that.

MS. GOSS: Great questions, great commentary, and Jo, you set up actually a segue to my very specific question I think without realizing it.  My question is around grandfathering those who have blazed the trail, and we have heard from Sheryl and from others that there is this constant churning and variability and those who were learning along the way and evolving as we are going are bearing the biggest cost currently and are having probably even a further challenge on justifying ROI and sustainability, but whenever you are a trailblazer, it’s a twisty long switchback trail.

So wondered your thoughts, and I think you have already said it, which is there’s some kind of grandfathering that maybe or implementation if they are going to pivot, those who have really paved the way should be given some consideration in that case.  Did I hear that right?

MS. PORTER: Yeah, I mean, I’m not sure that there are many — so certainly, if the mechanism to get what could be a data gap around self-funded data requires states to move to this common data layout, then I think that even the old states are willing to do what needs to be done in order to minimize the gaps in their data.

I would say that I think there are states that would probably rather seek support for shifting to the common data layout than support for grandfathering themselves into the old system, and I can’t say that that’s true across the board, nor would I propose to speak for every single state, but I think much of what has been talked about today for the value of standardization around common analytic tools and common reporting and comparative analysis, both out to other states and to regional approaches, states are very interested in that and in fact have been interested in that for a very long time, but again, the impetus is on the state to figure out how to make that happen, and every state is working in their own environment.

So whether or not I think the states would want to be grandfathered to do what they have always been doing or supported to shift to what other states are doing, I think it might actually be the latter.  But in the absence of support, you have to stay grandfathered because you can’t afford not to.

MS. INSKEEP: I think there would be some incentive to move to a standard for states, a, because again there is this interest in the ERISA self-funded and there are many states that are going through a process of continuing to try to require and so those are legal topics that I’m not going to touch on as I’m not an attorney.

However, I can tell you from an operations perspective for a health plan such as ours, we can separate our fully insured from our self-funded, and then in order to separate our non-ERISA from ERISA, that’s relatively manual.

If you ask us to reach out to each and every ERISA self-funded group and try to track all of that and all the moving parts and I’m telling you, we did this recently in Colorado, and we had many groups just change their mind, and so they start something and then they, wait, wait, we didn’t really mean that.  We want to stop.  So when you are talking about the ramifications of what’s essentially a breach in data based on the fact that people are changing their mind, we get a little squirrelly, because we are really trying to protect our members’ privacy, and there are serious ramifications if we don’t.

However, if you multiply that by all of the different ways that we are submitting data, it’s really staggering in terms of complexity.  So I do think that it helps the conversation of standards because this whole ERISA issue is complex in and of itself.  So it does, I think, lend a lot more credence to the fact that the more we can streamline, minimize the variation, minimize the potential for data breaches or anything else that would be adverse, I think that and I hope that everyone would be more open to it at this point.

The whole financial impact for the states, though, I mean, that’s very concerning, because that’s certainly something we broach all the time.  It is hard to submit in so many ways.  So I can’t imagine to change their databases it would be very difficult.

MS. LOVE: I am going to ask a quick question to follow back on ERISA, and it may be a really dumb question, and I am always embarrassed to ask the question that I should know, but how — and I should have asked HCCI and Blue Cross, but also the other private databanks, how are they handling the ERISA issue, because there are private aggregations.  Are they filtering?  How does that work?  Do you know, on the voluntary databases?

MS. INSKEEP: It is done so differently with business associate agreements, data use agreements.  There’s usually —

MS. LOVE: So, but there is a workaround.  How do they filter it?  It’s hard to filter it.  But then they have a process for filtering it that must be efficient.  I am just trying to —

MS. INSKEEP: I don’t know if HCCI is getting self-funded or if it’s completely de-identified before it is sent.  I don’t know.  I can’t speak to that.  I’m sorry.

But I can tell you for things like CHPE or the smaller regional voluntaries, there are fewer self-fundeds involved and there is an enormous amount of legal work done up front where if we had to do that group by group to try to submit that into a state, I can’t imagine how long that would take.

PARTICIPANT: I am just curious, because if there was a process that they were following, that could be —

MS. INSKEEP:  It is a very elongated process.  It’s really elongated and it’s really, really narrow.

PARTICIPANT: If ERISA are not in those private databanks and they are not going to be in state databanks, so we don’t know what’s going on at all.  So it just, from a public policy standpoint, that just alarms to me to no end.  But that is an editorial comment on my part.

DR. SUAREZ: I do have a question about the standard, because it seems like we had the chance to develop the standard.  We developed the standard, but the standard didn’t seem to be responsive or fulfilling the business need of each of the states, and in reality, with respect to the standard of the data that is being is being used or the methods and methodologies and electronic standard being used, I wondered what’s the degree to which there is consistency across — I didn’t get a sense of — so we have 17, let’s say 17, 18, state APCDs at the moment, and there is let’s say three basic databases being collected by each state, and the eligibility file, the provider file, and then the claim file, which includes all the claims.  So claims for medical, for pharmacy, for dental, for everything else.

So let’s just say to simplify, those are the three major groups of files: eligibility, a provider file that identifies the provider types, and then an information on provider, and then the actual claim file itself with all the — what would you say is the state today of the consistency across the states, or the differences across the states, on those files?  Are we talking about when you see one you have seen one really and there’s just no?  Or are we talking about, well, there’s like 80 percent of consistency and 20 percent difference, or 90 and 10?

MS. INSKEEP: I think it’s more like if you have seen one, you have seen one.  So some of them are similar, but they are not exact.  So we have to code from the ground up for every submitter for every state.  So in some states if we have 12 submitters, we have to start from the ground up, analyze those requirements, and kick off that project, and each submitter has to do that separately.

I think the really hard thing is most states believe that, well, if your eligibility file, if this field is this and this field is this, then this field should be this.  So we are actually filling in data, which is not any data that we collect, but it is based off of the interaction of a couple other fields.

So depending on how much of that a state has, that’s a lot of coding, and I don’t know how coding breaks, but it periodically does, and we have proven that, time and time again, and from what I understand from different vendors in different states, we are not the only ones.  So the more of that that we do, the more convoluted it gets.  So there may be commonality like for the provider files, the last time I did an analysis, and it’s been about a year, there were about 10 percent of the same fields.

So we have one state that has a really narrow this is the provider, and then we have some states that want when did the doctor have — where did they go for their residency?  What year was their residency?  Where did they go for their fellowship?  What year?  And all of these things that, you know, if you are submitting a monthly file, you are thinking, well, that’s not going to change, and why don’t you maybe get it from the doctor?

Because at the end of the day, if this is truly an all-payer claims database, which is primarily the data that a health plan has, when you get into all of some of these other extraneous things, they are in transactional systems that we can’t even access.

So when I talk about our claims data warehouses, which is primarily where we get our data, these are data warehouses that we can pull reports off of versus our transactional systems where we do not want to touch those, because that could complete derail and disrupt our business if something occurred.  So it really causes quite a bit of administrative work to try to explain some of these discrepancies and disparities.  Does that make sense?

MS. PORTER: Yeah, just to add a couple of thoughts to that.  I think like anything particularly measure based depends on how you calculate that.  You know, what’s your denominator?  So if you took all of the data elements across every state’s data collection, you would probably have a not very high number, because there are some states that have a lot additional that most states don’t have.

I think where we have tried to flip that coin a little bit is to look at all of the data elements and sort of say to ourselves, okay, if all 14 states are asking for this, this is probably one we want to keep across all of the states, and then maybe there are some that’s only certain states have done, and it’s a very direct question back to that state to say why are you collecting that?  Do you need it?  Are you getting any data?  Is it a 0 percent threshold, or is it 100 percent threshold?  That’s the sort of nuts and bolts of that.

So I think from a quantifying the answer to the question that you pose, Walter, it’s a little hard to figure out what we want the denominator to be.  I think at the end of the day, what we have seen and maybe Bernie and I would see this a little bit differently is most of the states are trying to collect most of the same stuff.  They are not necessarily doing it the same way. 

So there is this process issue that Bernie has talked about, and there are some anomalies where there are pieces that have been added by states that most other states haven’t added, and those states I think now in the processes that we have talked about are starting to say to themselves maybe we don’t have to have that information and maybe we thought we would be able to get it there, and it turns out we actually can’t, because we have had it on our data layout for 2.5 years, and it turns out we never get that filled in.

So we have been in this process that I was talking about really trying to go literally element by element and say, you know, is this a valuable one and taking into consideration some of those anomaly pieces and figuring out if there is value to those.  So again, I think at the end of the day most of the states are collecting most of the same information, and there’s an opportunity around process and there’s an opportunity around if you are going to collect it, can we at least collect it in the same place every time.  I think that’s still an opportunity.

DR. SUAREZ: What made me scared and worried about this is that if we are having differences in the collection of the data, not just the type, not just the methods of collecting it but actual data elements, we are using the systems to evaluate, for example, health reform and how the impact to health reform when you look at state by state or state to state comparisons or across the nation, the picture will not come together very well, I assume, because the numbers are coming from systems that have different data.  I mean, I’m just worried about that.

MS. PORTER: Well, I am going to go back to Bruce’s question, which is again, where is the opportunity at the federal level?  I think there is an opportunity at the federal level, and we could talk about standardization of any number of things, but standardization of interpretation of the measures is probably one of those places where we could do standardization.

So there’s probably an opportunity for standardization at every level, at data collection, at data use, at data interpretation.  There is no shortage of opportunity around that, and I will go back to where we talked to the states in the data use standpoint.  You know, states, they want to be able to have comparators.  They want to be able to have comparators to other states; they want to be able to have comparators to regions and those sorts of things, and again, I think there is opportunity around whether we call it standardization or something else around that activity, as well. 

I get nervous about calling it standardization, because I have been somewhat trained by my standards-related colleagues that there are formal processes for that, and I have to be careful about my language.  So again, I don’t know what we call it, but I wouldn’t want to imply that all of that sits with formalized standards bodies, because I don’t know how long, how big their purview would want to get around those things.

DR. SUAREZ: Just one follow-up on that point that was made earlier about one of the recommendations or one of the ideas would be to bring together groups.  I guess one in my mind somewhat logical place would be the National Governors Association, if they are the group that brings together all the governors that are implementing.  Has the National Governors Association been involved in any of this and are they playing any role in facilitating cross-state harmonization to call it one way instead of standardization?

MS. PORTER: Thank you.  I appreciate your language.  At this moment, no.  Historically, there has been involvement from the National Governors Association.  I can’t remember what year it was that we sat here with them and talked with them about the potential of the role of the National Governors Association, and that technical advisory panel that I mentioned, they were present on that as well.

So I think historically there has been interest at that level.  I’m not sure if they would find it — we met with them a year ago and talked a little bit about the APCD world and where we were, and there was a lot of interest whether or not they have the means to convene.  I just don’t know.

MS. LOVE: Thank you, panel, what’s remaining of the panel.  Survivors, and the survivors of the committee.  I think we just got a flavor today of the challenges that are involved with building any data initiative on a broad scale and state to state in this case.

So the next portion of the conversation will be for the committee.  When do we go to public comment, and then after?

DR. SUAREZ: I think it is a good time here to take public comment.  We should do that.  If there is any public comment or anyone interested in expressing any point of view or comment or ideas.

Just state your name and affiliation.

Agenda Item: Public Comment

MS. WEIKER: Margaret Weiker with NCPDP.  If we go back to standards and what was done and the history behind them and the use of those standards once developed, the SDOs — and I’m speaking for NCPDP, but I will throw in just a little bit ASC X12, because I was the chair of the insurance subcommittee at the time, when Denise and Amy and others came to ASC X12, and I formed the special appointed committee, and Laurie Burckhardt thanks me all the time for doing that.

But it goes back to the standards organizations are basically there to meet a business need and a business requirement.  We bring together the experts, or we hope to bring together the experts in the particular business arena that we are working with, the business requirements, the stakeholders associated with those requirements, and develop the actual standard.

NCPDP’s model is a little bit different because we have staff, paid staff, that does a lot of the heavy lifting in regard to the NCPDP standards development.  Most of my standards development staff have pharmacy background experience, whether it is from a provider point of view, a software vendor, a clearinghouse, or a payer.

So we have more help there to assist with gathering the requirements, grading the layouts, et cetera, where if you look at the ASC X12 model, it is a little bit different in the fact that there’s not paid staff to set and help flesh it out and develop and write the guides and attend this meeting or that meeting kind of thing, because it’s all made up of volunteers.

So we just, yes, and I understand there are some companies, some individuals, some states, that say we can’t afford to participate or we don’t have time.  Well, drop us an email.  Tell us what’s wrong.  We can fix what’s wrong, but if nobody comes and tells us what’s wrong, we can’t fix it.  You know, it goes back to a statement I have made at this committee many, many times.  Help me help you.  So whatever the committee decides in regard to standards development, NCPDP stands ready to assist in that endeavor.  Just tell me where and when.

I cannot speak for ACS X12, though.  The chair of the subcommittee of ACS X12 will have to speak to that.  But we stand ready to assist, and just let me know how and where.

DR. SUAREZ: Thank you so much.

Any other public comments?

Is there anyone still on the phone?

DR. STEAD: Bill Stead here, and I have been trying to listen, but I don’t have anything to add.

Agenda Item: Next Steps for Claim-Based Database and APCDs

DR. SUAREZ: All right, we are going to go to our comment session.  So we can jump in.

MS. HINES: I was going to propose; it’s basically laid out in the agenda that we gather our comments on — we heard many different perspectives on the value.  That would be one set of things that we could lay out.  Key challenges, and then the opportunities and areas for recommendations.  So those are three kinds of themes that we could organize, and I would be happy to put them on the flipchart if that’s helpful.

DR. SUAREZ: I think that’s a good way to start.  I have like with yesterday’s hearing, I think I have and probably others have a list of at least topics to consider.  You know, items and elements to consider, some of which we would need to discuss and determine whether there is any recommendation we can make.  Some of them might not have a recommendation, but at least we want to note them here.

So we can start.  Certainly one way to do it, we can start like we did yesterday, which was go around and begin to have people identify one item.  That might be a good way to get us going, and then we can see how we —

MS. GOSS: Could I propose an alternative method for today, Walter?  Sorry to interrupt.  Linda was gracious enough to send her list.  Okay, good, you have those.  I haven’t had a chance to go back through my notes and pull things out.  So I would be amenable if somebody already has a list to offer that up and then go from there.  We have 20 minutes left.  This is just the beginning.  This is the initial synthesizing after a very long week.

DR. SUAREZ: We are going to go to 4:30, but that’s all right.  We just were talking about that here.  So we will go along and see how it evolves.  We might end up being Denise and me sitting here, but that’s all right.  That would be a good way to reference it.

So I would suggest just go around and say, like, we can start with Linda while people are framing their own thoughts, but we will say, for example —

MS. HINES: Can we do it by categories?

(Comments off mic.)

DR. SUAREZ: I am not sure how to fit this 10 into those three categories yet.  So we have 10 here from Linda, and I have a list of another 10 or so.  I am not sure how to fit them yet into those three categories, but we can at least start to list them just the way they are, and then we can map them against this at a later exercise.

Let me just start with Linda’s and see, people, if you are —

PARTICIPANT: The focus just so I’m clear is for next steps for this committee?

DR. SUAREZ: Well, the focus is what are all the topics that we heard that would be important for us to discuss as we frame a letter of observation and recommendation.  I think that is the goal.  Is that what you had in mind?  So the goal is so what we did yesterday is come out with all the different themes that we heard and areas, topics, that we can consider as we deliberate in developing our letter of findings and recommendations.

MS. HINES: I can imagine this letter organized by saying here’s what we heard the value is; here are what the challenges that all the different perspectives, whether it’s the state or the providers, consumers, employers, what-have-you, and then here are the areas.  So it seems to me that’s probably — it’s one option for organizing what would go to the secretary, and like you are saying, if we don’t want to try to organize it that way right now, that’s fine.  We can just take it all down.

DR. SUAREZ: Exactly.  Those three categories we will look into as we flesh out all these items.  Let me read; I’m going to read Linda’s.  So for the notes, I’m just going to say two or three words and then I’ll expand on those three words.

The first point that she said is claims dash clinical data integration.  She added this is inevitable.  Is this really about APCDs or public use databases?  So claims dash clinical data integration is the main topic.  I don’t think we need to —

MS. GOSS: I think that actually the rest of her portion of it is too limiting.  I think it’s not an or in her second sentence at the end of it.

DR. SUAREZ: So the second point is APCDs and payment reform.  There’s a whole statement here, but you don’t want to make a note of that.  It is just APCDs — I just want to keep it short in the notes there.  We have a longer statement that people are framing, but I just want to keep it short.  So she says how relevant will the APCD be as payment reform goes forward?  You know, those kinds of things.

The third one is the — let’s just call it collision of self-interests among stakeholders.  So here I think we find that like in other topics, we as a national committee, there’s multiple stakeholders.  Each one has a different perspective, and we need to find some of the commonalities.

She points that we didn’t hear too much of the provider perspective.  We had invited and unfortunately we couldn’t confirm some of the providers, but we are certainly interested in their perspective and she also pointed out there’s much resistance to sharing, particularly as health system ACO dash population health efforts continue to advance.  So that is the whole framing.

PARTICIPANT: (Comment off mic.)

DR. SUAREZ: We will make sure to put on the record the other one.  So number four is privacy, APCDs and privacy.  So there’s some point that she makes about, for example, self-insured employers and data about the employees.

So number five, data governance models and practices are lacking.  She is just saying that.  I think we have many different data governance models, and maybe consistency in the models is what’s lacking.

So then number six, I think it’s about scarce dollars.  So spending those scarce dollars in areas like uses and not really reinventing the technology infrastructure.

I would maybe group number seven inside this one which is what’s — she points out maybe there’s a benefit of modeling the cost for setting up a typical state APCD application.  So modeling the cost for setting up a typical APCD application.

So number eight, she has the ERISA issue.

Number nine she has data models and normalization standards, the need for data models and normalization standards.

Those are hers.  So maybe we can start going around.  Bruce?  By the way, we don’t need to go and do like what we did last time, which was — if you have five, you can go ahead and throw five.

DR. COHEN: The last discussion around data harmonization versus standards development, I think, is a key issue.  I am troubled or concerned about the federal commitment and the federal role in developing standards and this sort of echoes one that was mentioned, resources to do all of this.  And timing the process to do this.  Those are my key takeaways.

PARTICIPANT: (Comment off mic.)

DR. COHEN: Well, there could be commitment without resources, and defining the process.  Thank you all.  This was great.

DR. SUAREZ: Thank you.  Thank you for being here, Bruce, for the entire week.

Helga?

DR. RIPPEN: I think the big one from my perspective is how do we define the purpose and the scope?  We talked about payer, claims database versus an everything database.  So we really do need to discuss scope.

MS. GOSS: Scope and kind of name.  Are you tying those together?  Scope and purpose and how we label it.

DR. RIPPEN: Yeah, whatever, yeah.  We can label it after we figure out what it is.  The other thing is the question of how this all ties into all of the other initiatives capturing lots of different data and using data.  That includes the HIEs.  That includes just research.  That includes the quality improvement organizations.  All of these organizations that are grouping and have data, how does that play into this ecosystem?  Again, it depends on the scope and purpose.

I do think there is a nuance as it relates to — I know there was a concern about EHRs and billing systems and what’s within scope or not, and I guess I just wanted to — it’s just more a consideration that EHRs and billing systems are linked and built that way.  So thinking a little bit outside the box.

Then the last part really is about privacy and security as far as who.  Where does this fall under HIPAA, covered entity, which again is governance, but it is actually maybe one step higher.

DR. SUAREZ: Okay.

MS. GOSS: I am going to start or share first Vickie’s.  I am reading from Vickie’s notes.  Critical to think ahead to take into account changing landscape.  Some of these may be duplicative, but I’m just going to give you what she offered.

Suggestion of converging data healthcare and social determinants in an example.  I think it’s converging.  I think the point is getting an example.  Developing a data model for collection that would make its use flexible and has most use across datasets.

I will start again.  So developing a data model for collection that would make its use flexible and has most use across datasets.  Example was how height and weight are captured so that it’s comparable at both the national and international level.  We just took it up a level, international.  But it’s coming.  We’re going to have to do that.

Standards that include — I’m going to shorthand this.  Remember we had the who, what, where, why, when?  So what is collected, how it’s collected, when it’s collected, why it’s collected.

DR. SUAREZ: I think it was what, how, when, why, and who.

MS. GOSS: Yes, I agree with you.  I expanded on Vickie’s comment.

Okay, her number 5 was actually going to be one of my ones to comment.  So I will give you hers, and then I will take a slightly different twist on it.  She would like:  need to comment on the centralized versus distributed and come up with the pros and cons, is what Vickie suggests.  So the pros and cons of federated versus centralized or distributed, as it was being called today.

From my perspective, I felt that that conversation was something that was a bit of a distraction today, and that theme depending on stakeholder type was pretty prevalent.  I think the concept of centralized versus federated is something that has been — the tires have been kicked pretty hard on over the last five years, and it is sort of the in-the-weeds technology option, et cetera.  There are a lot of things — there are differences between the two, and I think we probably need to accommodate both, but to me it is just a basic assessment issue in the beginning, and it’s influenced by a whole bunch more than just technology.

PARTICIPANT: (Comment off mic.)

MS. GOSS: Not in our country.  There’s not one size fits all and we need to be very respectful of that.

So I think we may have had something about the interoperability roadmap needs to bring HIPAA back in scope.  I don’t want to just think interoperability roadmap from a national — from a clinical perspective.  I want to think the whole kit and caboodle.  We have to get our arms around all the datasets and how it’s all fitting together, because they are all linked and it all starts with patient care.

How people use the word standard and what it means to go beyond a base framework, like a standards development organization like NCPDP and X12 offer.  NCPDP offered to help, and X12 is giving guides as well, as we heard from Margaret.  There is more to that than just the core SDO, because the way we do standards development across our business models is handled in multiple ways, and I heard a lot of different kinds of standards that need to be addressed.

I heard a theme of guaranteed access to data by employers and their leadership role at that table, of what’s happening with the data.

PARTICIPANT: (Comment off mic.)

MS. GOSS: They want their people healthy and happy and good, yes.  So they have a lot of good motivations.

I also heard a theme of neutral intermediary or convener to help — this kind of goes back to one of Linda’s comments.  Somebody has to come bring everybody around the table, and manage that process or else we are not going to make any progress.  Process of building APCDs or whatever their name is.

DR. RIPPEN: So, it goes back to what is it as far as the program, again going what’s the scope and purpose, and then based on that, leveraging or thinking about existing but then also opportunities for a new way of thinking about things, too, right?

MS. GOSS: Not necessarily putting, to add a bunch of new stuff to it, either.  One of the themes I heard was about the distinction — the complexities of identifying the parties in healthcare.  So first of all, just knowing who is who in the zoo and how they all relate to each other.  Provider ID and attribution, thank you, Denise.

But I also heard that that then, there’s the added dynamic when you de-identify the data before it goes into a database versus the complexities of what happened when you come out of the database, and that was a very interesting nuance, and I think Patricia made that comment.

I think that’s good enough.

MS. LOVE: Well, I am a grouper, not a fish.  Though I feel like one right now, out of water.  So, I just really have three domains, and I can give you the piece of paper so Rebecca you don’t have to write them all, but I really mapped it into federal state overlap.  I mean, there is clear areas where the feds are needed.  Whether it’s 42 CFR OPM or Gobeille, Medicaid, or ONC.  It’s just very clear.

Then the other grouper thing I had was standards, and we get into not only the core, the data elements, but all the other things I heard people say, linkage, identifiers, and future standards, as we have existing, but what do we build for future, and I think that is something that needs to be addressed, whether it’s alternative payment, whatever.

Then the third group that I mapped what I heard was national state roles.  I mean, clearly the states have the business need and are building and paying for these things, but what is the national role?  Is there?  I think there is.  Funding, who pays?  Is it shared?  Medicare/Medicaid, how is this interplay with what the states are doing.

So those are the three domains with some of the subdomains, and I have it written and I’ll hand it in.

DR. SUAREZ: I have not said mine.  The ones that I read were Linda’s.  I read Linda’s.  We’ll go to Bob first and then come back.

DR. PHILLIPS: I think there’s been a lot of ground covered.  I heard an opportunity for Medicaid, which is saying it will do 90 percent funding to help normalize.  What are the best in class doing?  What can Medicaid really help fund towards?

I heard an opportunity to do some normalizing.  We really had some outstanding states, some of whom I didn’t know what they were doing, but this idea of what are they doing that informs governments?  What are they doing that informs practices, and specifically around supporting the move towards MACRA?  And what are they doing to support the payers and even under the payers, the ERISA, the self-paying employers, how can they really build on a return on investment in terms of what the information provides so that everyone really wants to continue to participate?

I got really concerned about the idea that once — that the payers are providing data to multiple groups, and they kind of lose control of it once they do.  When you start talking about integrating clinical data with claims data, that can’t happen.  I mean, really the stakes get much higher in terms of privacy.  So that actually touches I think back on our minimum normal and privacy issues.

I heard that there’s a real opportunity to pull social determinant data in, and while I don’t want this to squelch innovation, it would be awfully helpful to this to help normalize that process, to at least create a standard across all states about what gets pulled in and how it gets used, and I think that actually comes back to our roadmap about social determinant data and what does the federal government can do to support the availability of data, because when ASPE comes out in 2017 and says this is how we are going to pay for social determinant adjustments, not many states are going to be prepared for that.

That’s my list.  So this is the role of our population health committee and the roadmap of social determinant data, what data are available at what level?  I think this ties into it.  What this actually I think supports our continued effort here, and I think CDC’s role specifically in supporting data availability so that it can tie back into what social determinant data get combined with claims data.  Trying to relate it back to the other work we are doing.

DR. SUAREZ: All right, I will go with mine.  I am going to start with the big V as in value.  I think the point that we have to make and review, it’s just like when we recommend adoption of a standard and we evaluate the value of the standard how it is meeting the business need and how it is fulfilling the purpose for which it was being created, I think we have an opportunity to highlight and to identify and to document some of these value points.

Value not in the sense of having a lot of data to play with, as some people have said, but actually data to point to issues that can be fixed and then really does it get fixed?  In some cases, the role of the entity that is conducting or maintaining or supporting in a state these type of databases might not be the responsibility — the responsibility might not be that entity, that same entity as the one that carries over.  But value is a big topic, I think.

The second one is standards.  A lot of people have said a lot of things about standards.  I think standardization of data elements, of quality of the data, thresholds.  We heard a lot about thresholds of reporting, data analysis, standardization of data analysis, standardization of the data dissemination itself and the release of data.

In the federated versus central, I think it was a valuable discussion, but more the way I saw it, more from the perspective of the kind of example that Bruce gave.  So I think federated versus central was more meaning like what we do with vital statistics systems in this country.  We don’t have — well, anyway that’s an example.  So federated versus central continues to be an important one.

Really another big one is state level versus multistate level aggregation.  So at the end we have this wonderful, incredible state level databases, but we are going to try to look at the nation as a whole, I don’t think we have the ability to do it yet, and that’s part of the value that we would need to think about, because it’s great that we have these state resources that every state has that can use.  I’m talking about the APCD.  I am just saying just like conceptually we have these data resources at the state level that are used to analyze at the state level, looking at some sort of aggregation that is sort of like the hospital, the HQAR(?) project.

DR. RIPPEN: I think the nuance is I don’t know if we have to figure out how to aggregate, because it could be aggregated database, but it could be using tools to analyze and aggregate the analysis.  So there is a nuance.

DR. SUAREZ: My point is really we have a great resource at the state level.  Should we have it at the national level?  That’s the main point.  We have an opportunity to highlight that in a letter.

Roadmap into the future.  I think the roadmap, the way I was thinking about it, was more about into the future, how does the current system that we are using today fit into payment reform and delivery, care delivery reform?  So with ACOs and with bundled payments and pay for performance and those kind of things, do we have the right data ultimately to achieve that, or are we thinking fee for service mentality to try to solve a pay for performance issue, on population health?

Governance models, I think it’s another topic, but I think it has already been highlighted.  Data governance I think is a big, big important element, and harmonization of data governance is a critical one.  It’s sort of like having — when you have seen one, you’ve seen one in terms of data governance, and it’s ultimately the same data across the board.

The limits in the level of comprehensiveness of the data I think is still a challenge.  So we should discuss and have an opportunity to comment on how to fulfil those gaps that exist in the level of the comprehensiveness of the data.  I’m talking about, you know, Medicare data; I’m talking about self-insured data.  I am talking about non-claims type data that has been mentioned before, too.

Big data privacy considerations include, to me, I think as really the patient-identifiable data and unique identification of individuals in the data and how it can be really expanded, if you will, or what is the implications of having the patient identifiable data and the unique identifiers or not having them, if that’s the case, because in some cases they don’t get collected, in order to achieve some level of understanding across systems that have different identifiers for individuals.

What is the level of individual identifier information being collected, and how individuals are really identified and matched across systems, because I don’t think there is a single solution there yet.  There are multiple approaches.

Sustainability is another one, and we didn’t touch on this, although a couple of people mentioned it.  Sustainability and the selling of data in order to achieve certain level of sustainability.  I think that is an important point to discuss.

DR. RIPPEN: I would say that — and maybe the nuance, because selling is one thing, because if you ask for CMS data, it could be 24,000 or 48,000, and the position is that it’s the cost of actually providing the data.  Because it could be selling to make money or it could be —

DR. SUAREZ: It is a charge that includes the cost or includes a cost plus or whatever.

MS. LOVE: In our experience, pricing is also a way to control the casual purchases, but that is a different argument.

DR. SUAREZ: The last one is data aggregators.  We didn’t talk too much about them, either, and I think it would be valuable to bring that discussion.  Third party data aggregators.  So that was my list.

DR. RIPPEN: Just one, and it was touched on to some degree, but it’s the question of the role of incentives.  It might be federal payment, but I know it was mentioned a few times, and I say that because if you think about some of the things that are happening, for example, PCORnet, where you have 14, 15 different systems and regions across a state, country actually, competing to standardize and normalize data and with probably I would argue not sufficient funding to do so, and again, getting the alignment quickly and them feeling that they have to, because otherwise they won’t be able to participate in something.

So again, the question of what role do incentives play in actually accelerating or shaping or directing value.

MS. LOVE: So I think we are —

DR. SUAREZ: Bill, do you have any additions, Bill, if you are still on the phone?

DR. STEAD: I am still here.  No, I think you have plenty, and as I checked what I had heard, it seemed all covered.  I do think figuring out scope is an early parsing opportunity.

MS. HINES: When you say scope, Bill, could you just say a little bit more, because I’m a little braindead.

DR. STEAD: How far is this moving from claims?  I think that in many ways that will influence all the other things you have listed.  I am just saying I think it is an organizing principle.

MS. LOVE: So, scope.  Aspirational scope or practical scope?

DR. SUAREZ: Defining our scope as we discuss it.

DR. RIPPEN: Actually maybe again, because life is full of nuances, because we already know scope has changed, depending on who you ask, from the presenters.  So there might be scope as it relates to roadmap.  So the scope in the short term, you know, what are the big issues today to set the stage with understanding where do people want to go so that you are aligning not only kind of your prioritization but addressing the scope as it is today as you then also prepare for kind of where it needs to go, because that way you don’t have to rip and replace potentially.

DR. STEAD: That makes great sense, because if you can work out how to deal with standards and architecture for a relatively small scope, you can then build it out.

DR. SUAREZ: Great, okay.  So I think the next steps are basically going to be we are going to wrap this around a document that will summarize and de-dupe these comments and go back to our purpose of the hearing and the questions that we asked also, and then structure this, all these ideas into some groups as Rebecca identified.

We should type all this up.  And then what we will do is I would suggest to convene a group.  We will ask for three or four volunteers, because this is a committee-wide activity, but we want to have three or four or maybe five individuals that can get together and develop a document that can be then presented to the rest of the committee for discussion and for comments.  But I think that’s what we will do, and we will do it via phone calls and emails over the summer and hopefully we will have back something in September to the full committee.

All right, I think we are going to get our meeting adjourned.  With no other people in the room, we are adjourning.  Thank you, and thanks, everyone.  Thank you, really, staff, for the incredible support.

(Whereupon, at 4:16 p.m., the meeting was adjourned.)