[This Transcript is Unedited]
DEPARTMENT OF HEALTH AND HUMAN SERVICES
NATIONAL COMMITTEE ON VITAL AND HEALTH STATISTICS
SUBCOMMITTEE ON POPULATIONS
June 20, 2007
National Institutes of Health
Natcher Center
Bethesda, Maryland
Proceedings by:
CASET Associates, Ltd.
10201 Lee Highway, Suite 180
Fairfax, Virginia 22030
703-352-0091
P R O C E E D I N G S (3:55 P.M.)
(Introductions)
DR. STEINWACHS: Why don’t we go through and finalize an agenda of what we’re going to do. The agenda that is in your book in front of you has “Finishing a Draft Letter” so I thought we ought to try and resolve a few points on that. Then we need to talk about next steps on the two current activities – Linkages and SURGE – I put at the end this 21st century because I wanted Larry and Garland to have some chance to talk a little bit about the ideas that you were raising with me about Population Subcommittee being more involved in looking specifically at issues with public health data, and so I wanted at least to start that discussion because we need to be planning for what we are going to do over the next six months to a year. So we don’t have to make final decisions now, but it would be good if we want to have hearings or workshops late summer into September to try and make decisions about those things that are beyond that a lot time in September again to be talking about it or we can have conference calls to talk about these things.
Anything else that ought to be on the agenda?
Let me just run over the changes that I heard for the letter and what I thought I would do is fine, type those up this evening after this group makes a decision and then we can distribute them tomorrow morning, first thing so you can look at them before the Committee looks at them again.
So, one was to put in a “Re:” line. We had – that generated more discussion than anything else I could think about the letter. I had copied down a number of different things, the last one I think might have been, was Leslie was increasing use of link data sets while protecting confidentiality and security. I didn’t get Larry’s – Larry you had one –
DR. GREEN: I like Leslie’s.
DR. STEINWACHS: Okay, so why don’t we go with Leslie’s, unless there are other comments? Put double stars by it.
The other was on the NCHS recommendation, the first recommendation stands as is, was to put a couple sentences before the first part of what would be a split recommendation. So we were going to take what “the NCVHS also recommends” which had two sentences, was to split those into two separate recommendations. The first one would be amended – or maybe we don’t have to put an actual couple sentences. It was other agencies within the department that sponsor research data centers including AHRQ, CMS, and SAMHSA review their access policies and procedures to ensure these data can be utilized effectively and the suggestion was that it be not only effectively – accessible and effectively? No, that doesn’t sound right. Let me use effectively – leave it at that. Does that need a couple of sentence lead ins or does it? Does it need a couple sentences that say that there are others besides NCHS there are —
DR. SCANLON: I think maybe, that since there is no setup maybe just a sentence, but it would start off “other agencies also make collective valuable data and maintain and operate data centers” – something like that; and then accordingly NCVHS also recommends that these agencies do this because –
DR. VIGILANTE: You might want to link it back similarly to NCHS other agencies operate data centers which should conduct the same –
SPEAKER: They wanted us to name them such as AHRQ, CMS, and SAMHSA I think was the other one.
SPEAKER: So the names, should that be
in the introduction and not put it in the recommendation itself?
DR. STEINWACHS: We just adopted your suggestion for the “RE” line, so you’ve done well – you want to just introduce yourself?
DR. FRANCIS: I’m Leslie Francis and I thought I’d start joining this group as well as – you mentioned privacy?
DR. STEINWACHS: Yes, we really need some privacy and confidentiality in this group because –
DR. GREEN: I really like this letter and
I like the way this is going. I wonder about the following, when we start using these agencies these are the usual suspects. What I have begun to realize is the real power of this thing is when do you move out into the census and move into the Department of Education is when you start moving into corrections and you move beyond the family that’s the opportunity that’s not seized, and we need to shake that occasionally.
DR. VIGILANTE: Really interdisciplinary type of research that comes from different research paradigms and perspectives.
DR. GREEN: Over lunch I was talking with guys that are here reviewing journals for the National Library of Medicine and they were talking about geography.
Geo-coding and the publication of relationship revealed that geo-coding had been under our nose forever and how interesting those were and how the new journals – it is that sort of idea that I had in my mind when you had a conversation in there about a couple sentences that set up what we’re trying to get to and then someone talking about the “RE” line used the world ‘value’, something about enhancing of the value of the taxpayers’ investment in these various agencies that we’re really after and I still don’t think the letter quite –
DR. VIGILANTE: This whole conversation started a year ago and one of the things we talked about was the highest level of educational attainment is probably a better predictor or health outcomes than almost any other metric yet its not a health metric and its really by consolidating or linking these data and hence these fields, the endeavors of research that produce an added value.
DR. SCANLON: I wouldn’t disagree about the idea of agencies outside of HHS but I think we also have to be mindful that we’re writing to Secretary of HHS and we’re the advisory board for the Secretary of HHS. As in the last part here where we talk about trying to standardize the linkage agreement sort of among HHS agencies and we make a mention to other government agencies. There’s an issue of how one deals with the non usual suspects because they’re not –
DR. GREEN: We should do it the way you recommend.
DR. VIGILANTE: Like you did in the FERPA letter that we didn’t know there was an outreach between this Secretary and other Secretaries.
MR. SCANLON: We usually just say HHS is easier to work with.
DR. VIGILANTE: Right. I think that makes sense. It’s got to be in that kind of content
DR. SCANLON: There’s another issue with respect to the splitting of this recommendation which is I’m not sure that I came away fully understanding Simon’s point – let me put out one interpretation – you wanted sentences about the new technologies that we’re talking about taking full advantage of – what are these new technologies that are going to make data more – protect confidentialities for the better. I guess in thinking about this after we went through this discussion, Don, what you were raising about DOD may really have been more about security than confidentiality. The issue that I can remotely access classified data, but I’ve got a security clearance and so it’s fine for me to look at the classified data; this is different. This isn’t about stealing the data or anything like that, this is the issue of the output that I get protects confidentiality. There was some discussion about that in which went in some respects went both ways during that workshop. On the negative side was that the algorithms for in some respects mining data to figure out who a person is are becoming so good that it’s becoming more and more problematic to have a lot of variables – sort of data file because that’s going to ultimately lead to identification. We talked about at one point that maybe public use files will only start to identify at best census region or maybe if even just that they’re in the nation, or maybe we won’t have public use files at all.
The workshop involved a lot of discussion about we may be moving in the wrong direction here because we’re really losing the ability to protect privacy. And then there was some discussion, but it wasn’t necessarily that strong, about well can we find sort of ways of basically encryption or screening in terms of releasing results that will protect privacy while allowing sort of essentials of analyses to be done. I’m a big believer, because of my background in economics, that the markets that people are operating in matters so I want to know what market are they in and what are the characteristics in this market but that’s geo-coding down to a very fine level of detail, and that’s getting precluded more and more. So, I don’t know where to go with the sentences that he wanted to precede the second half of the recommendation because we’re way down what the new technologies are.
DR. STEINWACHS: I am too. I have another strategy – Kevin, you had suggested that we layout a little more detail that currently the process that secures the privacy of the information leaving the data center involves a person looking at the output and making a judgment and I assume the guidelines – I don’t know but judgement are made by people who are doing that — that the privacy is being protected, because the cell sizes aren’t so small that you can begin to identify individuals or the multivariate tabulation doesn’t allow the variables.
So maybe we could set the stage here by talking about that and the importance of, I guess when you talk about technologies there would be two types – and I guess maybe we can just drop it at this point – one technology has to deal with what your competent statistically but indeed is still not identifiable and HIPAA said that you could have identified data either by virtue by eliminating all these items from the data set or you could make a statistical statement – you could make a statistical argument that the probability of identifying someone is less than some number. The other is that instead of having a person review an output could you have a computer review an output and using an algorithm make a decision that this meets reasonable criteria for not being identifiable when you have this cross tab, or three way tab or four way tab.
DR. VIGILANTE: So maybe the way to do it is
to kind of discard the “as is” state and say currently the data center serves two functions; one, to provide security but then also to enhance confidentiality by having reviews, by having individuals or trained experts review the outputs to provide a level of assurance that identifying individual’s is unlikely. Then there are two ways then to proceed. One is to continue to require people to go to the data center or then provide them with remote access to the data center and continue to have that review, or perhaps even use technologies to provide that review to assure confidentiality. I think that might make the whole caustic or transparent to reader then understand the distinction between security and confidentiality and also understand what you’re proposing and I think it could do fairly economically.
DR. GREEN: The security issue, for another part of my life, the certifying boards or positions are now distributing examinations to computing centers all over the country and they are secure exams, and I don’t remember all of the technology but I am confident that the technology is sitting there available for application that it will allow you to put out something like 300 questions with possible answers and take it back and it was never there. That’s what it seems to me, that’s what went through my head when Simon was talking about this. We need to get smarter about how to let this thing out of Hyattsville so some guy in Albuquerque can use it without having to rent a hotel room and fly out here and use it and then suddenly it is back in Hyattsville.
DR. SCANLON: I think it is the algorithms that Don is talking about in reviewing the output that is what we have to be able to develop because —
DR. GREEN: But even after that is done the distribution of the data —
DR. SCANLON: See I think actually in some respects, that’s not the real issue.
SPEAKER: That is a security issue.
DR. SCANLON: It’s not necessarily a big security issue because the real benefit of having people come to the data center is you don’t have to have too many people review their output because few people are going to come to the data center. But if you’ve got everybody around the country keying in, and you have got this queue of requests and you have got to start assigning people to review these things that’s – so you’ve got to have something to control the demand and one way to control the demand is to say come to Hyattsville.
MR. SCANLON: You could submit a remote
data request but then it goes into a relative manual system where its run against the data set and then a statistician reviews it to assure that there are no confidentiality breeches; and then its released to you and you basically have the statistical output, you don’t have any of the records.
DR. SCANLON: And that’s where I think we’re at now, with respect to remote access and what could improve this for everybody would be the algorithms that Don is talking about which is to say the statistician doesn’t need to review this manually, its going to be done by machine, and lets be honest, if you are looking at pages and pages of output to see whether or not the cell sizes were sufficient, do you think you would be paying attention all the time? So that the algorithm is probably going to do it better and provide better protection for confidentiality, and I think that’s the direction you want to go. I don’t think the security issue, I mean – that has been broken by DOD and some others, that’s been solved. It’s this issue of do we have the algorithms that can protect privacy —
DR. GREEN: Are those actually being used in HHS?
DR. SCANLON: The algorithms or the security measures?
DR. GREEN: The security measures for access. Is HHS –
DR. VIGILANTE: There is not remote access now, right?
MR. SCANLON: Well, you have to literally – it is not remote access. You can remotely send in your data request but then it goes to into the regular system. And the security measures include, it’s not on the network, these are — the computers through which these remote requests are others are sent, they’re not linked to the main database, things like that. There are certain – you could look at cell sizes you know that’s a simple thing to do, anything below 10 or anything below three, that’s not the main issues, its more up to now there’s been sort of a need for it depending on what the variable is and what variables you’re looking at. I think NCHS has an automatic suppression for less than 30 cases, that’s their statistical viability, that’s pretty easy to do, but I think it’s more than that. Sometimes you might combine a few codes – the higher income codes for example, and you might be able to get the person a decent tabulation.
The other thing is some of these are tables and others are the result of statistical type things like a regression or principle components or whatever, so you’re walking away with a statistical output.
I think at the workshop there were a couple of issues, one was, even if you go to the site it could take awhile, so it’s hard to do and it’s cumbersome. For a graduate students and others who don’t have a lot of resources it’s complicated – you know you have to pay for travel, pay for a hotel. So that was one issues in terms of access. The other was you have to go one of these sites and the remote submission is still, while they were working on it, its not like it was still fairly early stages.
DR. GREEN: Our recommendation is to allow secure remote use not remote submission of a request form.
DR. STEINWACHS: Maybe we just say remote ability, and drop secure, because I think Larry, your discussion had been useful – we really are not talking about security of moving the data — we aren’t moving data out of the data center we’re moving end results of analysis out of the data center so that product that comes out the data center should be something that anyone can look and no one could-
DR. VIGILANTE: I guess what I’m saying at the end of the day one of the obstacles is you still have to travel from Albuquerque to Hyattsville. Will what we’re proposing address that?
DR. SCANLON: Yes
DR. VIGILANTE: So you will not have to travel here?
DR. SCANLON: It seems to me the preface to this recommendation the preface to it actually should be, in some respects, about travel. For most of the better or more use of these data that the greater the ability access them remotely needs to be developed and maybe we shouldn’t be talking about new technologies, maybe we should just be talking about remote access because there is nothing very sophisticated about this. What’s coming in from you in terms of your data request is not confidential, what’s going out, in terms of your output, is not confidential. The data are never leaving the data center so, we don’t have to think about things that are too sophisticated.
DR. VIGILANTE: You know when we get research on data sets, I mean, usually its by getting your hands dirty and digging in yourself and combining things in different ways and slicing and dicing, by requesting something from somebody else and having them send something back to you and you’re looking at saying “well, what if we did it this way” that is isn’t happening, right? I mean how does that guy or that woman, the person in Albuquerque sort of accesses data sets from these two different agencies and then kind of really, work them the way that you would if you went to the data center, which I believe you could.
DR. SCANLON: I think that is like the old days of batch processing. You submit your data, it comes back you look at it you say “Oh my God” and you go –
DR. STEINWACHS: It’s good that you still remember that.
DR. SCANLON: It’s not the way you would want to do it but this is what the cost of protecting privacy is.
DR. VIGILANTE: And that’s the only way to do it to protect the privacy.
MR. SCANLON: Now there’s a whole set of other access products that are available. You can go to the websites down from our agencies, you can go to the HHS portal for example, and you can get access to whatever they published. You could get access to – they’ve already done, what are the most 100 popular tables from this data set, that’s already there as well, and that’s already been looked at for confidentiality. You could even make up, you could even sort of put together a cross tab, so if you’re wrong usually it’s just taking one of the 100 most popular tables.
This was the next step where you really had to tailor — what most researchers want is would be able to get the actual 50,000 households record, micro record data, plus the Social Security files plus the Medicare claims files all linked together. And the problem is you walk away with that and you pretty much know who they are, someone could figure out who they are. So that’s what the researcher wants, the researcher wants to look at this, get a feel for it, look at the distributions.
So the only way you can do that now on these large link files where you can get tabulations, you’re not going to walk away with an individual record even if it’s the de-identified, you have to sort of experiment through, ask for these tables, ask for these cross tabs or maybe even do some. When you do some of these techniques you get a lot of correlations anyway. So this is really, all of the other data access mechanisms are there but when you get to data linkage and you now have administrative claims files and other things linked, those files exist publicly as well, so now you’ve got a situation where you are getting more and more detail, in some cases someone could get that data for other reasons administratively, so they can actually trace back who that person is if they’re so inclined.
Its no so much that people do that, is that it can be done, so that’s I think the threat. I think you’re right. It wasn’t so that you had to go to Hyattsville, that‘s the way you had to go originally, now you can go to one, and I think they are associated with the 10 census sights as well. So now you can go to San Francisco.
I think you’re right. You could argue that this
should be Internet – that this should look like the Internet in many ways – you ought to be able to go to a site regardless of where you were, regardless of where the home site is, there ought to be technology. If they want to have human eyes looking it over before they hit the send button queue, that’s fine. Maybe that’s the technology that they’re thinking of, that there ought to be a way to submit requests. You are still not going to walk away with the source files but you’re going to get the tables and you’re going to get the – whatever —
DR. SCANLON: They probably are talking about some security procedures too because even though the data that are going to go out are not going to be confidential, there’s still a process in terms of saying this is a legitimate use of these data and you’re, in some respects a legitimate researcher and they are not going to let just anybody access the data center. So they probably do want the output even though its not confidential and privacy’s not at risk to only be going through a legitimate requester. So there is probably a security element —
MR. LAND: Even with remote access, is a human being looking at reports released?
DR. SCANLON: At this point —
DR. VIGILANTE: So what’s needed is to really work on ways that we wouldn’t have to have a human being look at it at all.
MR. LAND: There’s really no full proof system in that regard.
DR. STEINWACHS: Even human beings are not full proof.
MR. SCANLON: It depends on the variables.
MR. LAND: I will give you an example. When I worked in Missouri we had an interactive system that we thought was pretty good in terms of — You could create one table and it would blank out cells, you could create another table that blanks out cells, but you could put the two together and you could not figure them out.
DR. STEINWACHS: Garland you’re just too damn smart.
DR. VIGILANTE: But there are ways, you know, rather than using date of birth, using week of birth and month of birth. I mean there are other ways to obscure detail that actually is very effective in frustrating your ability to actually identify people. There is software that does that as well.
MR. SCANLON: And if you submitted a request to NCHS or whatever and they could – that’s why its hard to think for eliminating the human element because they might decide – they could meet your request as long as all you needed was your at list birthday. As long as you did not need date birth, then they could tabulate that for you, or income for example, is an obvious one, or a level of geography. So the human looking at it could say we can’t give you that exactly but we could give you this which is close and it may not matter to you. But again, I don’t think that matters. There ought to be a way, you shouldn’t have to go – originally you had to go to Hyattsville now you can go to 10 locations, the next step should be from any website you ought to be able to submit this request. Now, what happens at NCHS; that really you don’t need to know particularly.
DR. SCANLON: Well I think it’s a concern of ours because of a combination of some things, one is that we could make automate this that would speed a sort of better response; we wouldn’t have to wait for a backlog of requests. Secondly, the second related to back order requests is in the whole NCHS budget. Then in sort of another context you’ve heard about, the problems they’re facing is the issue of allocating resources to this report.
MR. SCANLON: Well, that’s resources, that’s a different issue you could handle all of these much faster if, I mean, there’s a queue now because there are probably more persons doing this. That is a different recommendation.
SPEAKER: Move the ball back.
DR. STEINWACHS: Let me see, its two steps forward, two steps back process. Well why don’t we do this, I thought while the Quality meeting goes on, I would sit in the corner and see if I could come up with something, and then catch those who were around this evening to look at it, and Kevin, catch you and others who are here this morning and see if we can get there because I think I’ve got sort of the gist of it even though I’m sure its still in the implementation –
MR. LAND: By the way, I don’t know if you want to leave the word “technologies” in or out but a lot of states have developed interactive systems, data retrieval systems – CDC has their Wonder system, National Center is now using the 20/20 system. I don’t know if these types of systems have been looked at in terms of this need or not where you basically do some limiting in the system itself. So the system will only retrieve age by five year age groups or 10 year age group; you can’t get down to a particular age.
DR. VIGILANTE: It lowers it sufficiently.
MR. LAND: I don’t know if that has been looked at. I mean that’s a technology issue, but it does give you the freedom to create tables on the fly within the constraints of the system. I don’t know if that was mentioned in the hearing or not, but I think its something that could be explored. Maybe that’s what we mean by technology, I don’t know.
DR. STEINWACHS: Lets see how far we can go because we have to stay consistent with what was said in the hearing, at the same time we can talk about areas we’ll exploring. Larry.
DR. GREEN: I’d like to just chase that Jim said, Larry talked to him about resources and forgot the recommendation about add staff, can someone just indulge me for a minute, what happened at the hearing about getting our arms around the value proposition. You spend $30 million to produce another MEPS and you hire one FTE to provide people access to the data. Was that sort of thing talked about? Is there a database in which to make a response about? Unencumbered by any information here, what I suspect is that we are now a little long on the data collection process and the organizing data sets and we’re short on enabling the field to use these data sets after we will spent our tax dollars on getting these things, then we don’t have really ourselves lined up so we don’t get to the value proposition, the amount of work over the cost and everything, we don’t have it quite right – that could be tweaked. Was that in the hearing or not so that this can be strengthened any?
DR. STEINWACHS: We tried to get at value because it was part of the hearing and its also part of the motivation for why we pursued it. In the introduction we tried to talk about big investments in data versus – and I think we need to keep coming back to the value, so –
DR. SCANLON: It really is a sort of ROI issue and we don’t have the answer. We did talk about ROI in the last paragraph. We talk about making the initial investment then not spending enough in that sort of good use –
DR. STEINWACHS: But we don’t have enough data that sort of says how the trade off’s made, we don’t know the different data centers, how they balance that and indeed, I think what Bill was saying is they have ways of discouraging access so in a sense one persons’ good enough.
DR. VIGILANTE: Although, maybe its too late in the game but to take an example off of a couple of very large surveys and actually cite how much money that survey cost and really talk about many millions of dollars and yet they’re not being maximally leveraged for lack of the fact that we cant integrate them or link them for what would be a fairly modest investment compared to the original cost of these surveys.
MR. LAND: This first recommendation is kind of –if the National Center doesn’t even have enough money to continue surveys in the field and they cutting back on sample size, they are cutting back on all sorts of things and here we’re saying we ought to go out and hire more staff to get access better. Well, if they had the ability to hire more staff I’m sure they would do it. I’m not sure this is very helpful to them.
DR. STEINWACHS: Well, it may not be. Remember this is a letter to the Secretary and the Secretary has influence over the recent budget recommendations that go out of the department.
MR. LAND If we were going to tell the President to get more money for National Center, I’d probably be talking about some other things too.
MR. SCANLON: I know what will happen; it will come from somewhere else at CDC. Or if you say that vital statistics should be supported, then that’s where they will put the money – and it will come from somewhere else. But again this is the result of a process and it’s a workshop, you could, as they say in the regs world, you have to sort of stick to what was discussed there or a logical sort of outgrowth. The resources, I think, came up.
DR. SCANLON: I would say we really should be talking about, and this was going to be in our next steps discussion, other things about what’s happening with NCHS, in terms of the surveys, in terms about vital statistics, that we should be thinking about addressing because they really do have implications for populations and statistics. The second thing about this letter is that I think its important to maintain sort of a balanced perspective here because yes, there’s a big investment in the data but there’s also a lot of use in these data and we don’t want to go overboard in saying that they’re not being used because they are being used a lot and you can just keep tripping over the results, so the issue is we can do a better job and its not like either NCHS or AHCPR have not devoted resources once they’ve done one of these surveys to try and have –
MR. SCANLON: You don’t want to undermine the value —
DR. GREEN: Did you say AHCPR?
DR. SCANLON: Oh, I live in the past.
DR. STEINWACHS: Why don’t you say an NCHS R&D?
DR. SCANLON: I’ve been there, too.
DR. STEINWACHS: It’s good to have a long term perspective and fresh young perspectives.
MR. SCANLON: I do think that the, just to show – just bringing some attention to the data linkage issue a lot of the other issues they just weren’t doing because a lot of the other issues apply whether they are – HHS and others – but it is very hard to accomplish the data linkage even when everybody agrees such should happen, and that doesn’t happen that often. It just, you have a lot of procedures to go through – its quite expensive – and then when you have it your limited in terms of who can have access to it for various reasons. So bringing attention to the issue, even symbolically, is worthwhile, and then again I think there were some issues with the NCHS data center about whether it is link data or HIS data, its kind of hard for folks who don’t have a lot of money to do this, so easier access there would help.
I think the technology is getting there to where you can probably do this and the resources, I think, is helpful. That’s probably the most limiting factor of all.
DR. STEINWACHS: Well let me just take a stab at it with this advice and get it back to all of us.
Let me take you on then, since we’ve been talking about data linkages, why don’t we talk for a couple minutes about what the next steps are. We did have a telephone call, and small participation unfortunately, one of the issue, and Joan Turek who works for Jim, was on the call and couldn’t be here today, she raised the question about do you really need an approval process to get access to the data, which is sort of interesting to think about. It might be different if you’re in the government or outside, but is there a need for that process, and you were just talking a moment ago about is it a legitimate use, well do you or don’t you need to justify that, and so one idea on the table was trying to explore more what was involved in the approval processes and essentially the motivation was for saying before you can get access to the data center data what’s the approval process supposed to do? Does it control a scarce resource or does it try to assure that people have certain credentials or certain uses and partly because, I think, Joan’s suggestion was if it was possible to bypass that approval process either for some categories of people or generically, then you took sometimes a six-months to a year delay out of a process so that graduate student that researcher, that government agency, could just immediately go in and start accessing and using data that are part of these data centers. So that was one idea for a next step. We would have to bring together then the agencies, particularly in HHS, and people who know about that processes and what they are, and why they are and to look at those.
MR. LAND: I can tell you a little bit about those in terms of vital statistics. There’s an agreement between the states and National Center for Health Statistics on how data can be used. Part of the agreement on certain types of research does require approval even of the states, not just the National Center. So, at least for vital statistics there is a requirement for approval.
DR. SCANLON: This is one of those kinds of question where you would like to know more before you address it fully and I m wondering if there would be anything that would be written down about the approval process that we could look at that would give us some kind of understanding about it before we decide to –
DR. STEINWACHS: Well, why don’t I talk with Joan and see if we can pull some things together because right now we’re talking really about NCHS, AHRQ, SAMHSA and CMS are the ones and some of those I know at least for the linked, the cancer registry Medicare data, that process I think we can do on the web, and just complete something and get access very quickly to that linked data set, where other things seem to take a lot longer time. So probably examples – some of these may be driven is what you’re saying.
Okay, so I will do that and then a conference call since I hear everyone’s not taking vacations this summer.
There was another issue raised in the letter was about standardization of agreements and whether, I guess I’m raising the question, is this something we ought to do more with and explore? That may get into then also some of the issues about the laws and regulations under which different data sets have control and therefore why agreements vary and need to vary in that process.
The actual, what you need to do in order to get data sets through the linkage, I guess its sort of taking a step back, was a concern and those seem to drag on that’s what essentially people were saying. Joan’s giving examples out of your office Jim, of things that just w4re not done.
MR. SCANLON: Each agency has its own requirements often because they have their own statutes. I mean in CMS you have to agree to their own data use agreement and its pretty much coming out of HIPAA and their world, and then AHRQ and NCHS had theirs as well. Maybe there’s a model, I mean it would be worth looking at possibly what the content of these criteria for these agreements are now, maybe there’s the possibility for some common model or agreement. I think its generally because for data linkage database that include linked data, its just complicated – you’ve got Medicare data, you’ve got Social Security data, you’ve got survey data, and its, every one is sort of a unique one. Maybe after the first one there ought to be a commonality.
DR. VIGILANTE: Maybe there are components of these agreements that are standard and maybe there are components that are variable and just identifying the variable buckets and standard buckets, and then what the variance occurs around might be useful because going really deep into the variance around individual laws and it might get rather complicated.
DR. SCANLON: Not that you have an empty
agenda, but would this be something the data council would think about looking at?
MR. SCANLON: Yes, it could. We actually had a session on research data centers like where agreement was going, yes you could recommend it. I mean the committee could recommend that the data council take a look at the potential for modeled interagency agreements. Now it wouldn’t help us with census unfortunately.
DR. SCANLON: I’d like to make a recommendation to Alan – do you want to look at it first? – we need a response from the department about our recommendation for standardization, because I’m assuming we’re operating on the knowledge that we have and the department has a whole different set of information available to is and, so before I would want to set out on a next step, I’d like to know what the department thinks about what we’ve suggested in as the first step.
MR. LAND: What is the problem I’m trying to solve?
DR. STEINWACHS: The problems express – use inside the government but also externally and you want to look at something that requires the linkage of data sets that the process of negotiating between – you may have one of those data sets in your control you may not, with the other agency of the group that controls the other data set has become a very protracted activity and something the people talked about six months or a year in order to do that. Well part of at least in the context of the kind of work that Jim’s office does, then it becomes policy irrelevant. The questions here today, we want to know within a month, two weeks, or whatever that time frame is if you were able to link together Social Security information on income with health information, does that show you some gradience, some importance of understanding.
MR. LAND: I was wondering if the agreement itself is the hindering factor or is it the process of lawyers and everybody else getting involved.
MR. SCANLON: To be honest a survey within HHS like the Health Interview Survey or MEPS that wanted to link to Medicare claims data, the agreement is not the issue there, that can be done fairly readily, but it cost funding and it just take a bit of time in terms of queue but if it involves the Social Security or another group or census, it does get more complicated.
CMS over the years, they’ve perfected this and they protect the data and the data use agreement is actually – we have data use agreements with them in ESPEY(?) – identified the data, actually. There are standard elements and then there are elements in terms of we have to agree that that we will try re-identify – that we will protect the data, that we agree with certain confident as we would point everyone to do.
So I don’t know that the agreement, within HHS, I don’t know that the agreement language was the issue there but if you get into Census data or Social Security it really does get more complicated.
DR. STEINWACHS: Well, I’m going to take on Joan because I know this is one of the issues. Why don’t we stop unless there’s something else on data linkages, because I wanted to cover other things in the agenda, not just that. Before Kevin leaves I wanted to turn to SURGE? I gather that you and Bill and Doug have found out about what are the next steps.
DR. VIGILANTE: I know you talked about it this morning, I wasn’t there so if I’m being redundant. As some of you know, we had a fairly large meeting several months ago in which we heard from a broad spectrum of stakeholders regarding a lot of issues around preparedness and response generally to events that evoke responses from the health care system, we summarized that by using the word SURGE, although frankly its even a little broader than that.
We got a lot of input and one of the challenges really is sorting through all of it to say what here is significant, what here is something we can actually do something with that’s within the scope of this committee and with both in terms of mission and its boundaries and also its available manpower to do stuff. I think we’ve kind of in further conversations and reflecting on it really thought we should go down to a smaller subset of the folks that we had at that meeting, particularly from the government side, to kind of look at it from the government perspective of agencies that either through surveys or the grant process or other mechanisms request from health care systems data and information to really get a better understanding from these disport agencies what do you need? What are you getting? What are you getting that’s redundant, and what are you not getting at all? Because I think what we’re trying to get at really is what, because NCHS really comes at this from a data perspective, so to narrow the scope of the work to kind of look at it from that perspective. So, the folks we are really going to reach out to early are AHRQ, particularly Sally Phillips, her agency did a fairly comprehensive survey hospitals, civilian hospitals, nationwide. Her survey were up to 3,900 hospitals got 2,500 responses which is pretty darn good, it was a fairly comprehensive survey about the status of the individual hospital‘s level of preparedness. It hasn’t been published for a variety of reasons but its useful baseline information. Talking to the folks from ASPRs, Secretary for Preparedness Response, that support the hospital preparedness program because they provide about a half billion dollars in grants that go out every year but they also request from hospitals certain benchmarks to be achieved and reported on. There is also DHS that gathers data, we’re not sure exactly what DHS is gathering from health care systems. Who else are we going to talk to, I think somebody from – but we had a short list of government agencies to talk to – we’re going to actually go out and interview.
MR. LAND: Didn’t HRSA do a survey?
MR. SCANLON: That program is moved to hospital preparedness.
DR. VIGILANTE: The hospital preparedness program moved from HRSA to ASPR and that money used to be administered through Department of Health, now it’s going to be administered on a regionalized basis – its going to be a different funding mechanism – and so their benchmarks are going to change. Then to understand others to consider would be JCAHO, American Hospitals – ask hospitals so on the one hand to get a sense of what is needed, what’s redundant, where are the gaps or what’s being asked for in three different ways and then to understand the burdens.
DR. FRANCIS: For a set of perverse reasons, I went though all of the states pandemic influenza plans – that are available on line. Well the perverse reason was I was actually looking to see whether it said anything specific about how much they anticipated there would have to be refusals of care. I was looking to see what their prioritization was –
DR. VIGILANTE: They would have to refuse caring for individuals.
DR. FRANCIS: Exactly, so I was interested in the Tamaflu and in the vaccine question but I was also interested in the question of what they were expecting to have in the way of – say intensive care unit beds.
There are a couple of states, but its very spotty, who’ve done anything about having full information from their hospitals about how much excess capacity they have, if any they have, there are a couple who have thought of things like tents.
DR. VIGILANTE: Alternative immediate care sites, that sort of thing.
DR. FRANCIS: Alternative care sites. There are a couple who have done things like say, JCAHO has an emergency standard which is so called “need based” and there are a couple who have thought about that. New Jersey has a dial-a-psychiatrist plan, I don’t know what supposed to do. New Jersey considers itself a leader in tele-mental health, but its really varied.
DR. VIGILANTE: The variability is amazing.
DR. FRANCIS: It’s extraordinary and it’s kind of instructive to realize how shockingly little actual data the states had when they were doing whatever they purported to have done about influenza planning.
DR. VIGILANTE: Its actually surprising how little good data is out there, so the question is, is it because: A) we’re not asking for it, or B) we’re asking and getting it in isolated pockets that are not shared appropriately where – it kind of gets back to, if not linkage, data sharing among agencies who have common missions that overlap, that cause them to do things but not really integrating the operational aspects of those missions.
DR. FRANCIS: Well if pandemic planning may
be all done state by state so you take an urban area like St. Louis, you have an Illinois pandemic plan and a Missouri pandemic plan.
DR. VIGILANTE: And in reality of course, the
response may be regional and not state – it might be a tri-state area. And these are all sort of the complicated realities of emergency planning and the questions is from the point of view of this committee what can we do to influence a more rational collection of data and distribution and sharing of it with our purview?
DR. GREEN: Since the hearing I’ve had three experiences from this afternoon that have altered the way I was thinking when I left the hearing. One had to do with the manuscript I sent Bill that came out of Dallas/Fort Worth, where the was a very nice mixed methods report of what happened when the Katrina disaster with the people that came to Dallas/Fort Worth where they took four digit numbers, people in the thousands coming into their system and elaborating a proactive triage system that came out of a primary care base that dropped the numbers going to the emergency rooms to double digits and dropped the people going to the hospital, I think, to nine.
We didn’t talk about that sort of stuff in the hearing at all and that sounds important to me, that sounds like – you intersect this with about the sizes of the platforms of health care delivery that we’ve got to mobilize here. The physicians office and the primary care platform that exists that might be worth some thought. The second thing is I had the chance to listen to some presentations from some of the doctors that were in the Astrodome and there was agreement among them that Walgreens and the pharmacy databases were the most useful databases they had to operate off of. That came up in the hearing – someone said something about that, but that pharmacy database and access to it might merit some follow up attention. Then I was talking to Jonathan Tempkey, he’s in Wisconsin, but he’s the guy working the interface between virology and Public Health and primary care and state wide data collection surveillance systems. He’s got – if he had four hands he’d have a hand in each of those four areas. There’s some pretty interesting stuff coming out of Wisconsin about bi-directional communication concerning pandemic influenza where they’re getting this darn close to real time about what’s going on in communities and all over Wisconsin in practice settings and having feedback pumped into the practices to a single system, and that struck me as something that did not come out in the hearings either. So, I have three nominations for follow up:
MR. LAND: I was just in a meeting yesterday on bio-sense and CDC is reconsidering the direction they are going with bio-sense, talking about hospitals and major redirections, and questioning what is the purpose of bio-sense and if they should actually be looking at some of the issues you’ve actually been talking about here rather than ED data and discharge data maybe they should be looking at beds and utilization and these types of indicators. I was really surprised at the bio-sense it was originally – (Simultaneous comments.)
— how they should be monitoring the system after the event actually happens. That is getting into more of what you are talking about.
MR. SCANLON: That would be what I was going to mention the same thing that bio-sense is really, if this is the CDC Program, after a number of years is sort of banging their heads against this wall, I think there is a desire to change so just looking at where it is now and what – I believe we had capacity data, now Katrina – there are folks that say don’t plan anything based on Katrina because its sort of like a one-in-a-century – so it doesn’t really apply to most situations.
So there was very little information on what sources were – the hospital as you know, in an emergency like this tends to be – that’s where people go, its sort of the symbolic in the actual sense of this is where people think they can go to get refuge and care and the emergency room and so on; unfortunately in New Orleans, entire hospitals, charity and others, were out of commission. Other hospitals had services closed and you couldn’t really tell on any given day, or even several days, what the capacity was, and we had folks there and we still couldn’t. You could argue that the state, while it has an involvement is really not going to be all that helpful in a real. I mean there’s a whole theory of how emergency response works, but preparedness is really at least from the hospital side, the emergency department side, where people will look symbolically and others, that’s going to have to be – I mean I don’t think the states are going to do that much and the Federal Government is going to do that much other than support it, and where you were heading in terms of how do you know you’re prepared, what are the standards, what are the dimensions, what do people think it means anyway, and then what data support that, how do you measure it, that’s not a bad way to look.
DR. VIGILANTE: I think people particularly just abstracting chief compliance discharge diagnosis you’re never going to get the kind of sensitivities. Its almost like quality where you just cant abstract from the narrative text the data elements you need to actually measure quality, its sort of a similar thing, so I think when we try to figure out our scope here I think we have to sort of keep it at a somewhat elevated level and hopefully a story will emerge here as we talk to people. One of my own biases is that we will never be able to SURGE to any – once we tell people what the expectation to SURGE II is, they will now be able to plan to get to that point. In Israel the national plan is very every hospital to SURGE 25 percent with the understanding that with the first 10 percent there’s no decrement in care, the second 15 percent there’s going to be different nursing ratios and those kinds of things. It’s a very, very difficult and unpopular thing to do here, but frankly unless we get to that point there will be no way to distribute the load evenly and it could be that we’re better off having a fewer number of more meaningful metrics than a lot of detail about things that just don’t matter that much.
MR. SCANLON: I would add that you should talk to the NCHS service folks who do the emergency department survey because they’ve done one previously on prepared hospital preparedness.
DR. STEINWACHS: Maybe when we together in September there might be a report that comes out of talking to these people that really should go in the next steps somewhere.
DR. FRANCIS: If you want to pick the state that’s done the most I think its Massachusetts.
DR. VIGILANTE: Well I think we might want to interview you. I mean, not too many people have actually done that –
MR. SCANLON: A contractor actually looked at all of the pandemic plans of everybody, they didn’t go into that much detail Leslie, but they looked at what the plans contained or not contained, and what was the range of – and you might just want to have the researcher come in and talk.
DR. VIGILANTE: Sorry to run but I’ll see everybody tomorrow.
DR. STEINWACHS: Next thing that I would like to do and that and it probably puts both Larry and Garland, the newer members of the committee, and Larry leaned over to me and said, public health data – where does that fits into NHIN and where we’re going and how this committee is dealing with that public health and that’s come up, as I told Larry, in some of the discussions when you talk about particularly EHRs and how does that support population health measurement, what are the problems/issues, but I think in part, Larry was laying to me briefly and I thought it would be good to have some discussion, I know Garland has similar interests in — are we doing the things we ought to be doing in trying to support that next stage development that assure that the data sources that are out there feed both public health needs, as they feed clinical care needs, as it feeds others, and so I didn’t prepare you for this Larry, but I knew you were ready anyway.
DR. GREEN: It’s very spontaneous. Let me be quickly biographical. My career mentor, Carl White who was the Chairman of this committee, has been in my face for 35 years and still is down in Charlottesville. He’s alive and well, not really missing much of anything near as I can tell and about 20-25 years back when he and one of his friends was writing that book of his about healing the schism, I felt like I got a Masters’ degree in the origin and the nature of the schism between medical care and public health. It’s been sort of hardwired into me ever since. I bring that to the tour of duty at NCVHS and I’ve now just explained why I’d be interested in having the opportunity to work with you guys on this subcommittee. I’m a physician, I still see patients. I do not have an MPH or a PhD in Epidemiology, that’s not the route I traveled to get here and I often say I am cursed with the mind of a practicing physician. I’ve become more predictable to because I tend to think like a doctor trying to solve a person’s problem. So with that background, I’ve noticed things in the last few year that have created in me a somewhat idealistic spirit of optimism that we might be arriving at a point in development of our health care system and in our knowledge basis where that old schism is might be trying to be healed.
I’m very interested in that personally, but that doesn’t mean the committee has to be but that’s where this comes from. You can see all over the country the active blurring of the distinction between quality improvement and clinical research. The old ways of think about these being nice, pure buckets that an IRB Chairman can use to decide if something needs to be reviewed or not, it’s collapsing. This is falling apart, its not going to hold. It’s going to have to redesigned, its going to have to be rebuilt. But when you start trying to improve the quality of care people get while you’re doing the NIH road map to improve clinical research, all of a sudden tiny little silos between which are schism, I think, start to shatter. A river that flows through it all, to mix my metaphors here, are the statistics, are the data requirements that underline all of this. It turns out that often times the same measurement that you want to make for your quality improvement program is also the same measurement that you’ve got in MEPS that is also the same measure – this just keeps rolling out – increase overworking from the same database.
The data base is distorted and scattered and absolutely insecure in most instances, scattered all over the country, and whatever, but it looks to me as almost trying to happen is a complex adaptive system trying to go to a new level and somewhere in there is a pony. The Population Subcommittee of NCVHS might be around which there could be some type of further activity that tries to reunite – its not that that’s the end game, its sort of a consequence but sort of a guiding, one of maybe many, but as a guiding principle for the work Population Subcommittee is what is it going to take in 21st century information type system, which we all say but we don’t operationalize and we don’t have an operating definition of what do we mean by that, but what if we go to some sort of understanding of an operational definition of a high power, high class 21st century information system – off of which public health, mental health, and traditional medical care were all operating off of the same database. Now that’s a scaling to value of something that is worthwhile in my view and that’s really what gets me in the room. I was just blurting it out to you so is that – I should stop now?
I would be very interested in this groups thinking about what, if anything, can be done to enable this for future use, for the lack of a better term, don’t we need to get back together – get mental health and physical health and the medical enterprise and the public health enterprise – aren’t we really at a moment of opportunity to try to pull this back together and isn’t the information system and the data and statistics that conform from that, isn’t that a mechanism to do that?
DR. STEINWACHS: Well, I think Larry you were probably the transition point to the next committee meeting here. Otherwise Justine is soon going to take me out to the shed. Just to close up, I wanted you to introduce that because you see that also falls into this idea of looking at the conceptualization for health statistics in the 21st century and to try and make that more operational. I also wanted to get Garlands’ part of the conversation – I guess I’ll get that next time.
MR. LAND: This meeting I was at yesterday about bio-sense, DR. Yaznof, he made an interesting statement talking about the health information that you are all moving towards he says “don’t assume that that’s going to be helpful to public health” and I was just astounded when he said that, I mean he probably knows more about this than anybody else and if he’s saying that kind of word then there’s a disconnect someplace in what’s going on in developing that system and so it kind of follows – I see your dream, but here’s an expert who’s still has been developing it saying it may not be there. I didn’t get a chance to talk to him more about it but – it would be interesting. I understand where he sees the disconnect already even before we even have the system that public health isn’t at the table yet to ensure that its needs are being met.
Of course, we have the issue I’m concerned about, your lack of funding for the National Center for Health Statistics, I mean it’s just drying up. The surveys are in real jeopardy, the biostatistics systems’ in jeopardy. It’s the backbone of health data for the nation. They’re talking about some very drastic measures right now because of lack of funding.
DR. STEINWACHS: I think we ought to draw Population Subcommittee to a close. Sorry we didn’t take more time but that’s because our letter took up more time.
(Whereupon, the subcommittee adjourned.)