MEDICAID MANAGED CARE: WORK PLAN DEVELOPMENT AND PRESENTATIONS
PARTICIPANTS:
Subcommittee:
Lisa I. Iezzoni, M.D., M.S., Chair
Hortensia Amaro, Ph.D.
Richard K.
Harding, M.D.
George H. Van Amburg
M. Elizabeth Ward
Staff:
Carolyn M. Rimes, Key Staff
Olivia Carter-Pokras, Ph.D.
Patricia
Golden
Ronald Manderscheid, Ph.D.
Opening Remarks and Introduction - Dr. Iezzoni
Integrated Data Collection: Current Status, Gaps in the Collection Process and Recommendations - Mr. Hitchcock, Mr. Mendelson
Using NHIS to Monitor Access to Care for Vulnerable Populations - Dr. Cohen
Medicaid Managed Care - Ms. Rosenbaum
Performance Measures and Data Needed to Support Performance Measures - Mr. Koshel
Policy and Data Issues and Gaps: Balanced Budget Amendment, Medicaid Managed Care and Welfare Reform - Dr. Ku
SAMHSA: Project Overview, Ties to Core Data Sets and Existing Limitations - Dr. Buck
Encounter Data and Recommendations - Ms. Dodds
Medical Expenditure Panel Survey: Overview and Potential to Analyze Specific Population Subgroups - Dr. Banthin
DR. IEZZONI: Why don't we get started, because the earlier we get started, the earlier we will finish. I don't want to waste the time of Leighton Ku and Jeff Buck, who have come to meet with us.
I know Leighton, you at least were at this morning's session, and so you kind of got a sense of what we were doing. Jeff, I think you have talked to Carolyn about where we are heading. Great. You are here to give us advice and let us learn from you about what you saying within the Medicaid managed care.
So Leighton, do you want to start?
DR. KU: Sure. I initially was told I was supposed to be talking about welfare reform and Medicaid. Then it changed over to managed care. Then the Balanced Budget Act passed recently, so I thought I should maybe say something about. It sort of ends up being a mish-mash of stuff.
Actually in thinking about data on Medicaid, I realized that several years ago I had written a paper that was sort of similar; an overview of data in the Medicaid area, and the overall state of data is kind of shaky, some good spots, some bad spots. My overall impression several years later is some things got better, some things got worse, but the questions have become much more complicated and much tougher. Because of it, researchers who are trying to understand what has been happening in Medicaid have I think, a very difficult time.
I have some little thoughts of things that can be done, but as far as is there an overall solution, a particularly simple one, I'm not aware of it.
Medicaid policy right now is fragmenting at a very rapid pace. The policy locus has shifted even more so to states than it ever did before. Medicaid has always been sort of a state-oriented program, but the emphasis on state flexibility that has occurred means that this has just gone way beyond anything before.
The fact that managed care has now been set up, and generally speaking managed care systems are being set up for adults and kids, but not so much for the elderly and disabled means that in addition, the way we provide care to moms and children can be quite different than the way we provide care to the disabled and the elderly in Medicaid. So it is not just 50 programs; hundreds of programs are breeding, and each little HMO may be a little different from the other.
These can be good policies, and there are lots of reasons why I think state flexibility is a good way to go, and customizing the programs is useful, but it makes it much harder for researchers who are trying to understand big pictures in terms of well, so what is going on, particularly if you want to understand it from any perspective of national policy, or how the variation in states relates to differences of people and outcomes.
What I thought I would do is sort of try to skip through three basic topics: what are some of the main policy areas that I have seen recently; what are some of the types of outcomes that in general I would think health researchers are interested in; and then how well suited are the databases that exist right now to answering those questions?
The first policy area is welfare reform. In welfare reform there are certainly a number of things that have happened as far as coverage to people. Because of the welfare-related changes, the number of people who where on AFDC before are losing AFDC. This has been happening for a while, so people are losing Medicaid coverage because of that.
Perhaps an even bigger shift are the changes in the rules on immigrants. A number of people, really millions of people who used to before have Medicaid coverage, will in the next several years, no longer have that coverage. What are the implications of that? It's sort of hard to say.
Medicaid managed care is another big area. In the Balanced Budget Act, the two big things that happen in the Balanced Act from my perspective -- or three things let's say. One is the creation of the Kiddie Care Program, variously called CHIP, SCHIP or CHIP. I'm not sure what the right acronym is at this point.
States were basically told, put the accelerator on for managed care or given a lot more flexibility for Medicaid managed care. Then there was a reduction in disproportionate share hospital share funding, which has peculiar effects.
On coverage, one of the things that we thought the easiest things to think about in principle are what is happening as far as who has health insurance coverage. Are we gaining coverage or not? So some of these things in principle changes, we might see losses because of some of the welfare reform acts.
We might also see some gains that came from the Balanced Budget Act; things like the kiddie care programs. There are a number of other state initiatives which have been going on; things like the Section 1115, state health reform expansions like TennCare, Quest, Minnesota Care, and so on and so forth.
Last on the scene there is this tobacco settlement looming in the works, which has hundreds of billions of dollars involved, much of which is earmarked to go for health programs, and even more may actually go to it. Potentially, the tobacco settlement may have just a bigger revenue impact on health programs than these other things we are talking about.
There are interactions with private health insurance also. People are worried somewhat about the prospect, as Medicaid is reaching up into upper income grounds, that we may find people who previously had private health insurance opting out of private health insurance to choose Medicaid or some other state subsidy program, because it is less expensive.
There may be some other changes where people -- we said welfare reform -- people might be losing Medicaid due to welfare reform, hopefully the people who left welfare are getting jobs. Some of them will get health insurance. The economy is booming, so one might believe that there are other ways in which private health insurance might not fare quite so badly.
Then the trickier questions become even once you have established that there may be changes in coverage because of some of these Medicaid policies and related policies, what is happening as far as health access, utilization expenditures, and then eventually health status?
The big area where people have focused in terms of thinking about changes in access really have been in what I gather is supposed to be the topic of today's meeting, is Medicaid managed care. As I'm sure people have told you, right now about a quarter of people on Medicaid are in capitated managed care plans. About 40 percent are in managed care, including the PCCM-type programs, and it is growing rapidly.
There are a variety of hypotheses talked about conventionally as far as what managed care does in Medicaid in terms of how it will change access of utilization. I laid out all the different kinds of services that Medicaid provides: primary care services, medical/surgical, inpatient care, ER care, preventive services, allied and social services, mental health/substance abuse.
People worry very deeply. Medicaid was not known as being such a wonderful program to begin with. To what extent is it going to get better or worse in any of these areas? On one hand we would expect that managed care has the promise to strengthen most of these areas. There is also a tremendous sense of foreboding among many people, particularly health care providers that in fact many of these areas are going to become worse.
What I'm calling safety net providers, so places like community health centers, public health departments, large public or charity hospitals that have traditionally been major providers of Medicaid care think, well, gee, we have learned over the years how to provide this care. We were never making a lot of money doing this, but on the other hand, Medicaid was paying it.
As managed care comes in, they are worried about a tremendous loss of revenue, and in addition, worrying about their patients. If their patients migrate to private providers, that the private providers won't know how to provide care as well. So again, they particularly have been particularly vocal in expressing a lot of the concerns that they have about the implications of Medicaid managed care, both on their revenues and on the welfare of their patients.
I mention health status in passing, because it is traditionally always very hard to study health status measures, and to link it very well to insurance coverage. The areas where in principle one might think that it is possible to do something in terms of looking at the implications of Medicaid managed care, and there have been some studies done on things like what are the implications for things like birth outcomes or immunization rates. So that is something that is doable. How well it works with other chronic diseases, I guess I haven't seen that much other there in the literature yet.
Let me skip on to my fourth page now, and start talking more seriously about what the variety of databases are that are available to answer some of these questions, and what, to my mind, they can do; what they can't do; and what has been happening with them recently.
The first set deals with Medicaid databases, and here we have already spent a fair amount of time talking about, and we will quite reasonable spend more time talking about claims encounter data and eligibility data systems that all go together.
So where people have thought the most when they are talking about Medicaid managed care is getting the detailed information that explains what was the nature of care that was provided, what were the diagnoses, et cetera. How can these fit together in some ways fairly comparable to the claims data?
They can be used to answer a variety of interesting questions. I laid out a couple. You could ask, for example, do the kiddie care kids receive the same types of care as the kids who were already in Medicaid beforehand, or do they get something different? How are they in terms of their risk nature, or the level of expenditures that are necessary for them?
These are things that in principle you could answer with those data. You can ask does the content of care change when you switch from fee- for-service to managed care in a given state? This is, in principle, something that is doable. It's what we are trying to do in some of this evaluation of Section 1115 waivers that we are doing right now for HCFA, and there are some problems, and I will get to those in a moment.
The generic problem with the Medicaid databases is that they only deal with Medicaid. They don't deal with people before Medicaid or after Medicaid, so some of the things that I think are the most interesting questions, like what happens to the people who lose Medicaid because they have lost welfare, or because they are immigrants, what do they do for care? Got me. They fall out of the system.
Similarly, what do we know about the kids who are brought in because of expansions, what were they doing for care beforehand? Again, the big black hole.
If you want to compare how does Medicaid compare in terms of the types of coverage or the types of care provided relative to people who are uninsured or commercial insurance, again, this highly detailed, very huge, expensive data system really can't help you at all.
The other big disadvantage is that they are state-specific and they are very cumbersome to use. HCFA has done a number of things, and I gather because of the Balanced Budget Act there are more mandates to try to turn things into a somewhat more uniform data system and eventually have national coverage. That should be helpful. They will remain huge, mammoth databases that are not easy to use, and because of that, there have really not been all that many analyses that were done to the Medicaid claims bases heretofore.
In the issue of switching from fee-for-service to managed care, which is at the heart of where many states have been going in their Medicaid programs, in principle you should say, well, gee, if encounter data are capturing pretty much the same kinds of information as claims data, that is, there is a transaction that reports in some manner, shape or form, this patient had this particular problem, received this set of services from this provider, you would think you should be able to go back and forth and say, well, what exactly did managed care do?
Did it really reduce the level of inpatient care? Did it really reduce in some manner, shape or form emergency room use? Did it increase primary care use? Did it increase preventive care use?
The problem is no one is very clear yet on whether claims data and encounter data are comparable, just to the extent that there may be enough changes in terms of how the data are collected and filtered, that some of the may become quite touchy. Certainly, as was mentioned earlier today, certain areas like especially inpatient hospital care should be pretty consistent from place to place, before and after in systems. Even that may not be thoroughly the case.
The other big problem that people have mentioned about encounter data is it takes a long time to get the data in. Our Section 1115 projects, which Sue Dodds will be mentioning, out of how many states are there in the first five state project -- five states. We only have data from Tennessee at this point?
MS. DODDS: Right.
DR. KU: So we're in the fourth year of this project now?
MS. DODDS: This is just the end of the third year.
DR. KU: End of the third year, and so we have information from one state in at this point.
We are involved in another Section 1115 evaluation also that has another set of five states, of which one state has data ready yet. That means it is there. We don't know how good it is yet, but there is something there.
So we don't know how long it takes to get good encounter data. Certainly, there is a fairly long time gap. The lapses between the time a policy decision is made, that is the decision let's move forward into managed care, and the time the data comes out from that pipeline, to be able to evaluate what is going on.
So answering some of the most immediate and interesting questions, it is not clear how feasible you can make those comparisons between pre and post, and for sure, you are going to have to wait a while for those answers.
I guess the particular problem -- maybe other people mentioned this -- that comes up in encounter data is the risk of what happens when people and providers themselves are capitated. For inpatient hospital care, it is easy to believe the system should generate the same types of data, but for the primary care doctor, if now he is getting $15 a month to manage the care for this patient, and he is going to get paid regardless of whether he submits claims or not, what incentive is for this provider to ever fill out any claims forms? Consequently, are those systems going to disappear completely?
There have been traditionally a number of administrative reports that most Medicaid analysts at the national level have relied on heavily. The biggest data system, which I gather was discussed in an earlier meeting is the HCFA 2082 statistical reporting system. It is an annual report that compiles a lot of statistics on the number and type of Medicaid recipients, enrollees, the services they use, and the expenditures.
Frequently, it is coupled with another form whose lovely number is 64. The 64 report is an expenditure report. Sometimes you try to make the expenditures line up with the people.
For years and years and years these have been the major ways that people could figure out how many people there on Medicaid, how much we spent on them, and most of the sort of monitoring information that you think is really the most straightforward and simple-minded kind of thing that you could hope for.
Capitated managed care has played havoc with the 2082 system, because people are no longer in the regular care system, so you can no longer link individuals to specific services, like getting inpatient hospital care, or getting physician care, or anything else. So the 2082s have become less and less useful over time because of this. It is not that there can't be ways to get past some of those problems, but it has been a problem.
There has also been the problem that in expansions, things like some of the Section 1115 expansions, and then I'm not exactly sure what is going to happen with kiddie care, there have been inconsistent federal policies about how to report these expansions, with the net effect being that you might have some states that have just actually had tremendously large expansions, apparently have massive decreases in enrollment because they believe the expansion populations that we moved into managed care, we don't need to report them anymore. They are not part of the regular Medicaid system.
While there are other states that have had expansions, suddenly you see their caseloads apparently increase radically, because they assumed, well, you are supposed to report them. Simple things like common decisions about how to report these new expansion groups would help a lot.
Another source of information which has been increasingly important in recent years is their annual Medicaid managed care reports from HCFA. The basic thing they do is they report the number of enrollees in each plan on a given day and in a given year -- June 30, 199x -- with a relatively small description of who is in the plans.
I gather that HCFA is planning to revamp that system. Again, it is useful as being a fairly basic count of how many people are in managed care, and that's where we can generate these simple statistics. The lack of information it has terms of how much we paid for the care, what kind of services are covered within that care, et cetera, et cetera is astonishingly broad -- the gap. So there is really very little information in terms of the types of the people in the plan in addition.
Where I guess I think that there has been potentially useful information that can be helpful in understanding the changes in Medicaid from the perspective of contractions and expansions are some of the national household surveys. These are things like NACJS(?), the current population survey, maps, so on and so forth.
To me, their key strength really is the fact that they have broad population coverage, so you can look at Medicaid people compared to non- Medicaid participants. You can do things that are then interesting in asking were the policy changes, things like expansions, et cetera, or contractions, how did they affect people's lives in the sense of how much care they got, and so on and so forth.
There end up being two big constraints for the same virtue of their national breadth. It means that there are limited sample sizes, and so the number of people who are on Medicaid, or who are low income in general ends up being limited, particularly if you want to represent states.
Again, because policies by state are changing, and at least at this particular point appear to diverging, there is a clear and reasonable interest to believe we would like to have state identifiers going on with each of the databases, so that people can understand, well, gee, is this because there was a special policy in Texas versus some other policy in New York?
On the household surveys, their big constraint is the respondent is someone who in the household. The person who is in the household has an unlimited amount of information from the perspective of the true medical researchers who would actually like to get in the medical records and see all the diagnostic information and so on and so forth, and the clinical data, that's just not there at all.
But there can be some limited information in terms of things like coverage; a little bit of information about utilization; but it is limited. One of the things that is a problem again with Medicaid managed care that got brought up by some others is that people in some cases, are not sure if they are in an HMO or not, or some other kind of a managed care organization.
In addition that, may not be sure if they are in Medicaid or not. Again, if they have a Kaiser card, they may think from their perspective that they are in Kaiser, and they are not in Medicaid. It's not that they didn't go through the Medicaid eligibility process. They know they went through that, but somehow in their minds the word "Medicaid" evokes something else; evokes specifically that you have just a Medicaid card, and you went to the fee-for- service system.
So simple definitions have become trickier and problematic. It was nice to hear that there is some testing going on to figure out whether those questions are valid, and how to improve them.
Even when you have some of these basic questions resolved, I think an interesting question is how do people on Medicaid differ from those who are uninsured, and a number of parameters. You still have problems with selection bias. The reasons that you might find substantial differences between those on Medicaid and those not on Medicaid might not have anything to do with Medicaid, but might have to do with the fact that there are different sorts of people.
For example, routinely there have been analyses that have been done that show that people on Medicaid appear to have worse health by various measures than people who are uninsured. You have to think, well, gee, did Medicaid make them sicker? Wouldn't that be a terrible problem?
Then eventually you think, oh gee, it really must be that there is a selection bias going on, and that in fact the reason that people Medicaid appear to be sicker isn't because Medicaid made them sicker, but because they were sicker, and that's how they therefore got on Medicaid. That's the downside with the national household surveys, is you can caught that in particular rut in terms of how the design is set up.
Let's go through at least three of major surveys that have been used heretofore. The most commonly used to measure insurance coverage has been the Current Population Survey. It has fairly good socio-economic information. It is done routinely, has big sample sizes, and yet even so, runs into limits on a number of states.
A big problem there has been their routine insurance question has been essentially did you have these different kinds of insurance. Then if you care about who is uninsured, it has to be the residual of you didn't answer any of the other questions. It is done in the context of the prior year.
There have been a number of people who questioned is it really picking that concept up correctly, people saying they have insurance or don't have insurance right this very moment? So the concept is good, and it's been the most routinely used database for that purpose, but it does have some measurement problems.
As we heard earlier this morning, the National Health Interview Survey, another very big database. It has been reconfigured to be at least somewhat more state-specific, and to ask a lot more health care questions. I think it is very promising.
A problem that I know other people have had in the past in using NHIS data, and hopefully this is getting better, is for researchers to get access to these geographic identifiers to try to append the state policy variables onto the NHIS files. If you can't add in or identify states, it is really quite hard to try to figure out what the context of Medicaid policies or other regulatory policies that affect health care are. That has been a fairly big limitation from the perspective of researchers heretofore.
I gather a little later we are going to hear about the MEPS Survey, which is replacing the National Medical Expenditure Survey. It again, looks very promising. The sample is not as big as NHIS or the CPS, so it is not so good for state-specific analyses, but on the other hand to the extent it can provide medical care expenditure information, which is almost certainly one of the hardest topics to cover, it can be quite useful.
I guess one thing that I'm not clear on is how MEPS is handling medical expenditures and gathering information on medical expenditures for people who are on Medicaid or who are uninsured, or how it is being measured when people are in capitated care systems, and there are no bills or something to look at. Maybe you can talk about that later. I don't know the answer to that.
One general class of survey that I haven't heard people talk that much about in this context, that actually in some respects strikes me as one of the most useful classes of surveys are provider surveys. The reason why I think provider surveys are worth considering is that ultimately whether it through managed care or fee-for-service, or whatever new system that is as of yet uninvented or not disseminated, there has to be some level at which there are health care providers -- and here I mean doctors, hospitals, clinics -- that are providing care to people.
If there are better ways of gathering from them, information about what has happened, how much care they have been providing, the characteristics of people they are caring for, et cetera -- and here I mean very simple characteristics like Medicaid versus uninsured versus commercial insurance -- that in and of itself could help on a real time basis that you can't do with the encounter data.
Again, one of the big problems with encounter data and claims data is most analyses that you read about are looking at data that are a few years old, and a few years is actually relatively soon in many cases relative to these time frames. Provider surveys could be done relatively quickly.
So in principle you could answer questions like is Medicaid managed care associated with fewer Medicaid inpatient admission or with shorter stays in hospitals. That's a useful question. In principle you might ask, are some of the welfare reform eligibility changes leading to more uncompensated care?
The two main data sources that can be used right now on these is the American Hospital Association has an annual survey of hospital, and the Bureau of Primary Health Care has a database for community health centers which is now called the Uniform Data System.
Each of those has some basic information about Medicaid volume that goes on that can help us figure out simple things in terms of how is Medicaid affecting these types of providers, and therefore if you know how it is affecting the providers, you can infer what is it doing for the system of health care for the poor.
Managed care might be messing up the AHA data. It is not quite clear. Again, the problem that we mentioned where patients don't necessarily know if they are on Kaiser, whether that means that they are on Medicaid or not on Medicaid, also applies to hospitals.
When I have gone to many hospitals and talked to them about things, they keep on pointing out, well, gee, right now we still have the Medicaid managed care under their commercial lines of business and their basic accounting system. So when in AHA data it is asking about things like Medicaid inpatient admissions or stays or revenues, it is not clear to me how much of that is including the managed care component or not. I think it could be done fairly simply if AHA would make that definition a little clearer.
It seems to me that from a provider survey perspective, the big gaps are surveys of on one hand, physicians, because they are the people who do provide care. It seems to me in principle one could design a survey about physicians that could really be done on a routine basis, that can answer a lot of the useful questions. They are the ones who will, under whatever system of care -- are in charge of a bunch of the care.
The other, to my mind, just amazing gap is lack of information at a national, somewhat uniform level about how much and what types of health care are provided by state and local health departments. As I have gone on talking to a number of states and health departments, there are tremendous fears and anecdotal stories about what is happening to local public health departments because of Medicaid managed care.
It is exceptionally difficult to find information. It just seems to me that that is an amazing gap considering that these are people who we do think of as being safety net providers, and in fact they are primarily publicly funded. That is something in principle we should be able to get information on more easily.
There are a number of special projects that have been begun in past years to look at a number of topics like these. The Urban Institute has a fairly large project called Assessing the New Federalism. Some of you may have heard of it. It goes by many guises, because it is such a huge project.
Basically, the purpose of this project is to look at changes in states and across the country as welfare reform and other things that have devolved authority to the states proceed. We are doing a large household survey in 13 states, and then a wrap around component for the rest of the country.
There are a number of questions about health insurance status, health care utilization, access to care and health status, as well as questions that are relevant to welfare, social services, so on and so forth. It is a fairly good survey. It has a sample size of about 50,000. We are doing one survey right now. We plan to do another survey in 1999.
There are other components to the survey that involve site visits to the 13 states, development of a 50 state database, which I think is actually accessible on the World Wide Web. So this is a fairly large situation where we designed a survey that can be done as a pre/post survey, focusing in certain states, but with some national wrap around capability.
The other project that I have been involved in along with colleagues at Mathematica and the Center for Health Economics Research are a number of evaluations that HCFA has been sponsoring on the Section 1115 projects. I list in my handout, who are the various states that have fallen into these projects. There are three projects, two of which have five states -- Tennessee, Hawaii, Rhode Island, Oklahoma, Maryland in one, Minnesota, Los Angeles, New York, Illinois and Vermont in the other. Then Oregon has its own project.
Despite the fact that we are fairly far along in these projects, the primary source of information that we have had so far really has been case study site visits. The process of getting the encounter data has been quite slow. We are planning to do phone surveys of program participants, and in some states, low income comparison populations that cover a number of the basic questions, that try to look at the impact of the expansions and of the shift in managed care.
I think that there are some other projects that I don't necessarily know a lot about. I know that Sara's shop at G.W. has a fairly large project to look at the impact of welfare reform. I think one of the main things that they are just trying to keep track of are what are the state policies that have occurred in adopting these. I think APWA is doing something fairly similar.
The last area in my mish-mash of stuff is thinking about potential new data sources. Again, some of these we have been trying to dig at for some of the purposes that we are undertaking, like our Section 1115 evaluations. Managed care is routinely generating certain types of routine reports, which in principle have information that should be useful.
A lot of them, because they are still new systems, the data quality is not so great, or there are non-uniform formats, which makes them hard to compare from place to place. Nonetheless, in principle more work can be done on them.
I think one basic one, which I haven't heard people talk very much about are mirror financial reports. Things like how much is a health plan spending on new health care, versus on administrative costs and profit, and then potentially can simply define some basic categories, things like inpatient care, physician care, drug expenditures, simple things that would help us look from plan to plan, or across time, how much of the basic content of care.
Again, this is aggregate data. It is not wonderful, but on the other hand in a real time basis can give us some useful information relatively quickly.
A lot of people have talked about HEDIS data. HEDIS data I think is again promising, but the initial reports that I have seen so far coming back from state are pretty disappointing, lack encounter data. It is hard to get the systems up and running, and the first generations of the data lead to peculiar findings that make you realize the data systems just aren't working quite right.
Even when they are up and running, you may still have some puzzling questions about well, is the care better or worse than it used to be? For example, I can think of one state that had some findings that indicated that their immunization rate was about 60 percent, whereas they hoped that they would get to a standard of around 80 percent.
A question that I asked them, well, gee, do you think a 60 percent is better or worse than it was before under fee-for-service? There was no answer, because they never bothered to collect that information beforehand. So you would have a number, but within context.
The other area of information that can be used potentially, and I don't know if anyone has tried very seriously to use these from a research perspective, are the consumer satisfaction surveys. I think in many cases the response rates tend to be fairly poor, so they may not be the best quality surveys.
On the other hand, to the extent that they can be used in some meaningful fashion to compare plans within a state or across states, again, there is potentially information that could be very helpful in at least understanding how managed care plans stack up against one another, or over time, though again, the comparison to fee-for-service would still be pretty much up in the air.
So that's it. There are lots of holes. I don't know that there are any perfect solutions.
DR. IEZZONI: Thank you, Leighton.
Before we move to see whether there are any questions or comments from the committee, can I just ask are any of the other panelists pressed for time? No, okay, Jeff, we have run over a little into your time.
Any comments or questions for Leighton?
DR. HARDING: Is the issue of financial reports that you brought up, is that a proprietary problem?
DR. KU: In some states it is; in some states it is not. I mean I know in the four states that we have recently been looking at financial reports, when you ask the state, is this proprietary information, two of them say yes, two of them say no.
I don't know that there are any uniform rules. Again, it seems to me that in principle, public funds are paying for all things. I don't necessarily know why they should be proprietary, but they have agreed to it in the first place, and so therefore it might stay that way.
So the confidentiality and proprietary information is potentially a problem. In some cases though, you can bypass some of these problems by aggregating even further up, like aggregating at state levels, as we have been doing for one recent study.
DR. HARDING: The other thing I would just comment on is the last one, the consumer satisfaction rates. I haven't seen any data lately about that. In our area what they found when they did those was that people were satisfied, as long as the didn't utilize the services. The cost was down, and so about 80 percent were satisfied, but they hadn't used the services very much. Those are tough to figure out.
DR. KU: That's a problem. We have run some projects -- we ran a number of focus groups in which we found that people who were low utilizers of care were generally quite satisfied, and the people with chronic diseases were less satisfied. Again, the pre/post problem; were the people with chronic problems less dissatisfied when they had them before?
DR. IEZZONI: Hortensia, you had a comment?
DR. AMARO: Yes. Thank you for your presentation. It was helpful, although it does keep getting more overwhelming as we hear of all the problems.
I think one of the additional things of the limitations of the National Household Survey is the fact that they don't really give us longitudinal data. For everyone there is a subsample, I understand from the National Health Interview Survey that gets followed. Otherwise am I correct in assuming that none of these would allow us to follow people over time to look at access to care over time and how that might change?
DR. KU: The ones that you are talking about -- I'm not sure about the longitudinal component for NHIS. Does anyone else know about that?
DR. IEZZONI: For some special studies there have been, like the disability survey there was a follow back.
DR. BANTHIN: The MEPS follows people for two years.
DR. IEZZONI: NHIS doesn't do it, but these special surveys do.
DR. AMARO: Right. So that would be the only one that would allow us to follow. So that's an additional problem, with just relying on the information they give us.
DR. KU: That's true. SIP is longitudinal, but SIP doesn't have that much health care information.
DR. IEZZONI: Leighton, it seems like you are part of this kind of industrial network of researchers who kind of have this on their radar screen. I guess Sue, you represent Mathematic, and we had George Washington this morning.
Are there any specific areas that you as a researcher have come across that you think more policy visibility that our committee could potentially help with, specifically around the issue of measuring the impact of Medicaid managed care?
DR. KU: Well, I mean as far as I can tell the main tool that your committee has is sort of raising visibility, as opposed to actually ordering anything. To the extent that you can do those things at a DHHS level, I think that is useful.
DR. IEZZONI: Specifically on what? What are the gaps?
DR. KU: If I had three little things to go for -- and again, it's lots of little disparate things -- one, I really would like to see some state identifiers coming out of NCHS data, and the ability to link that more clearly, and other national databases. If you can't get state identifiers on the data, you can't append the policy variables of interest for Medicaid or other regulatory policies.
I think there needs to be more thought about certain types of provider surveys. Again, it seems to me that the two gaps that I thought that are the biggest gaps are areas like physician surveys and the amazing lack of information on what state and local health departments actually do as far as providing health care services are areas that could use some help.
There are lots of people thinking about encounter data, and so I hesitate to say that necessarily there is that much that can be done in a simple way there, though Sue I know will come up with some very good recommendations in just a moment.
It might be thinking are there somewhat faster ways of getting data about Medicaid managed care like some of the financial reports or so on and so forth, that might get us somewhere close to only being a year or so out of step with the time, rather than being several years out of step with the time. Sort of quicker data collection modes and reporting modes I think would be helpful.
MR. VAN AMBURG: Leighton, can you describe a little bit more about what you would put into a provider survey?
DR. KU: Well, it depends on which kind of provider. I mean I think you could start with the base of information that AMA has in a socio-economic monitoring system, but there is not much that is relevant to asking some simple questions about things like which Medicaid managed care plans, or how many Medicaid managed care plans do you participate in?
If you accept Medicaid patients, under what terms do you accept Medicaid patients? Do you accept any who walk through the door? Do you limit the number of patients that you see? If you have just changed from fee-for- service or managed care, there might be questions along the lines of have you changed, whether you provided the care or not before.
To me, one of the simplest questions that we ought to be able to answer that is unfortunately hard to answer is, are there more or less doctors providing care under Medicaid managed care than there were before? Certainly there were a lot of providers who didn't want to go anywhere near Medicaid.
The anecdotal impression that I get today is that there are fewer providers, there are fewer doctors providing care, but I can't say that for sure. It seems to me a survey would be an easy way of getting at that -- or maybe not so easy, but a direct way of getting at that question.
DR. IEZZONI: We're running a little over, so why don't I not call for audience questions at this point.
Leighton, thank you. Hang around.
DR. BUCK: This might be good following Leighton, because I realized that a few of the things that I am going to say are elaborating upon some things that he brought up, and I think on one or two points, disagreeing with him. Actually, I think in such an exciting area as data, having one or two disagreements actually helps make things more interesting, and makes us think more about what it is we are trying to do.
Let me tell you just briefly a little bit about where I am coming from. I am the director of the Office of Managed Care in the Center for Mental Health Services within the Substance Abuse and Mental Health Services Administration. I think by the way I got to this presentation is that before I was in SAMHSA and in that position, I worked three and a half years in the Office of Research in the Health Care Financing Administration, where part of the time I was the project officer for the Medicaid tape-to-tape project.
Before I was there, I spent five years working both as a researcher and as a program manager in the Medicaid program in Illinois. So even though now I am outside of HCFA and outside of Medicaid, I have a fair amount of familiarity both with Medicaid operations at the state and federal levels.
Also, I have been, in my bureaucrat hat, a supporter and funder of research projects on Medicaid. Right now in SAMHSA, our office is participating in about half a dozen that in one way or another deals with Medicaid managed care or mental health and substance abuse services in Medicaid.
Also, personally I have authored about a dozen studies that have dealt with Medicaid services or Medicaid financing. So hopefully that will help you understand kind of my orientation.
I thought I would make my major points first, and then talk a little bit about it, because generally although I understood that the focus of your interest here today was on Medicaid managed care, I think perhaps this is an area that is like some other areas of health care in the sense that I think we now more and more think that there was some magic, golden time when everything was kind of working right, and so on and so forth, and all these evils with managed care came along, and now we've got this big mess we've got to address.
When in fact there are some very long standing problems with Medicaid data, and certainly Leighton alluded to a number of them. I kind of want to talk about what I think some of the reasons for those problems. By Medicaid data I'm not talking about all sources of data that can produce some sort of information about Medicaid, but I'm rather talking about data that is a direct product of the program, of the ones that come out of the state systems, and the ones that are maintained by the Health Care Finance Administration.
The problems with Medicaid data, I think the other major point I want to make, and this is where I think it might really be helpful for a panel such as yours, because the current problems with Medicaid data as I view it, result from issues such as what are priorities for HCFA, and also priorities for the federal government, and for the Department of Health and Human Services.
What is the role of the states vis-a-vis the federal role in this area? What is the reasonable amount of resources that should be devoted to this kind of effort and area in order to do it well?
Then I think another general area where there is a problem is that there is confusion about what it is we are trying to do with this data, and issues about who is responsible, accountable, and owns it.
I think these kinds of issues are more important than ones that I hear talked about more commonly right now, which has to do with things about standardizing encounter data, or dealing with issues of confidentiality. I don't mean to say that those aren't important. They are important, but rather the issues that are involved in trying to answer important questions about the Medicaid program, those are not the most important issues that need to be addressed.
Another point I wanted to make was that I have only recently learned about this, but I understand that there are now some requirements for states to participant in the MSIS system, and also to provide encounter data. Leighton said that this would be helpful. My general first reaction to that is that it is going to be unhelpful. In fact, it is going to be detrimental.
The reason for that is that unless some of these other issues are addressed, particularly issues of resources -- let me add a cognate. I want to make it very clear that the things I'm saying are not meant to be a criticism of the particular staff at HCFA.
Generally when I have worked with staff at HCFA, both when I was in HCFA and also without, that people are very helpful, but in fact are very hampered by issues such as the lack of resources, and the failure to assert priorities. So it's not like these issues are because people aren't doing their jobs.
I think a lot of it is that there is a lack of investment and direction at the very upper level of the department and at HCFA, but unless some of these issues are addressed, the legislation now that appears on the surface to be improving matters, in my view is going to make them worse.
I just want to mention a couple of things. This is partially important, because I am just really recapping just a couple of things; maybe just two things here. We sometimes I think overlook how really a fundamental part of the overall health system Medicaid is. It represents about 15 percent of all personal health expenditures. It is a major item in the federal budget. It is the second -- at least this is of a couple of years ago -- it was the second largest item in state budgets.
This is not some little public health clinic program. This is a major part of our overall health system. For us in SAMHSA it is very important, because if you put all our spending in SAMHSA up against estimated mental health and substance abuse spending in Medicaid, we are this little itty-bitty sliver. It is not even a decent helping for dessert in terms of the overall part of the total public mental health and substance abuse spending.
Yet here we are, we have a whole agency devoted to mental health and substance abuse services, and yet the dollars that we control or direct or otherwise have oversight from are dwarfed by what is provided within Medicaid. So we are obviously very concerned about issues about what is happening in Medicaid, and have a high need for this type of data.
The other thing I would point out is Medicaid is the health plan for about one-third of preschool children in the nation. I'm not talking about poor kids; I'm not talking about people who are getting income support or anything like this. This is for our entire nation's children, a third of those at the preschool age level have Medicaid as being their source of support.
I think even if you don't care too much about the mental health part, I think you should very much care about what we know about how well this program is working, because of where it is focused.
Well, I think a very important possible focus for your group would be focusing on, and articulating what it is that we should expect to come out of the Medicaid program in terms of data. In particular, the kinds of data we ought to have and maybe also sometimes I think even before you get to that point, the types of questions that the Medicaid program data ought to be able to answer, because I have not seen that articulated anywhere, and I think the failure to articulate that kind of standard, which in some ways you don't have to know anything about Medicaid and Medicaid data to be able to do.
You can just say, look, if you have a decent running program that is providing information about how it is run, here are the types of questions are kinds of questions that you should be able to answer with it. If you can't answer these questions, regardless of how you generate these data, you're falling short. Nobody has ever articulated that standard, and then assessed the current data system against it.
I'll just outline some basic options. Certainly there is one part of the program that seems to think that the only people that really have the use for this data and need to worry about it are the states. You can certainly take the attitude that because of the state administration and participation in the program, that if you want to know something about what is happening in a state or states, we'll go to the states and ask them for it. That shouldn't be a worry of the federal government. That is a legitimate point of view.
If however, you think there is a need for federal data, what sort of data do you want to try to get? I guess throughout all of this there is an assumption of how good should that data be? Do you want just some general administrative reports that gives you for instance data that says how many people are enrolled in managed care, and not much else; that says here are some categories of spending by major service categories, and not much else? That is one level of detail.
You might also go at it from saying, well, it's very impractical to try to do this for the nation, but maybe we will focus our efforts on a few states, and try to in essence come up with a general sense of what is going on in the program from those few states, et cetera.
So I think one thing that would be very helpful for you to do is just to say, you know this is what is reasonable at the federal level to do with these data, because I have never seen that debated. I have never seen it settled. I think many people assume that Medicaid is there at a certain level of comprehensiveness and quality and so on, but I think often it's the case where they are disappointed when they find out what is there and what's not.
So what are some of the major problems with Medicaid data, after having that sort of conceptual overview? I think there are several in my experience. A very important one is that there really isn't any centralized accountability for the overall Medicaid data system. You have a set of responsibilities that the states take on. The people that work with the states that have an overview of the state MSIS systems, who assess the data in those states, are not the same people who are responsible for maintaining the data at HCFA's level or processing that data.
Those people are not the same people who use the data within HCFA. There isn't anyplace that I was aware of in HCFA when I was there or since where you can go and see a single point of concern, responsibility, accountability, however you want to characterize it, for the entire system. Because of that, there are just all sorts of opportunities I think just as a natural outcome of a very large organization such as this, for in essence, different points of view, orientation and so on and so forth, to result in gaps, failure of coordination, et cetera, et cetera.
In fact, there is not even a real impetus to do coordination. There is nobody saying that it's a problem that certain things don't necessarily feed into certain other things.
Second, and I think if I was trying to push you towards something, I'll come back to this again in a few minutes, there is no comprehensive plan for Medicaid data. By comprehensive, I have a particular view of what that means.
First of all, there is no plan that first says, this is what we are trying to do by having Medicaid data. This is the kind of detail we are trying to have. This is the kind of quality standards we are trying to have. These are the kinds of questions we want this data to be able to answer. Do we want to do it just as a sample or state by state? Have detail by diagnosis? Not have detail by diagnosis? There is no articulation of a standard by which to measure what is the output of this entire process of generating data.
Additionally, there is no comprehensive plan in the sense that it addresses every single level of data and following it through all the way to what would be an end user product. The people that maintain files, they do have plans for how they are developing SMURF(?). They do have plans for how they are developing MSIS, but their plans do not extend to the standards that are used in reviewing MSIS systems, because that's not part of their job. That is somebody else's job.
So there is not a plan that says, gee, if we are really going to have this work well, here is our plan for addressing every single level of the system.
Thirdly, and that sort of goes along with the other, just as there are not goals for the overall system, the existing goals and standards -- and there are many -- are basically piecemeal and inadequate, depending on which level you are looking at. Once again, I am assessing this from the point of view of a potential user or researcher, as opposed to people who may be doing other things.
I'll just give you one example. I don't know if this has changed. Some of this I haven't stayed on top of, so it would be wonderful to know if some of these problems have been addressed. The standards that were used for reviewing the adequacy of the state management information systems primarily focused on issues such as is this a legitimate claim? Was it paid timely? Is the person it was paid to a legitimate Medicaid provider, and basically looking for things that largely focus on was this program being administered well in the context of receiving and paying claims.
It didn't ask questions like is this a garbage diagnosis on this? Is this a valid service code? Are there other kinds of things that are missing from this file that would be very important for the questions that Leighton was outlining? But the standards existed. There were standards, and so on.
I think another problem here is that it's not clear who the customer is supposed to be for this data. Is it supposed to be primarily for HCFA's internal use? Is it supposed to be available to policymakers to be able to answer policy questions? Is it supposed to be there for researchers, to ask questions for instance about the adequacy of current care for long income children or pregnant women or mothers? Who is it supposed to be good for? That is not well articulated.
Related to that, and I think what is giving rise to that particular dot point, there are a number of current practices that tend to discourage the outside use of the data. I understand the reasons for many of them, but they nevertheless serve -- you have spent all this time developing this data. You have certainly put a significant amount of resources into it, yet then it's hard for people to use it.
People can't necessarily tell you when data is going to be available, at what times, for what states. The charges that are imposed for using the data most people view as very, very high, so that you really had to be a very well funded or have a very rich agency just to look at a few states for one year. It gets hugely expensive to purchase data if you want to do much more than that.
Related to that I think is something you might also address, this actually says to me as I started to talk to a few people about this is that there is no universal department policy for sharing data or using data between the different parts of the department. To give you an example, generally within SAMHSA we do not charge people to use data that we generate from our efforts, and have several databases.
Now granted these databases are fairly puny in size and complexity compared to Medicaid data, but generally the policy is to make them available to people who might be able to use them. In contrast, for us to use Medicaid data from HCFA, we have to pay money, and we have to pay a lot of money. I think it is worth questioning whether this is to the department's advantage to have this kind of policy.
Another thing -- and I think this maybe gets to the heart of the problem -- if you understand the general characteristics, and what is kind of necessary, if you would just kind of go through in your mind what you think it takes, as the legislation as it has just been passed requires all states to provide data to HCFA, and for all that data to be processed, and to do all the things that need to be done to accommodate the different programs, variables, coding, data problems, missing data, data gaps, et cetera, et cetera.
Compare that to the general similar kinds of processes that are needed for Medicaid, where there is only one program, one set of policies, where HCFA has much more potential to influence the nature of that data, you would have to know that if you were going to do an equally good job with each, you would have to devote more resources to Medicaid data than to Medicare data, yet the reality is exactly the opposite.
Medicaid just simply does not get anywhere near the resources for producing that data that Medicare does. There are some reasons for it, but what happens then is that you fall way, way short in terms of quality. This gets to the issue I was talking about, about how this legislation could be worse.
Because in my mind the legislation that says that everybody has to provide data to HCFA means that inadequate resources that are already there for processing and developing Medicaid data files are going to be spread in a certainly thin level across a much greater number of states; approximately double the number if I understand things correctly. I doubt that there are going to be even double the resources to maintain what is already an inadequate level of resources for that function. So that is a major, major problem.
Finally, the other major problem here kind of gets at also how serious you are about this, because generally there doesn't appear to be really good incentives for states to produce good data. As Leighton already mentioned, obviously people are just enormously curious about what is happening in Medicaid managed care. There are all sorts of concerns.
Advocate groups are really worried that essential services are going to be lost, or they are not going to have access to things they need, so there is just an amazing hunger for information about what is happening with this program. As Leighton said, they have only gotten so far, encounter data from Tennessee.
HCFA and the department, because these are demonstrations, just to get to the bottom line here, these demonstrations can be terminated at any time by the secretary. There is no requirement that they be granted in first place.
So the fundamental authority there is there within the department to require good data, or to have other types of incentives to get it. The question is, how serious are you about the need for it?
DR. IEZZONI: Jeff, we should try to finish up.
DR. BUCK: Okay. Actually, I'll stop right there then, because I think you have a copy of my other stuff. I think most of the things I have outlined in terms of barriers lead fairly clearly to suggestions for how you want to go about things.
Do you want me to spend a minute on those?
DR. IEZZONI: Why don't you give us the highlight on the what is needed page?
DR. BUCK: Fine, yes, that sounds good. That is the very last page.
Let me just focus on a couple of these, which I have already touched upon at one point or another. I don't know if anybody has talked about GPRA. If you haven't heard about it, let me mention it, the Government Performance and Results Act. There are people I'm sure in this room that know it in much more detail than I do.
Fundamentally this is an effort with the administration's support by Congress to get more accountability for federal programs. If you can't answer such things as -- just very basic things in my mind -- are people getting more or less access to services? Are there more or fewer physicians providing services in the Medicaid program?
If you don't have a data system that allows you to answer that nationally, what do you think is an adequate assessment of how well the federal Medicaid program is doing? Because this is a program that is paid for -- the majority of funds come from the federal treasury.
I don't know what the current plans are. I have not seen a GPRA plan for Medicaid, but I would think that a Medicaid data system would be central to it, and that it clearly ought to be able to answer certain types of questions.
I think there needs to be some examinations about how to improve incentives. Well, several things -- there need to be some standards by which to measure the ability of data to do that. There need to be effective incentives -- and there are incentives now. The thing is -- and we have this problem within SAMHSA too with our programs -- the bigger the hammer, the less likely it is that anybody is going to wield it.
So actually I think the solution to incentives has to be explored in looking at smaller things that people aren't afraid to exercise, as opposed to big consequences like we're going to terminate your entire program if you don't do this. Those kinds of sanctions never get employed, but smaller ones can be effective and help prod people towards improving things.
Then I think the other major thing -- I'll mention one that I didn't touch on before, the very last one. It would be very nice if there could be an independent -- by independent I mean not only outside of HCFA, but outside of the department -- to say we're going to look at this every year, and we're going to separately assess if you are getting or worse, or the areas that you need to get better or worse in or whatever.
That does not now exist, and I think as a result, that helps let people off the hook, and helps I think fail to bring attention to certain issues that might benefit from some attention.
DR. IEZZONI: Thank you. As Hortensia said earlier, it gets more sobering as we hear from more speakers.
Do any of the other committee members have comments, questions for Jeff?
DR. HARDING: Could we get some information on GPRA?
MR. SCANLON: Yes, there is a standard fact sheet.
DR. HARDING: Yes, that would be helpful just to have that. I didn't realize there was such a thing.
DR. AMARO: Could we also get copies of the studies that you mentioned that SAMHSA has sponsored?
DR. BUCK: Well, most of them are still ongoing, but we can certainly provide you --
DR. AMARO: The completed ones, and then abstracts for the ones that have been funded, that are ongoing.
And maybe just in real bullet form, if there is an answer to this question, what has SAMHSA learned thus far about the impact of Medicaid managed care for substance abuse and mental health on access, quality, and health status, just in case there is an answer?
DR. BUCK: The short answer is very little at this point. For instance, one of our sources of information is participating in the 1115 evaluation that Leighton referred to in his presentation. So the same things that are hobbling -- whatever he says he does or doesn't have is what we have or don't have, because that's one of the projects that we are dependent upon to give us information.
DR. AMARO: Well, for example, I know that in the SAMHSA News I just read -- I was at the SAMHSA Women's Mental Health and Substance Abuse Conference in Phoenix last week, and I read in the SAMHSA News that two studies were just completed, one by Dr. Rosenbaum, right?
DR. BUCK: Yes.
DR. AMARO: I know that just in the abstract of that it says that one of the things that she found was that just the evidence in the contracts, by looking at contracts, inadequate coverage for mental health and substance abuse, a lack of stipulation of what services -- I'm just thinking off the top of my head.
I was wondering, things like that, are there sort of bullets that you can say you found so far? I couldn't remember if her study compared access, quality impacts in terms of the mental health and substance abuse for Medicaid managed care versus other managed care plans.
DR. BUCK: She looked at two things. She looked at Medicaid managed care contracts between states and managed care organizations, and then a second part of the study for us, and this was part of a larger effort that they were doing at G.W., looked at contracts between managed care organizations and providers, regardless of whether it was Medicaid or not.
All she looked at was simply a paper review of those contracts. So we still don't know what that actually meant. You can say, oh, this doesn't look good, but until you actually -- you don't know if that in fact resulted in poorer access, quality or whatever. It is quite a jump to say just on the basis of looking at a contract provision, that we now know how services have been affected for people.
Certainly there were some things that she found in that that people have been concerned about.
DR. AMARO: So you're not at the point where you can say bullets yet or anything?
DR. BUCK: No.
DR. AMARO: All right, thank you.
DR. BUCK: Sure.
DR. IEZZONI: Jeff, you were very clear in what you thought we should be doing, which I thank you for.
Why don't we take a 15 minute break now, and people can go get their mid afternoon caffeine fix. So we'll reconvene at maybe ten past three.
[Brief recess.]
DR. IEZZONI: I'm going to get us started again, because Dr. Harding will not be with us tomorrow. I want to make sure that we have as much time as we possibly can or have the will for, to be honest.
Sue, you have been here all day, very faithfully sitting in the audience.
MS. DODDS: It's very interesting.
DR. IEZZONI: Good. We're happy to have you here, and Jessica as well. So can we hear from you?
MS. DODDS: First of all, I wanted to just give you a little bit of my background. I have been working with Medicaid data for almost 20 years. I was the HCFA tape-to-tape project director for 11 years, and worked with Jeff. He was the project officer, and we worked with data from five states, trying to make it uniform.
In addition, under that contract we provided technical assistance to HCFA in converting their MSIS system into a research file format, the SMURF system. I'm continuing to help in a consulting role in that activity.
For the last approximately three years I have been working on one of the 1115 waiver five state evaluation projects. I actually have worked a little bit on the second one, but primarily under the first one. My role in that project has been to work with the states to collect encounter data to try to understand what they are doing; to try to understand why they were having problems; providing feedback to them on the data quality problems that we were seeing.
I guess that the first comment I really want to make about that experience is that I think that a start up is always very painful. It is going to take a long time, but I think it is worth it, because I think in the long run that the states are going to be able to generate useful encounter data. I know that this is contrary to what everybody else has been saying. It's an optimistic statement, but I feel that certainly Arizona's experience -- they are producing high quality data.
I have been working with the five states in the five state project, and Tennessee is actually, in spite of their early reputation for having all sorts of problems, is actually producing pretty good quality encounter data. I really think that I'm very excited about the requirement that all the states showing MSIS and submitted encounter data.
I think that if there are adequate resources allocated to this activity, that this data can, in the long run be useful, but I think it's going to take time and money in order to do this.
So what I have done is I basically have a series of eight recommendations that I was going to go over, and just kind of read through, and then I'm going to go over them one by one. My major recommendation is that HCFA take very strong leadership in this role, and they require or strongly recommend the specific standards to the states so that the resulting data will be uniform across states, and useful for research.
In my conversations with many of the people in the Medicaid programs, I have repeatedly heard that they would have really loved to have had more leadership. Each state and each plan is out there busily developing their own different system to handle encounter data, and that it would have been very useful if there had been a set of specific standards that they were following. Now I can't speak for every state or every state official, but I have certainly heard that comment on many occasions.
My first recommendation is that we require individual records for each service instead of for each encounter. In that sense, the data would look very much like the old fee-for-service data that we are used to looking at.
My second recommendation -- as I say, I'm going to read through these and I'll discuss each one in detail -- the second one is to redesign HCFA's MSIS system to collect the needed service and enrollment information, and have that be in essence, the Medicaid managed care minimum data set.
I also recommend that states be encouraged in processing their encounter data through a modified MSIS system instead of setting up a separate system.
I think it would be very helpful if the plans used the standard forms such as the HCFA 1500 and UB-92 instead of developing all their different data collection forms.
Another area I think it would be very helpful to develop a uniform set of codes to replace the state-specific service codes, and that would be based on a review of the existing state codes. I'll say quite a bit more about that in a minute.
I also think that the states should be allowed at least a one year grace period to get their systems up and running properly, and probably really more than that. I think that is really minimal. I think it is unrealistic to think that they can immediately start generating high quality data.
I think that HCFA should provide substantial technical assistance to the states during the start up.
I think that they should develop and provide to the states a set of uniform standards for data quality review of encounter data.
Okay, I'll go back to my first recommendation, and that is there are two ways that encounter data is currently being collected. One is the summary encounter record, where all the services that are current in an encounter are put onto one record. So that there may be multiple diagnosis and service code fields, so there may be three diagnoses and five service fields.
The service detail records are like the old fee-for-service records that we are used to, where there is one record for each specific service provided. That is, if you go to a physician and you have an x-ray and an immunization, and have your broken leg set, this would generate three claims.
Right now there is no requirement to the states on which of these two systems are being used, and so of course both of them are being used. Most of the states fortunately have chosen the line item service detail system, but there is at least one state, Rhode Island, which is generating encounter records.
The limitation of the encounter summary records is that there is no way to relate diagnosis codes to service codes if there are more than one service on a claim. Also, the MSIS system currently requires the line item service detail, and these summary claims can't be disaggregated, so that states that are using encounter summary records cannot participate or cannot submit their encounter data in MSIS.
Also, Medicare in the past, have the experience of trying to use these summary records, and it was abandoned because it was not considered flexible enough, and it was not a good tool for research. So I feel very strongly that the requirement should be that the state submit service detail records, and not encounter summary records.
DR. IEZZONI: Can I just interrupt at this point as a point of clarification? How does HCFA in the standardizations of the transactions fit with Sue's recommendation here? I haven't kept up on that as much as Marjorie and Jim have. How would that relate to this specific recommendation?
MS. GREENBERG: Are you recommending that even if they saw the same provider, one provider gave a person an immunization and had an x- ray done, and did some other service, that that create three separate service records?
MS. DODDS: Right. That's how it is done pretty much in the fee- for-service world in Medicaid right now.
MS. GREENBERG: In Medicaid perhaps, not in -- it's my understanding not in Medicare or other claims-based systems. If you go to a separate provider, and if you defined an encounter as going to the clinic, and you went to several different providers, but several services provided by one provider I think usually end up on the same transaction.
MR. SCANLON: Pretty much, but they have different diagnoses and services.
MS. DODDS: But they are in different line items, right? There are different segments for each one.
MS. GREENBERG: You are talking about the line --
MS. DODDS: Right, so that diagnosis X goes with service Z, and et cetera.
MS. GREENBERG: Oh, okay.
DR. IEZZONI: We have a clarification from the audience I think.
MS. HARMON: I'm Jane Harmon from NCHS. I'm not the expert on X12 claims, although I spoke with Donna Pickett about this very issue, and I believe that on the X12 transaction, the 837 claim encounter record there is a way. They procedure codes and diagnostic codes, but I believe there is a pointer that connects the two. This is what Donna mentioned to me. So that you can link them on the new X12.
MS. GREENBERG: I think maybe this is semantics that we're getting into. It does associate the diagnosis with the service.
MS. DODDS: The difficulty is that most of the states are using the HCFA 1500 and UB-92, and on the 1500 in the case of Rhode Island, they are then taking that data and aggregating it up to a summary level that cannot not be disaggregated. Part of the thing is that the states are left to do this any way they want. This is just two different ways of handling data, but there are more. I'm sure there are many more ways of doing it. I just feel that it is important to mandate a particular structure.
DR. IEZZONI: I think HIPAA requires everybody to report in the same way, but how they aggregate it could be different. So what you are suggesting is that the aggregation algorithms be similar across states?
MS. DODDS: Right, and that they be like fee-for-service. In the encounter data world there was the sort of concept of the encounter claim was somehow a summary of everything that happened in that encounter. So a lot of the detail gets lost or can get lost.
MS. GREENBERG: You weren't suggesting the equivalent of three separate 1500 claims?
MS. DODDS: No, but on the 1500, there are three different lines, and each one, the diagnosis and service codes, and the quantities and costs and so on.
MR. SCANLON: We better check that. I think the X12 can accommodate that.
DR. IEZZONI: Because the X12 standard is where it is heading, right?
MR. SCANLON: Yes.
MS. DODDS: But the basic point here is I think the states need to be told what system to use, because right now they are selecting different types of file structures, which makes it difficult to compare states with each other. It's just another barrier to making the states comparable.
My second recommendation is that the MSIS system should be redesigned so that it will collect the needed service and enrollment information, and that that replace what was called McData, the Medicaid managed care minimum data set.
The MSIS system I think is a good basic system that currently collects most of the data that are needed for most of the applications, research and program evaluation applications. I think that with fairly minor modifications it can be used also as a tool for collecting encounter data.
There are also a few data elements that are in the existing MSIS system that I think need to be modified to make the files more broadly useful. I have a couple of charts on the next few pages, and I'm not going to go through these charts. What they are, are a list of data elements, and why I think they needed to be added to the various files. I will say that for the claims files, that we do need to add the plan identification numbers, the plan type, and the type of coverage.
One issue that is going to come up with managed care is in my work with Tennessee, approximately 15-20 percent of the claims they are submitting are denied claims. A decision needs to be made whether or not denied claims are going to be included in the system. They are kind of interesting, because the service is actually provided to the person, and the managed care was paid a cap rate, but the provider was not paid for the service. So if they are included, then there needs to be some sort of pay denied indicator.
On the eligibility side I think there needs to be the monthly managed care organization plan identification number, the BHO plan identification number, the dental plan identification number. Some states have people enrolled in multiple managed care organizations at the same time for different services.
I'm also suggesting the addition of the Medicaid case number, which is very useful, particularly for linking together mothers and babies. It is one the limitations I think the MSIS data set.
I am really not going to go over my next set of recommended modified data elements for the existing MSIS system, but I have tried to explain in this chart in some detail why I think these things need to be added or changed, and what the importance of these data elements are.
Another one of my recommendations is that the states either be required or strongly encouraged to process their data through a modified MSIS system and not in a separate state system. In my working with the states, the two states that have done that, that I have worked with, it has worked very much more smoothly than the states that have tried to create a separate system.
Basically what they do is bring both encounter data and fee-for- service data into the system, run them through some common edits, and then split them. There are some edits, some of the cost edits which are not applicable for managed care, and some managed care edits that are not applicable for fee-for-service; just split that and bring it altogether.
It means I have one data system, and it means the submission of data to MSIS goes more smoothly. It just seems to have worked better for the states that have chosen to do that rather than set up a separate system.
Another recommendation is that the plans use the standards, the HCFA 1500 and UB-92. It seems sort of a shame, because that has pretty much gotten to be the standard everywhere, and now with managed care, a lot of the managed care organizations are going ahead and generating their own data collection forms, which seems sort of a shame to move in the direction which things are less uniform instead of more uniform.
My next recommendation is to create a set of codes to replace the state-specific service codes. As we all know the standard are HCPCS, but in Medicaid the states have up to thousands of state-specific codes. Usually they are for things like durable medical equipment, immunizations, transportation; obviously the things that are not in the HCPCS system.
Tennessee has hundreds and hundreds of codes for immunizations for example. It means that if you are looking at immunizations across states, you have to get each states' procedure formulary file. You have to map all those codes. You have to identify which ones are immunization and map them into the proper categories.
Even worse than that, some researchers don't realize that they have to pay attention to states' codes. So they will say, okay, I need the following HCPCS codes, and they will go cruising through and select those, and maybe miss -- in Tennessee's case, probably miss most of the immunizations that happened in that state.
The states continue to use their state-specific codes, and the reason they use them is because they have different reimbursement rates for different services. The services are things like -- one of my current favorites is pesticide spraying in the residence of an AIDS patient. It has its own code and its own reimbursement amount.
States continue to use those codes, but they could be also required to map them into a more generic list of codes. Either the states could be given the generic list that could be developed by HCFA, or HCFA could get the codes and do the map, and I think it would be better for the states to do this.
It seems like sort of a nit-picky little suggestion --
DR. IEZZONI: To relieve you, we spent a huge amount of time on this. A lot of the private companies develop their own procedure codes too. So it's a mess.
MS. GREENBERG: I think the good news is that under these new standards, the HIPAA standards, there is a plan to migrate all the local HCPCS codes to a set of uniform national codes.
MS. DODDS: There is even another level. There is the local HCPCS codes, the level three codes, but then there are also the true state- specific codes that are outside of the local codes, and there could be thousands of those. The thing is it's a big undertaking at the beginning, and it is going to be a big pain in the neck, but once it is done -- researchers either don't do it; do it different; or it is just a huge duplication of effort. I think this would be a wonderful time to implement this.
DR. IEZZONI: Our committee -- I don't know if we are on record, but a number of us are in actually recommending going even a step further, which is getting rid of the distinction between ICD-9-CM procedure codes, and CPT procedure codes, and thinking about whether we ought to think afresh about an entire new procedure classification system.
It is something that we are going to be dealing with under HIPAA, which is the Health Insurance Portability and Accountability Act. It is one of the topics I guess, Marjorie, that is slated for a couple of years from now.
Somebody from HCFA had a comment.
MR. PATTERSON: I just wanted say in accordance with Sue's recommendation, one of the plans and one of the recommendations under HIPAA for the procedure coding will be to eliminate all of the local codes, or at least give states the opportunity to apply for a local code that will be documented, so that researchers and folks will have the opportunity to know what all those are. So I think we are going to take care of that.
MS. DODDS: That's great. That's terrific.
DR. IEZZONI: It's going to be -- how can I phrase this? -- controversial. It's going to be a very long haul, if it ever happens; it may not. I think that there are a number of people who feel it is time to try.
MS. DODDS: When I was involved in looking at Medicaid data in the early eighties, and there were all these different service code systems around, and everyone said that there never could be a uniform system, you could never get the providers to do it, but you know, it has for the most part happened. If there is enough will and resources, I think it can be done, and it is definitely worthwhile.
Another recommendation is to allow the states at least a one year grace period to get their systems up and running properly. There are so many layers of starting up these systems. The managed care organizations have to get their systems up and running. The states have to come up with the file specifications, and lay outs, and methods of transmission, and time requirements, and data quality reviews.
Then once this all starts to happen, then the states reject data, usually over and over. The plan sends in the data. The data is run through the edits, decides the edits need to be different, but decides plan data isn't right either, rejects it back and forth.
Then when it is submitted to HCFA, I'm sure that there is going to be another round of this kind of thing. I just think that there is no way to get around this situation. That as a state goes to managed care data, that they should have at least this one year grace period. They should probably be given a one year grace period, understanding it is going to take longer than that, but I think that is very important.
I also think it is important to provide substantial technical assistance to the states and plans during the start up. This will help catch problems early on, and I think make the whole process go a lot smoother. Also, if the states were provided with very uniform standards about how to count services, what is included, how to classify different things, I think that would be enormously helpful.
My final recommendation is that HCFA develop and provide to the states, a uniform set of standards for data quality review. I know that right now I've gone out to visit plans in states, and the plans have their own data quality review. Even under the MCOs, the BHOs have their own data quality review, and the states have theirs.
Not only is it a huge duplication of effort, everybody is doing it differently, and measuring data in different types of ways. I think that it is important to look at things like distribution of services, comparing differences between plans, and using some validation techniques.
In the case of Tennessee, we just finished or we are actually sort of in the midst of doing validation on their 1995 encounter data, and it actually looks pretty good. We looked at the services that are provided to people with chronic conditions before the demonstration, and the first year of the demonstration, and then two years before versus the year before the start up of the demonstration. They are surprisingly very comparable across many of the plans; not all of the plans, but most of the plans have done a very good job.
I think techniques like this should be standardized and provided to the states, instead of having each state and each plan develop their own set of data validations.
Then I was going to talk just for a moment about the back page, which is a little chart that I did with five of the states. Four of them from the first five states' evaluation, and one of them, Minnesota, is from the second. These just happen to be the states I have worked with.
In terms of claim structure, all of the states except for Rhode Island have one claim per service. Rhode Island, as we talked before, has one claim per encounter. The number of plans with the states tends to fluctuate, but right now in those five states it varies between five and ten plans per state.
Two of the states are using their MSIS, Minnesota and Oklahoma, to process their encounter data, and the other three states have a separate system.
In only one state are the plans submitting adjusted claims. You would think that adjustments are not so important in encounter data, but what the plans are doing are adjusting their payments to their providers, and so it is very important that you resolve these adjustment claims to get down to one final bill. It is not the dollar amounts, but you don't want a service in there three times; the original, the void, and the resubmission. You have to get to the original claim.
The next line was the date that the demonstration began in the various states, and the following line is an estimate of when the first year of useable data that we will be getting from these states. You can see that it varies from a year to three years.
I guess the only state that I can really talk about is Tennessee, because that's the only state we have really gotten useable encounter data from. It was 40 months from the start of the demonstration until we had one year's worth of useable data. Now they are in a production mode, and now it will be very timely. For the start up time, I don't think that was an unreasonable amount of time.
The really just basically concludes my remarks, if anybody has any questions.
DR. IEZZONI: It was very thoughtful and specific. We ought to make sure that Barbara Starfield and John Lumpkin get this handout, because it is very specific; very helpful.
Any of the committee members have comments or questions Sue?
MS. WARD: I have a comment about standardizing the data for quality evaluation. It seems like a fairly ominous task. Do you have a sense of whether that is something that has to be started? How long that would take? Do you have a sense that there are some standards sets that people are ready to do, and that could be in place?
MS. DODDS: Right. For the five state evaluation, that has been one of our tasks, to do that. So we have come up with a way of preliminary review of the data and some validation techniques. I think that in order to do this properly, it is a fairly considerable undertaking. I think it would be six months or so.
What we have done, I think is pretty good, but I think that for something to be given out to the states, it should be a very complete package of very specific standards of the kinds of things you look at.
MS. WARD: When you use the word "standards" and validation, are you saying you have six quality performance measures that you have identified, and then what data needs to be collected to say what that measurement is, or are you talking about something else?
MS. DODDS: What I'm talking about is first of all we looked at things like the distributions of type of service. Were there any mental health claims from a plan, and what percentage was it? It should be within a certain range to be acceptable. So we looked at the distribution of types of service.
Then we also looked at the pre-demonstration and post- demonstration use of services, but with people with selected chronic conditions. We're also looking at how many physician visits people have per year, and how many hospital visits, and looking at that across plans.
It is really interesting when you look at it across plans. You can really spot kind of the outliers, and certainly when you are doing the cross states and on a larger range. So standards could easily be developed.
MS. WARD: Can you share with us what those standards are that you have then, that you are using for those five states?
MS. DODDS: You mean right now?
MS. GREENBERG: These are data quality standards.
MS. DODDS: Right, data quality standards.
MS. GREENBERG: Not quality measurement.
MS. WARD: You're talking about the fact that you are measuring hospital to hospital, that's a performance measurement.
MS. DODDS: Right.
MS. WARD: You are making some assumptions that there is a good and bad, and you said an outlier. That means that you have a sense that someone is outside of something else. So I'm trying to get a sense of how you came to those assumptions, and what you are actually measuring.
MS. DODDS: To give you a really kind of -- besides the chronic illness -- to give you kind of a gross measure, for example in Tennessee in 1995, there were 12 plans. So I just simply counted the number of services that people had per person year of enrollment across the 12 plans. In Medicaid fee-for- service it is usually somewhere around 30. For most of the plans it was like between 25 and 30. There was one plan that there were nine services per person year of enrollment.
Clearly a state could look at that and know -- that's, as I say, a very gross measure. They could look at that and know there was a problem with that plan. Their claims can all pass the edits and all this, and there is a substantial problem. Either they are underserving, underreporting, or treating a very healthy population, or having great preventive care and no one is getting sick or whatever. It is an indication.
You would expect that there would be a certain percentage of lab claims, a certain percentage of mental health claims, that kind of thing, plus many more sophisticated kind of measures.
MS. WARD: Any of that, that you have quantified now and are using for those states would be very useful.
MS. DODDS: Okay.
DR. IEZZONI: Are there any issues that you think more visibility that our committee could help get that visibility around data systems for Medicaid managed care evaluation?
MS. DODDS: Well, I think that my main concern is the limitation of resources to put these systems in place. I think that with enough commitment and planning and enough resources, a good system could be developed.
DR. IEZZONI: Any other comments? Anybody from the audience have any comments, questions?
Thank you, Sue. That was really interesting.
Jessica.
DR. BANTHIN: My name is Jessica Banthin. I am from the Agency for Health Care Policy and Research. I believe I was asked to come speak to you today, because I have been the first researcher on our research staff to use the data from the 1996 Medical Expenditure Panel Survey to look at the Medicaid community population.
I am economist. I have been a researcher at AHCPR for six years. I haven't done that much research on Medicaid, although before I got my graduate degree I worked at Massachusetts Medicaid as a research assistant on a federally funded demonstration to enroll Medicaid patients in managed care health plans sponsored by community health centers. So I have experience using claims data.
What I am going to do today is present an abbreviate slide show, and describe our survey, which sort of complements some of the encounter data issues that you have been talks. I will talk about the survey and what we can do with the Medicaid population and other populations. After that I'm going to share some early results of research that I have done on the Medicaid community population.
My research is focused on eligibility and enrollment, looking at take up rates among the Medicaid community population, and the issue of crowd out, which is a controversial issue currently, which looks at whether Medicaid enrollees have turned down offers of private health insurance coverage in order to enroll in Medicaid given the recent expansions to the working population.
The Medical Expenditure Panel Survey or MEPS is the third survey. We had a history of doing similar surveys in 1977 and 1987, but we keep changing our name. We are now different though. We subsample from the National Health Interview Survey, so we sample people who were actually interviewed in NHIS, and we got back to them. We locate them. We back to them. We include them in our survey and follow them for two additional years.
We are now in continuous data collection. So we follow a sample of families for two years. Each year we add families to our survey; follow them for two years; drop them; add more. It is an overlapping sample design.
There are four main components to the survey. I'm going to focus on the household component, but I will also talk about the medical provider component, our insurance component, and I will not talk about the nursing home component today.
MEPS provides national estimates of health care utilization and expenditures for a nationally representative sample, U.S. population. It is the non-institutionalized, civilian population. We provide distributional estimates, that is we can look at people across the spectrum of healthy and sick, high users, low users, uninsured, privately insured at all income ranges.
We look at individuals. We link them to families. We link them to providers and to employers and to health insurance plans.
Since we now follow people for two years, we can look at their experience over a two year panel. We can also provide annual estimates year after year to look at changes in the nation in health insurance coverage and employment. As I said, we link individuals to their families, to their employers, to their medical providers, and to their health plans.
We oversample based on what NHIS is doing and some of our own oversampling. In 1996, we looked at minorities, especially Hispanics. In 1997, we are oversampling low income children with limitations, disabled children and adults, high expenditure cases, and elderly; then in 1998, minorities again.
These are the sample sizes. Our samples are smaller than they have been in 1987 and 1977 surveys, but they are large enough to provide most analyses that we want to do.
The interview lasts on average, two hours. It can go much longer than that if someone is really sick and uses a lot of health care. It is a very detailed interview that starts with looking at all people in the family. We also ask questions about health conditions -- that is not shown here -- health conditions, health status. Then we ask about all types of health use. Then we ask the family how much they cost, how much they spent. As I'll talk about, we confirm these charges in our medical provider survey.
We ask them about their health insurance status. This is where we identify whether they are covered by Medicaid or private insurance or Medicare. People underreport Medicaid coverage a little bit. It is getting harder for us to ask someone if they are publicly or privately insured, because of expansions like TennCare, people don't always know how they are eligible for the program, or they may not wish to reveal that they are eligible because of low income. It's an expansion program that covers a whole lot of different people in one program, so the distinction between public and private is getting tricky.
We ask people about their employment status, since employment provides so much private health insurance, and extensive questions about health status.
When we ask about utilization, we go through all types of health care -- hospital, office-based physician, other medical providers, dental services, home health care, prescribed medicines.
If they have a medical event, we ask them what type of provider; how long they spent with the provider; what the reason for the visit was; what type of care they received; what health conditions the visit was related to; and we ask them about charges and payments. At this stage, and again when we interview providers, we distinguish between the charge and the actual payment in order to get at discounts and things like that, which are sometimes hard to get at.
We ask questions about the employment for main job; secondary jobs; additional jobs; wages; industry; hours worked.
We spend about 20 minutes on average talking about all sources of health insurance coverage. It is lengthy, because we ask about -- not everybody has several sources of health insurance coverage; most people don't. Our questions cover all of the public programs, and then all different ways of getting private insurance through employers, directly through self-purchase and things like that.
If you are offered health insurance through you job, and even if you do not get health insurance through your job, we ask people if they were offered health insurance. If they are offered, we ask them if they have a choice of plans.
We have supplements that we ask periodically. We have access to care that is a series of questions on ability to get care, problems or delays in getting care, and satisfaction with providers, office hours, the usual questions.
We also ask satisfaction with health plan and with the providers. These questions are asked of all survey respondents, so we have a similar set of questions that have been asked of the Medicaid enrollees and private enrollees and Medicare and CHAMPUS and everyone else. So you could make comparisons between HMO enrollees who are sponsored by Medicaid and HMO enrollees who are sponsored by private or commercial insurance plans.
We have supplements about long-term care. There is a self- administered questionnaire on health status. Then at the end of each year we ask about income and assets.
We have a medical provider component that allows us to confirm -- it surveys a variety of providers. It is directly linked to household respondents. We get permission forms to contact the providers. We use it to confirm the charges that were reported, and then we impute for people who do not respond from those who have responded in order to get complete response data on charges and payments.
The people whom we ask permission to contact their providers include all Medicaid households, because we don't believe Medicaid respondents can accurate report the cost of visits. So for the Medicaid population we just go directly to their physicians and hospitals and other providers.
We also include all hospitals and associated physicians. We include half of office space physicians, all home health agencies, because we have targeted that recently, and all pharmacies. I omitted the slide that includes the numbers, but I can tell you this is over 3,000 hospitals that we survey, over 8,400 office-based physicians, 500 home health agencies, and over 8,500 separately billing physicians.
DR. IEZZONI: Jessica, can I just interrupt? It's interesting that you use the provider survey as the gold standard corroboration of what the patient reports basically. How sure are you about the 8,400 physician offices and the home health care agencies? How easy is it for them, frankly, to get this information to be able to respond to your survey?
DR. BANTHIN: It is tricky. We go by what household reports for the number of visits. We don't change the number of visits that anyone has reported to us. So if a person says they went to the doctor once in January, and we go to that doctor's office, and we says they came in twice, we do not change that. We go by household reporting.
If the person says they believe -- let's say they are in a fee-for- service plan and it costs their plan $50. Because they hadn't met their deductible, they had to pay $50. We go to the physician and he says the bill was $65, we use the physician's estimate of the cost of that visit.
I think I have another slide here. This is the data we collect. The medical provider component is designed so that we do not have to ask the physicians themselves. It is designed to be answered by medical record clerks.
DR. IEZZONI: Knowing what I know of doctors' offices, I think you are very brave to go out and do this.
DR. BANTHIN: We would have response rate problems. But we get the dates of the visits, the diagnoses and procedure codes, and then the charges and payments. Then reasons for any difference. This is if any discounts applied.
Then of course the next question is what do we do about HMO enrollees who may not have all this information? We are going to do our best to get that type information from managed care organizations. We have plans for imputing data from HMOs to other HMOs, but not from fee-for-service to HMO. Those plans are not yet -- they are in progress. We have spent a lot of time thinking about this, and we'll probably spend more time thinking about it before we complete this imputation and hot deck process. These data, as I showed at the end, will be ready next year.
Our insurance component is really a survey of mainly employers who provide insurance to their employees, and also some health insurance companies where people have purchased insurance directly.
We get information about the employer, about whether they offer health insurance, and whether the respondent whom we have gotten permission from to contact their employer. It is harder to get permission from people to contact their employers than it is to get permission from people to contact their doctors. It's something about income.
But we find out whether that particular person was eligible for health insurance; whether they were enrolled; what choice of plans they had, if any; and then what the premiums were. It is important to get premium information. Employees don't always know what the full cost of the health insurance plan. They know what share they pay or what is deducted from their paychecks, but they don't know the full cost, so we get that from the employer.
We get some limited information about benefits in the booklet, and some information about the employer. We collect health policy booklets from all of our survey respondents. We pay them $15 to get a copy of their booklet. We extract the benefits that are in that booklet in a separate survey component, and then link it back to individuals and to their employers who provide this insurance.
We are committed to releasing this data as quickly as possible under GPRA. We have strict standards. We have already released two public use files. Round One, which is largely health insurance data, demographics, and employment status is out. We are releasing access and managed care status in a few weeks. We are editing the data a lot faster. It's a little less clean, but it is clean enough to use. We are releasing flat files to researchers.
Our schedule for next year, in May we will release the full year utilization tape. So this is 1996 utilization. Then in September we will release expenditures, sources of payment, and a lot more stuff.
DR. IEZZONI: Well, that sounds like an exciting survey.
DR. BANTHIN: If you have questions about the survey itself, I can talk about that. I can also talk briefly about some of the things I have found looking at the data that we have.
DR. IEZZONI: Quickly, because I think there seems to be a bit of fading around the table.
DR. BANTHIN: I just included some slides that talk about the Medicaid community population, that show facts that you probably know. I compared our 1987 data to our 1996 data. Over a quarter of children 0-5 are covered by Medicaid.
There have been increases in all age and sex groups. You can see though that some of the largest increases are actually among men 65 and over. The largest increase is among young children, the elderly have increased because of dual eligibility. The expansions into the low income working poor have actually increased the number of working men who are on Medicaid.
If you flip to the pie chart, this has some implications for the composition of the Medicaid community population. Children are a larger percent than they were in 1987. As you would expect working age adults are a smaller percent; they are 38 percent now. They used to be almost 42 percent. The elderly has grown as a percent, so that has implications for who you can enroll in managed care.
I'll skip the race charts, although you can look at those at your leisure.
The next three slides we put together for research on children. If you look at the one headed, "Are All Children Who are Eligible for Medicaid Enrolled in the Program?" that shows we have used our demographic and income data to simulate eligibility for Medicaid. That is, predict a person's eligibility for Medicaid. Then we look and see whether they are actually covered by Medicaid. So we can look at who is more likely to sign up, and who is likely to go uninsured, and who is likely to take private insurance.
I ran this table for this presentation. It just shows in a national estimate of Medicaid community population, who is enrolled in managed care programs. This confirms what I think most people know. It is younger. It is the children and the young adults who are more likely to be enrolled in a Medicaid managed care program.
There are no differences by gender. There are some differences by race and ethnicity. I don't have the standard errors for that, so I'm not sure whether it is statistically significant. There are differences by region that confirm what you would expect.
On the second page I look at health status. That shows that the healthier Medicaid enrollees are more likely to be enrolled in managed care than the disabled and sick, as we probably expected.
On the last page I just talk about some of the research that we can do with our survey. It is a different kind of research than one could do with state level encounter data. MEPS cannot make state level estimates, however, internally we have state codes. For external researchers, in the past we have done a match for them to match to state level information, because we cannot release the state identifiers because of confidentiality.
One of the things on the last page, in addition to some of this crowd out and take up rate stuff, in terms of managed care, the last two bullets talk about research we can compare enrollees and non-enrollees in a nationwide sample, and look at their use of services and preventive care.
The thing about MEPS is that we have collected the data the same way for everyone. So although it is not like encounter level data, if someone doesn't like the way we have imputed charges to managed care visits, at least it is consistent across everybody in our sample, including privately insured people.
We could compare Medicaid enrollees in terms of their satisfaction with their health plan to privately insured HMO enrollees and things like that.
DR. IEZZONI: Well, we might ask you to do it. That's exciting.
Anybody on the committee have any questions? George?
MR. VAN AMBURG: How big is your sample in the panel, and what is the drop out rate from the panel over the two year period?
DR. BANTHIN: We haven't even finished our second year. We're in the field, so I don't know what the drop out rate is. The response rates are a little lower than they were in 1987, because we have subsampled from NHIS. So they have been through the National Health Interview Survey, and then we go find them again and knock on their door and ask them to participate again. I don't have those numbers off the top of my head. Our statistician does. It's down in the eighties.
DR. IEZZONI: That's an interesting consequence of survey integration, isn't it?
DR. BANTHIN: Yes, and unexpected.
DR. IEZZONI: These people don't get paid anything, do they?
DR. BANTHIN: No, we are prevented from paying them, which is really unfortunate. It's a rule that OMB applies. They pay them for their health policy booklet, $15. For some people that helps insure cooperation, but it is difficult.
DR. IEZZONI: Any other comments? Great. I think that we might be very interested in hearing more or talking with you about doing some analysis specifically to inform what we are working on, so I'm sure we will be back in touch.
Thank you, Jessica, and thank you to the other panelists.
I forgot to ask, are there any questions from the audience for anybody on the panel?
PARTICIPANT: I'm from the Asian-Pacific Islander Health Forum. I was looking at the MEPS, and I was wondering whether there could be any data analysis conducted on the Asian-Pacific Islander portion of the sample? Is there Asian-Pacific Islander data that has been collected for MEPS?
DR. BANTHIN: We do identify Asians. It is very small though.
PARTICIPANT: Do you have an idea as to what the numbers are?
DR. BANTHIN: I don't know that off the top of my head. In 1987, we had a separate survey of Native Americans and Alaskans that was partly funded by the Indian Health Service, but we did not do that this time.
DR. KU: Can I just go over the imputation process for expenditures again? Let's say you are a Medicaid recipient --
DR. IEZZONI: Leighton, sorry, I'm concerned about Richard having to leave. Maybe imputation procedures, maybe you could talk about privately. Would that be okay? I'm sorry, I hate to express a disinterest in imputation procedures, but I'm just concerned about having the committee be able at least have some time to talk among each other before they have to depart.
DR. BREWER: I just had a question for Sue. In your work with different states on encounter data, I'm curious, are there any states who have actually tried to do some evaluations of their encounter data system by doing linkage with other public health data systems. Hospital discharge data systems would be an obvious example, to look at issues of completeness of reporting into those systems.
MS. DODDS: I know some of the states have done medical record review to look at data quality. I'm not aware of any states that have done that linkage, but they have.
MR. SCANLON: Jessica, beginning the day after tomorrow, states begin getting their share of $24 billion for the next five years I guess it is to expand children's health insurance. Does MEPS have the capability, at least on a national level, to be able to see what the impact might be?
DR. BANTHIN: Yes, I feel we are well situated, because we have collected data in 1996 and 1997, before anything has really happened. We have a good set of data to look at the new expansions, and then watch them unfold. So when we ask people how they are covered, we have a series of questions and them a mop up question to make sure we have gotten people covered by Medicaid. We also refer to other public insurance programs in case they are confused about the name. This has become a source of confusion now. We use the state-specific name for Medicaid when appropriate.
DR. IEZZONI: Great. At this point we would like to thank our presenters very much for an informative day.
The committee may or may not deliberate for very long, because I can see people looking tired. We have a lot to digest today. Basically we will talk for a few minutes at least, and then for those of you in the audience who are going to leave now, we will be reconvening tomorrow morning with just four of us.
It will probably be very much a conversation among the four of us in a public forum, and hopefully we will be able to have the tables arranged so the four of us are not all sitting in a line, but we're actually looking at each other. So those of you who arrange the room, if you could just make sure that that happens, so we're kind of closer together.
Hopefully, by the end of tomorrow we have figured out a plan for what our subcommittee will be doing for Medicaid managed care for the next nine or so months.
Basically, there are two handouts. One is an article from I guess Health Affairs that is basically just an overview of changes in health plans serving Medicaid that kind of can serve as a background for us just to kind of again, get a summary kind of lay of the land.
Then the other handout is kind of sobering. The other handout is a transcription of what we talked about in July. Do you remember that in July we Gracie White and the four other staff people from the National Committee trying to write on paper, and we tried to paste it up on the board around the room? Well, this is kind of what we came up with in terms of the questions that we were addressing.
We kind of took a very broad approach, and tomorrow what we really need to do is narrow it down. If folks remember what we did back in July is we just kind of brainstormed. We sat around and we said, okay, what do we want to know about Medicaid managed care? What populations do we want to know it about? What information do we need to be able to know these things about the populations?
We came up with a lot of different questions that we wanted to have answered, a lot of populations that we were interested in asking the questions about, and the data stuff we actually engaged in a variety of debates. I think George you were on the anti-encounter side, and Vince might have been on the pro-encounter side, or something.
So this just serves as a back drop. I think today has been very productive in terms of hearing more distillation from the experienced research community around what they think is important for us to be highlighting, since we are in kind of a convener, highlighter type of mode.
So I think that what I would like folks to do is just think about this maybe over night, and think about today. Then we'll start tomorrow morning at nine o'clock, and really try to have specific discussion.
Carolyn has listed for us on the second page of the agenda what our action items basically need to be. We need to decide what our product is going to be for this next nine months. We need to decide what contractors we want to contract with, and Jim has reminded me that virtually all of the people that we have seen today from the private contractor world are available to us in a contractual relationship to be able to do some work for us.
Then to kind of talk about our site visit strategy. I was interested that two separate people mentioned Massachusetts and Arizona as places that are doing a good job collecting data. Those were the two places that we had thought about doing our site visits, but that could still be very much open for debate.
So do committee members or Marjorie, Jim, Ron, Carolyn, do you have any comments, thoughts?
DR. MANDERSCHEID: I guess one thing I would add based on what I have heard here is that we need to get more focus on actual quality work and outcome measures to attach to enrollment and encounter data. No one mentioned any work in that area.
DR. IEZZONI: In the morning there was a lot more on quality I think.
DR. MANDERSCHEID: But actually getting things up and running in a reasonable period of time here. There is just an awful lot going on in each of these fields that really wasn't reflected in some of the presentations. If you focus on the research community rather than the practice community you miss all of these things that are happening. If you had the states here for example talking, you would get a totally different view than you get from the research community.
DR. IEZZONI: Well, that's a good point. You know what we need to first do though, Ron, is frankly decide that what we want to focus on is quality. I was going to propose tomorrow morning that we want to focus on two things, access and quality, which are part of the same point if you will.
If you go back to what we said in July, we were talking about cost, about provider issues. It's just a full range of things, and I just don't believe we're going to able to bite off everything. So people can think about that straw person that I've just floated, that access and quality is what we want to focus our interest in data systems around.
That's a good point. We need to hear from states, people that are out there doing it.
Richard, maybe because you are going to be gone tomorrow, if you have any final comments for us.
DR. HARDING: There are still a couple of just kind of basic things that I think about. One was Medicaid managed care will improve what, and how can we determine that? Is it being set up to improve the quality of care of individuals? Is it purely a cost containment issue? I would like to think it is both, but we heard from people that it is mostly the cost, because there is no way to judge the quality at this point, and that we need help in that some way to be able to say if we're going to keep passing out billions of dollars to organizations, that there ought to be some way to look at that.
I think I was impressed with Rosenbaum when she was saying that people in Medicaid don't have a point of service. They are stuck with what they have, and how important it is for us to make sure that they --
DR. IEZZONI: That's the access issue.
DR. HARDING: Yes, the access. They are stuck there.
The issue of performance data, and the agreements that are made. There are a lot of things that I'm sure you all will be talking about.
I'm from South Carolina, and I'm a states' right person and all of that stuff, but there have to be standards on this issue, national standards, or it is just going to be chaos. I think there are lots of companies that would really prefer that it not be standard, so they can give it as inexpensively as possible.
So I would be all for us working towards that on the committee, and helping people who know a lot more than we do about it, how to do that.
DR. IEZZONI: George, Hortensia comments? Not right now. I detect fatigue. Can I move that we adjourn?
Thanks in the audience for sticking with us.
[Whereupon the meeting was recessed at 4:25 p.m., to reconvene the following day, Tuesday, September 30, 1997, at 9:00 a.m.]