Hubert H. Humphrey Building
Room 705A
200 Independence Avenue, SW
Washington, DC 20201
Opening Remarks and Introductions - Dr. Iezzoni
Overview of Post-Acute Care - Dr. Kane
Post-Acute Care: Data, Findings, and Sources - Dr. Liu
Outcomes Research and Overview of Quality of Care Issues - Dr. Kramer, Dr. Webb
Panel Presentations on Pediatric Post-Acute Care and Quality Issues
DR. IEZZONI: We'd like to get started, if people could take their seats.
We are going to start, because we want to have ample time to hear from our guests, who have flown in a great distance and have prepared thoughtful presentations for us for today. So if everybody could have their seats, I'd like to get started.
We are the Subcommittee on Populations, from the National Committee on Vital and Health Statistics. We are here to be educated today. We are here to learn from you all about what we are calling for the moment post-acute care, but we are hoping to hear from you about whether there is a better thing to call it. Barbara Starfield isn't here yet. She will be the one that will react to those words in the most obvious physical, palpable way. But the rest of us have some concerns about exactly what we are going to call this initiative.
What I'd like to do at the outset is just go around the room and have everybody introduce him and herself, including the audience, so we all know who is in the audience. Then we will talk about what we're going to do for the rest of the day.
I'm Liza Iezzoni. I'm in the Division of General Medicine and Primary Care at Beth Israel Deaconess Medical Center in Boston. Kathy?
DR. COLTIN: I'm Kathy Coltin. I'm with Harvard Pilgrim Health Care, also in Boston.
DR. TAKEUCHI: I'm David Takeuchi, in sociology at Indiana University and a member of this subcommittee.
DR. NEWACHECK: I'm Paul Newacheck at the University of California, and also a member of the committee.
DR. BROWN: I'm David Brown. I am representing Ron Manderscheid from the Center for Mental Health Services.
DR. IEZZONI: Good, thank you. I was going to say, you're not Dr. Manderscheid. Good, welcome.
DR. LIU: Korbin Liu with the Urban Institute.
DR. KANE: I'm Bob Kane from the University of Minnesota.
DR. KRAMER: Andy Kramer, University of Colorado.
DR. SUMME: I'm Jim Summe with AHCPR, and staff to the subcommittee.
DR. HANDLER: Aaron Handler. I'm the chief of the Demographics/Statistics Branch, Indian Health Service in Rockville, and I'm a staff person to the committee.
DR. GREENBERG: I'm Marjorie Greenberg from the National Center for Health Statistics, CDC, and executive secretary to the committee.
DR. FRIEDMAN: Dan Friedman with the Massachusetts Department of Public Health and a member of the committee.
DR. MOR: I'm Vince Mor from Community Health at Brown University. I am a member of the committee.
DR. RIMES: Carolyn Rimes, HCFA. I get to staff this on occasion.
DR. IEZZONI: And we're happy that you do. Before we go around the outside of the room, I wonder, Marjorie, could you introduce Patrice to us, and Jackie Adler as well?
DR. GREENBERG: Jackie is a longtime friend and staff to the committee, and we have very pleased to be joined by Patrice Upshur from our office at NCHS. She is assuming the duties of Barbara Hechsler, who has been the staff contact with the NCVHS team to this subcommittee. But Barbara will continue to be working with us, because she will be working with the contractor, our support contractor.
DR. IEZZONI: Good, and we will plan to write a thank you note to Barbara for her service for the committee. She has been great.
DR. GREENBERG: That would be nice.
DR. IEZZONI: And welcome to Patrice. Henry, do you want to introduce yourself?
DR. CRAKAR: Henry Crakar, Planning and Evaluation for HHS.
DR. GAGE: Barbara Gage, the Urban Institute.
MS. CORNELIUS: Betty Cornelius, Medicare beneficiary.
MR. TART: Gene Tart, Quality Management with Beverly Enterprises.
(The remainder of the introductions were performed off mike.)
DR. IEZZONI: Good, and Dale Hitchcock is just coming in the room.
DR. HITCHCOCK: I am Dale Hitchcock, I'm with the subcommittee
DR. IEZZONI: Thank you. Well, thank you again, everybody, for coming to meet with us today. Again, we are hoping to learn from the panel this morning and this afternoon about issues that we might want to consider as we think about care outside of the standard acute care hospitals, standard ambulatory care model.
What I would like to do is, we have a schedule that we would like to try to follow, because we hope to hear from everybody. But I would like to also give the committee a sense that they can interrupt and interact, and also members of the audience, because I think this is our time to try to understand. So if people aren't understanding concepts or want to pursue certain threads that the speakers bring up, I think we should try to have the opportunity to have some interaction.
So maybe I could ask the speakers to focus initially on what they might want to tell us as a core, and then let me know when we can begin to think about asking questions, and interacting.
A whole bunch of folks just arrived. Maybe it was the security system at the entrance of this building which just let through a flood of people.
Barbara, do you want to just introduce yourself quickly?
DR. STARFIELD: Hello, I'm Barbara Starfield from Johns Hopkins University, member of the committee.
DR. IEZZONI: Okay, and we'll let the other folks just sit down.
DR. KILGORES: I'm on the agenda, but I don't see my name. I'm Carl Kilgores.
DR. IEZZONI: Welcome. Dr. Robert Kane from the University of Minnesota is going to start us off, talking about an overview. Bob, this is being taped for transcription, so we're going to need to make sure that you are miked at all times. We are also being broadcast on the Internet live, so that is another reason why -- we don't know if anybody is listening to us, Bob, don't worry. We might just be talking to cyberspace.
DR. KANE: They might be tuning into the impeachment trials and get this by mistake.
DR. IEZZONI: Well, they may be.
DR. KANE: This will be more interesting.
DR. IEZZONI: So the fact that we're on the Internet -- just while Bob is getting set up -- means that if anybody has any comment or wants to ask questions, you have to speak into the microphone so it can be picked up.
DR. KANE: What we thought we would do would be, I will present some concepts on post-acute care and Korbin will present some facts, and you can interrupt any time, in terms of this discussion. This is really designed to offer you some things to think about as you deliberate about this area.
I think the place to start is to recognize that nobody invented post-acute care. It sort of emerged without a great deal of planning, and without a lot of sense of what it was. For a long time, it lay dormant because nobody paid a great deal of attention to it, because hospitals could basically do whatever they wanted to do. Post-acute care rose to prominence after the imposition of the change in the payment system for hospitals that moved us to a prospective payment system, and suddenly it became very profitable and fashionable to shorten hospital length of stay. So a great deal of care that used to be given in hospitals was now given in other kinds of institutions which had existed before, but had not occupied the center of the health policy stage.
For practical reasons, post-acute care really can be thought of as post-hospital care. The problem with that definition is that things change. So what began as a program that was defined, as we tend to do in the United States, defining things either by who pays for them or where they are delivered, we now move to a situation where there were a set of services. People began to question whether indeed these services could be given without the intervening hospitalization.
Indeed, in some cases for certain kinds of post-acute care, you don't have to be in the hospital to get into them. As we move into new configurations of services under managed care programs, for example, we are now developing things that are called direct admission post-acute care, which is an interesting oxymoron just to confuse you.
So in a sense, I think at some point we are going to be doomed to make an arbitrary definition of what post-acute care is. Some people like to call it subacute care, but there are also reasons why that is not a good phrase. But either we are going to have to define it on the basis of something that is triggered by something else, like a hospitalization or some set of services that we can identify closely enough to define that they really are separate services. But as I think will become apparent today, many of these services really look the same, although they are paid for by different people. When they are paid for by different people, they are called different things. So this is a very murky area, and that doesn't help you much.
So let me start out then with where we began. For this presentation, I'd like to argue that what we are really talking about is post-hospital care, in that it represents care that is triggered by a hospitalization.
The problem with this whole idea is that at the moment, the post-hospital discharge planning phenomenon is a travesty in this country. There is no regularity or organization to it. So the danger is that you have people who are being discharged from hospitals, receiving no care, who look very much like the people discharged from hospitals receiving care. The people who are discharged from hospitals receiving one kind of care look a lot like the people who are discharged from hospitals receiving another kind of care.
So we have a problem. We are dealing with a major decision which launches this whole type of care that is being driven by a group of people who for no fault of their own are terribly uninformed about what is the right decision to make, so it is not a systematic process. Hence, the result is that unsystematic set of behaviors which then become very hard to model or describe.
So the probability of where you go may vary with your diagnosis, but it may also vary with a whole series of other characteristics.
Basically, post-acute care consists of three major vendor types. One is home health care, which for the purposes of post-acute care is reasonably defined as a Medicare certified vendor. We have nursing homes, who represent several different kinds of entities now, who may cater to the Medicare market, in which case they would presumably be engaging in post-acute care, but may also not be Medicare vendors and in fact provide very similar kinds of services under Medicaid, for example, or under a private system. Then we have rehabilitation, which has historically been given as an in-patient service, but increasingly people are recognizing can be delivered on an out-patient service.
All three of these to varying degrees have been greatly influenced in the way they deliver their services by Medicare regulations, which shape what they do, not necessarily for the better. In some cases, have made them extremely expensive services that could potentially be delivered in other venues or in other ways at lower cost.
The hidden hero of post-acute care is the family. An awful lot of post-acute care, particularly that which takes place in the home, is heavily dependent on informal care of one form or another.
Now, if we have difficulty defining what the beginning of post-acute care looks like, we have even more difficulty defining when it ends. At some point, it attenuates, but it is very difficult to do this.
If you look at it statistically, probably somewhere between three and four months accounts for at least three standard deviations in most of the post-acute care. But as you are all aware, Medicare home health care has made an art form out of becoming less and less acute and more and more long term care. Sot there is a growth spurt, or at least there had been in this post 120 day period, which was still called post-acute care. I guess people could get lifetime post-acute care if the thing continued long enough.
The other two major forms of post-acute care have generally much shorter tenures, largely driven by regulations and to some extent, tradition. Rehabilitation is usually over within three weeks. Medicare, the way the payment system goes, raises its rate to the consumer dramatically after the 20th day because of the co-payment requirement, so there is a big drop in the use of Medicare certified post-acute care after 20 days. But there is a tail out there, and it goes on for a fair piece of time.
Now again, we need to recognize that all of these definitions are the artifact of the payment system. None of them are clinically correct necessarily, or things that we would necessarily want to encourage if we were designing the system de novo.
So I'd like to talk about six issues that I think are relevant to your deliberations this morning. The first one is, who gets what kind of care. The second question is, what difference does it make. The third would be, how do we look at the costs across these different venues of care for different kinds of people. Then a basic policy decision about whether we want optimal care, that is, the kind of care that produces the best result or the kind of care that produces a good result at a cheaper price; are we optimizing or are we trying to become efficient.
What do we do with this problem of the fact that there are multiple users per episode, that post-acute care may actually have multiple people involved. Finally, would some kind of point of service capitation, that is, a sub-variant of capitation, work better than what we are doing. So we will take those on one at a time and explore those.
The first question then is, --
DR. IEZZONI: Can I stop you for a second?
DR. KANE: Sure.
DR. IEZZONI: So far, you have been talking primarily about Medicare, maybe some Medicaid as you begin to think about the nursing home. We have some private sector folks around the table. Can you just give us a sense whether the private sector paying for home health, et cetera, has mimicked what Medicare has done, whether it has paralleled it, whether there is great diversity.
DR. KANE: I think most of the care that we have seen has actually been among older people.
DR. IEZZONI: Like the home health has certainly not been.
DR. KANE: But the private sector has certainly done that. We have seen in private managed care, for example, more flexibility. Once we get away from the regulations, we see more of that.
But yet, I would say basically, the private sector for the most part has not shown any great burst of creativity, and has tended to follow the same models that we have created inadvertently through Medicare.
There is a strange thing about medicine. What we have done tends to be what we continue to do, even when we are given the option to do things with fewer constraints. So I think for a number of reasons, there is a risk to being too creative. The private sector has generally followed in those footsteps. Without some of the heavy regulations, for example, about the three-day rule that we have, where to get into a nursing home under Medicare, you have to have been in the hospital for three days, is not enforced in the private sector, at least not consistently.
DR. IEZZONI: Any other questions of the subcommittee before Bob moves on to addressing his questions? No? Okay.
DR. KANE: So the issue of who gets what care is a very interesting question. A number of people have tried to model this and have discovered that they can't account for large amounts of the variation, which is probably because a lot of the variation varies. It isn't a systematic behavior, so it is very difficult to model.
There seems to be reasonably consistent findings that several things do influence who gets this care. One is diagnoses. There are a certain cluster of diagnoses, things like stroke, hip fractures, congestive heart failure, chronic obstructive pulmonary disease, certain high cost, less frequent events, tricky kinds of wound care, tracheostomies, that seem to be more likely to lead to some kind of post-acute care.
Functional status seems to play a very strong role in this, both in terms of whether you get post-acute care or not, that is, people who tend to be more disabled if they get post-acute care, but also there is a Goldilocks phenomenon here, where there are people circling around trying to find the bowl of porridge that isn't too hot and too cold. So rehab, which is probably famous for being extremely selective of the people they will choose to treat in the best traditions of Hamarabi, go in there, basically trying to find people who are difficult, but not too difficult, who have problems but who have some potential for benefitting from this treatment. So depending on what your perspective is, there is careful case selection or cherry picking, depending on where you sit in the process there for doing that.
So the rehab folks tend to be more disabled than the people who go home, but less disabled than the people who go to nursing homes, for example. So one needs to do some case mix adjustments for this kind of thing.
Cognitive status is a fairly big predictor, in the sense that people who are reasonably severely cognitively impaired are not as likely to get rehabilitation, for example, and are more likely to go to nursing homes.
Living situation is a big ticket item, living situation here being the availability of informal care measured in various ways. The probability of going home is very much influenced by whether or not there is somebody to go home to, although there is data that people can go home without informal care and do okay.
One of the interesting phenomena is that hospital stays have come to be measured in nanoseconds. We are increasingly discharging people from the hospital at a very early stage in their clinical careers. So what used to be the process, where you would evaluate somebody and then decide what they needed, there is no point in evaluating them, because they are barely recovering from the anesthetic at the time the discharge planning process starts. So what you are really doing is betting on how they are going to do, on the basis of some sort of clinical prognoses.
But the problem is, there is no systematic prognosis going on. So the people who are making the decisions are making them with varying amounts -- a very generous and vague term, which really means no -- information about what their clinical status is. There is a whole area here that needs very careful attention in terms of what do people think these people really are going to achieve in terms of their clinical status, and to what extent is that going to be influenced by the kind of care they receive.
So what happens is that most of the decisions seem to be made not so much on the basis of careful clinical deliberation, but rather, the first train leaving the station is the best one to be on, regardless of where it is going. There are certain kinds of post-acute care that are easier to package than others, so there tends to be an attraction to trying to get those people into those kinds of care, simply because they are available.
Again, that may lead to some -- not irrational, but rational on a different basis decisions than what the ideal clinical situation would produce.
So we are making a decision without a great deal of information. Part of that problem is that we don't have a great deal of information. There are very few studies that really can tell you what difference post-acute care makes. So we can't exactly blame the discharge planners, because nobody is giving them the information to ignore; they just don't have it.
There have been on randomized trials. A couple of the people sitting around this table have done some quasi-experimental analytic looks at this to look at this. There is always the concern that the case mix is different because the people going to these different venues are different. There are overlapping distributions, different people with different philosophies or different religions, trust statistical adjustments more than others. This will be the great 21st century crusade, I'm sure, when the infidels of non-adjustment will do battle with the truth seers of adjustment.
When you look at these things with modest amounts of analysis, where rehabilitation is offered, it usually emerges as the most effective kind of care. That is, there seems to be a payoff for rehabilitation. The other shoe is, it is expensive. So this gets you into this question about, rehabilitation works, but it costs a lot, both in first order and second order costs, in terms of what really happens.
So that leads us then to the third point, which is that there is a big variation in what things cost. If you look at the direct cost, which is the cost associated with the care, rehabilitation is the most expensive. If you look at the indirect costs, which are looking at this per episode, and you can take different time cuts in terms of what constitutes an episode, if you look at it within three months or six months or a year, rehabilitation is still the most expensive.
It isn't as if you have this investment philosophy, where if you go in and rehabilitate the hell out of somebody, then basically they do so much better that they save money down the road; it just doesn't seem to happen. Part of that may be this uncorrected case mix difference, but again, what we have said is that rehabilitation doesn't get the worst cases. They get bad cases, but not the worst cases. So it is a hard argument to make very convincingly, that it is simply the bad luck of case mix.
So you will then come to different kinds of conclusions, depending on what is the criterion for choosing what kind of care you want. You can have optimal care or you can have efficient care. If you want optimum care, it turns out that home health care does a very good job -- not quite as good a job as rehabilitation in a number of areas, but quite a good job. If you are trying it by efficiency, then home health care is a much better buy than rehabilitation. If you are optimizing, then for certain kinds of problems that are rehabilitable, rehabilitation is a better buy. It is a question. What is the measure of success that you want to use in order to make your prudent purchase.
Now, if that weren't complicated enough, the other problem that we have is that what you get isn't necessarily what you see. People have what can be thought of as post-acute care careers. We followed people, tracking them on a regular basis for five diagnoses after they left the hospital some years ago, and discovered that people made up the six transitions in the first six weeks of their post-acute care careers.
Korbin was telling me that the record, he thinks, is 13 in a year, is that -- ? Eighteen in a year. Korbin is working off less personal data because this is just looking off the Medicare records, so some of those people may be having slightly separate episodes that look like one continuous episode.
I can tell you that at least in the short term, people may get a lot of different kinds of care. So one of the questions is, who do you hold accountable for this? You could have home health care after they have had two or three other kinds of post-acute care. You could have home health care right out of the chute. It is not necessarily all the same home health care.
We have for our analyses always argued that the first decision is the critical decision, and all the other decisions are essentially contingent on that first decision. So we have tried to trace everybody back for accountability purposes to that first discharge decision after they leave, but other people are looking at different kinds of effects.
Part of the inconsistency that you see in results is that some people may look at people who received any home health care during a period post-hospitalization, which is different than people who were discharged for home health care. So as you are looking at data, you need to be sensitive to what was the definition of what represents the care, because somebody who got home health care after they had either succeeded or failed in rehab, gone on to a nursing home for awhile and then got discharged back home may be different than somebody fresh out of the hospital, having gotten home health care as the first thing that they got.
The other complication with the multiple users thing is that Medicare set up this criterion for rehabilitation that you had to get three hours a day for rehabilitation, leaving aside the philosophical question of what is a day. It turns out there are weekend days and weekdays, and it isn't clear whether a day is every day of the week or just days of the week when people are around to give the care. Leaving aside those kinds of things, the problem is that some people aren't well enough to tolerate three hours a day, and so some of the people wind up back in the hospital, not because they need to be in the hospital, but because they can't tolerate the rehabilitation. It is a strange tangled web that we weave as we write these regulations for giving this kind of care.
Now, one solution to this rather tangled morass of post-acute care might be to say, can we develop some kind of a capitation system. It would seem that right now, we have created something that we never meant to do. What we have created has responded to economic incentives. As it became more desirable to shorten hospital length of stay, we then created this second round of services which of course we all appreciate means that HCFA is in effect paying twice for these services. They are paying the hospital under the DRG and then paying the post-acute care provider for the service that the hospital used to provide and that generated the hospital payment rate. So we have this interesting kind of thing.
So one solution to this that is attractive to some of this is the idea of doing what you can think about as point of service capitation. That is, at some point somebody would bundle together a set of services. There are great minds contemplating what is the right bundling to be done here.
There are two models that quickly come to mind. One would be to bundle post-acute care services, so at the point of hospital discharge you would be handed over to a post-acute care program that would then allocate the patients to the most appropriate post-acute care service and be responsible for their care for some designated period of time.
The second would be to say that in effect, the real problem that we have with post-acute care is that there is no locus of responsibility. It isn't clear right now, from not only a fiduciary standpoint, but from a quality standpoint. If somebody does badly after a hospitalization, is it because the hospital didn't do what they should have done, or is it because the post-acute care system failed to do what they should have done?
At the moment, you can point your fingers in both directions. If you take for example the case of stroke, with the exception of some of the thrombolytic champions, most people would argue that probably what happens post hospitalization is at least as important, if not more important, than what happens during the hospitalization.
So if you are looking at the outcomes of stroke and you are tying them back to the hospital, a lot of it may rest on what happens on things that hospitals may not have a lot of control over. So there is an argument to be made that what you really would like to do would be to develop some kind of a bundling system that would bundle the payment for the hospitalization and the post-hospital care, since the post-hospital care is largely this artifact of the way we pay hospitals. Before when we paid hospitals, most of the post-hospital care was given in the hospital. It wasn't called post-hospital care, it was still hospital care.
So one of the questions that is continually being debated is whether or not hospitals should then be more directly involved. Obviously, the pro arguments for this kind of combined bundling would be that it would create a much easier accountability system, both financially as well as for quality purposes.
That makes a lot of sense clinically, because these distinctions are really terribly arbitrary. It would in fact make the whole process of clinical planning or discharge planning much more realistic if the hospital had a stake in the outcome. If the hospital were now responsible for the financial and the clinical outcome, one might suppose that hospital discharge planners might behave differently, or that hospitals would encourage them to behave differently.
On the other hand, there is a great question as to whether hospitals are competent to do post-acute care. Hospitals have basically suffered from this pre-Columbian claustrophobia, that basically the world begins when people are admitted and ends when people are discharged, so everything else is a great black void. So they don't seem to have much competence in things other than high tech institutional care. Concerns have been expressed about whether hospitals should really be in charge of this kind of system.
It doesn't have to be hospitals in charge of this kind of system. One could create a new corporation that would in fact contract with both the hospital and the post-acute care system to manage it, so there are maybe ways to get around that. There are some models now developing to do that.
But anything less than doing this probably won't get you the real benefits in terms of the economic. The economic care is not just in dollars, but in terms of quality, the real economies of scale that you need to have to really make the changes in the path by which these people are truly managed.
We have become in this country experts at niche masters. We have developed our little boutique operations of various kinds of things, and fragmented the system. Unless we put it back together again, it is not likely that we are going to change very much for the better.
So since this is a group that is triggered by information, and your raison d'etre is to define information needs, I thought I would raise for you a couple of information needs that we have with regard to post-acute care.
The first one is that we have no empirical basis for making discharge decisions. We simply don't know what characteristics are related to outcomes for this kind of care. We have a little bit of information, but not a whole lot.
We don't know what kind of care works best for what kinds of people. We can define people in a whole variety of different variables. We need to find a way to collect information not only on traditional characteristics, but also on this clinically elusive but probably very potent issue of prognosis. Is there a systematic way in which we could actually collect information on clinicians' prognoses.
In fact, there is an interesting question about which clinicians' prognosis is most important. We tend to think that doctors know best, but in fact they may not. For many of the kinds of problems that we have in post-acute care, which tend to be things like strokes of hip fracture, things like that, the doctor who is treating them may not know the patient very well at all. It may be an orthopedist or a neurologist who is seeing them for the first time. It may be the primary care doctor, it may be the nurse on the floor. We don't know whose prognosis has the most validity in terms of looking at this.
We need to know something about what is the actual service capacity of these various post-acute care vendors. People have declared themselves expert in a number of areas. Nursing homes have quickly put up banners saying, we do subacute care, which is not quite the same as sub-optimal care. People have tried to study this phenomenon and have had a hard time identifying what this subacute care really is. It looks different in different places.
HCFA very wisely has refused to acknowledge it, which I think is a very good step, because it would just jack up the price for HCFA. But we don't know who does what.
Again, a lot of what we have done has been the result of HCFA mandates about what you have to do. If you want to be a Medicare certified vendor, you have to do thus-and-so, but there is no great body of science behind those dicta. What we might want to do is develop a much more flexible system that picks and chooses from some of the best of different modalities of care.
Why don't we do rehabilitation at home, and combine home care and rehabilitation? It isn't clear that you need a card-carrying rehabilitation person, whether it is a physiatrist or a physical therapist, to do a lot of this rehabilitation. You probably could do it in very different ways. So you could probably come up with some very interesting scaled-down models of rehabilitation that might be at least as effective and much more efficient in doing this. We just don't know.
We don't know what is a unit of service. Somebody goes and gives a home health visit. You could have a home health visit defined by who gives it, but that isn't exactly what is done. Again, who gives it may not be so important. It may be what is really done. So defining things by who does them may not be necessarily a very good strategy.
We don't know very much about who really is able to care for whom. We are learning regularly that there is an enormous capacity for downward delegation. We can keep giving other people responsibility for doing all sorts of tasks, but we haven't begun to really track the art of the possible in post-acute care. In fact, the payment regulations get in the way of doing a lot of that, because they require that you have a card-carrying X doing this sort of stuff.
So I think it would be very helpful to know more about what is the content of post-acute care. If we wanted to have something that was going to be useful for either analytic or payment purposes, it would be nice to have something called a post-acute care unit, that really was consistent across at least, a type of post-acute care and maybe across post-acute care in general, that would tell us what a real unit of service is.
Right now, we have home visits, we have hours, we have hours of different kinds of people delivering services. We have services in in-patient situations, services delivered in the community. It is a very mixed bag.
We need much more information about whether this kind of a unit of service really varies by the type of service given and by the site at which it is delivered. Is an hour of physical therapy the same at home as it is in a physical therapy suite? We just don't know. Is it the same in terms of its effectiveness?
We need to develop a whole longitudinal database to understand this question about what really happens to people over time. Post acute care has a long time perspective to it, and we don't have much information about that.
We need to understand how do we transmit from one vendor to another or one provider to another within the system critical clinical information. We know that there are a number of these transitions going on, but we aren't doing much to support that infrastructure to really provide what is the key clinical information that needs to be provided at these times of transition.
For example, it would obviously be important to know what medications they are on. It would also be important to know what was their clinical course, how is their situation -- presumably, if they are making a transfer, they are making a transfer because they are getting better or worse, so you ought to know something about what were they like before.
The best predictor in medicine of what you are likely to become is what you were. Very little of medicine makes you better than where you started. Some may argue plastic surgery is beneficial, but there aren't too many examples where you get better than your baseline.
So having a clinical course, particularly for complicated chronic problems, having some specific parameters that constitute that course, and doing well is not necessarily a rich set of information to transmit from one -- or doing badly, from one provider to another. We really need to find the appropriate clinical parameters to try and do that.
Then we need systematic case mix adjusted information on what are the factors that are associated with outcomes for a large body of cases. We don't have this. The dilemma in health services research is, we have a lot of data, but it is very thin on a lot of people. That is what we can get out of the Medicare system. But there is no clinical information there. It tells you a lot about services, but not a lot about people. Then we have some very rich data sources, where we have done in in great depth and collected a lot of information, but we obviously don't have it on a lot of people. So we need to somehow try and cross over those two.
Finally, I would raise some questions about quality that I think need to be thought about. One is this question about who really is competent to provide post-acute care, and can you provide the same equivalent care as an out-patient as you do as an in-patient. Remember, a lot of these definitions ar arbitrary and have been driven by various payment regulations.
One question is, does the type of post-acute care that a person gets really matter. Can we separate out the site of care from the type of care. There is a fair amount of confounding going on right now, again driven by the way we set up the payment policies. But it may be possible, for example, to look at the importance of physical therapy, the stink comes from whether it is given as part of a rehabilitation program. So one may want to focus attention on the smaller units rather than the larger units to disintegrate some of those things.
There is a question that I think is very interesting, that Andy and I have tried to look at at different times, this question about, does this idea of subacute care really mean anything. I watched nursing homes in California change their names from convalescent hospitals -- they went through three or four name changes along the way. Other than signage, I'm not sure that there was a great shift in activity. What you call yourself may or may not mean anything, in terms of becoming an issue of great concern.
There are now international associations of subacute care, so they at least have enough money to meet. I don't know that they do anything.
Very important question is whether the amount or the intensity, which are somewhat different concepts, of post-acute care really make any difference. We have taken all these strong stands about, how do you have to give this care, but of course nobody bothered to base it on any facts. So it would be worth going back and looking at whether you need a certain amount per day or whether you need a total amount of -- maybe we should make a rule that every stroke should have 217.7 hours of physical therapy. We just don't know. Does it matter whether they get that in the first week, or could it be given over a year? There are all sorts of questions about treatment. Treatment isn't one single black box. We need to think about timing and other kinds of factors.
Then I think the key to the quality question, which is probably less a science question than a policy question is, who should be responsible for the outcomes of care? How do we get some real accountability build into the system? Right now, what we have is a very fragmented system, where it is very difficult to create any real accountability. Everybody says, I did my job, but they screwed up over there, and so we are much more adept at finger pointing than we are necessarily at improving quality.
So I would suggest that those are all quality issues that need to be addressed. They really trace back to information needs to staff them, to serve them. In the end, a lot of this really comes back to making some decisions about what it is that we think this phenomenon of post-acute care is.
I'll stop there.
DR. IEZZONI: Bob, that was a wonderful introduction. Why don't you untether yourself and sit down? The subcommittee, can we start with questions from you all, and then we'll deal with questions from the audience? Paul?
DR. NEWACHECK: Yes, thank you for a very interesting presentation. You identified a number of different information and data needs. I'm wondering if you could tell us if there are any that you think are of extremely high priority. That is, which of these are most pressing, given limited resources to address all of them?
DR. KANE: I think a number of them feed off each other.
DR. IEZZONI: Bob, could you make sure to talk in the mike? Thanks.
DR. KANE: Sorry. I think a lot of these feed off each other. But it seems to me that the key question that we need to understand is what makes a difference.
Now, in order to understand that question, we need to have definitions of what are the various components of the service and the effects that would allow us to do this post hoc.
I'm starting from a premise that I don't think that we are going to have the luxury of doing a great many randomized trials in this area. I think that the temper of the times isn't going to permit it. So we are going to be relegated to doing quasi-experimental, deductive analytic studies. The question is, can we build up enough of a reasonable database that will allow us to apply these epidemiological methods to these kinds of questions.
We have been doomed in the past to be forced to use incomplete data to make rather extravagant kinds of conclusions, and the result has been that we have fallen back on non-data to make many of the policies.
If we are going to change that, there are two directions we can go. One direction is to say, we don't know how to control the system, so we are going to pass it over to another group of people that we'll call managed care institutions. Then all we'll do is pass the buck without a great deal of knowledge, and then we'll put most of our effort into somehow trying to calculate the fair rate to pay for this system. That is what we have done with most of medical care right now, not very well, mind you, but we've done it.
The second approach is to say, we really want to understand this system. We want to try and manage it intelligently. In order to do that, we need much more data about what makes a difference, so we know where to hold fast and where to give way.
DR. IEZZONI: Barbara?
DR. STARFIELD: Hi, Bob. You started off your presentation by talking about the lack of conceptual clarity. I think perhaps this committee can help with that over time. But we need help too in doing that.
You said you were going to limit your remarks to care that are post-hospitalization, and you did. That led you to some conclusions or some possibilities for data, like for example making hospitals responsible for the whole thing. But if you stop to think that if you don't limit yourself to post-hospital care, that becomes a very inappropriate solution. So it is not clear to me why you would even want to think about that solution, when you think about the broader field of the continuum of care.
That's one thing. Another observation is that part of the problem we may be having is that we are lumping things in the wrong way. When you stop to think about hospitalization, that is not a defined unit of service. Yet, we have not had any trouble coming up with data, mechanisms of collecting data from in-patient stays, even though it is not a defined service.
Hospital stays are similar to -- except maybe for the term and lack of end of the term, although there is always an end of a term, either death or something else. So that data for hospitals is very similar to data from nursing homes, for example. It is facilities oriented, whereas home health services and rehabilitation services are much more similar to ambulatory care data, because they are services oriented.
So maybe we just split the pie wrong. We have lumped things together that shouldn't be lumped together, and that is keeping us from coming up with a reasonable data strategy for them. Anyway, your comments?
DR. KANE: Yes, in the best of all possible worlds, we wouldn't have any designations about who gives what care. We would talk about people and we would have a people centered system that would attract people. But we are a long way from doing that.
Your observations about the nature of data collection are quite correct. In fact, I would argue that we have very bad data about hospitals. We tend to focus on things like length of stay and survival. We don't make very good distinctions about what goes on there. We tend to think that surgery somehow defines the type of treatment, when in fact it covers a broad multitude. There are alternative treatments to surgery now that we don't even look at.
So having backed our way into that kind of a designation, I'm not sure that we have necessarily created a very rich data system for doing that.
In the best of all possible worlds, we would equip each person with an implanted monitor at birth, and it would send out a signal when illness began, and then we could collect really good health services research data, by tracking them prospectively from the onset of this first indication of illness. But I don't see that happening in the near term, and there may in fact be for some reason people who would object to that.
So we are probably going to have to come up with some kind of an approximation of this. One of the approximations that we have tended to use is critical events, and whether it is coming to an emergency room or coming to a hospital or being admitted to a nursing home or even going to see a doctor, we clearly set up a system of health statistics that are driven by people's utilization.
Certainly, people like yourself, who have looked at the course of illness and recognize that some people don't avail themselves of services at the same time, if ever, there is a big bias built into that system. But it is a system -- one of the reasons why the United States has the best data system in the world, or used to have, was because we paid for everything on a fee for service basis. So we collected a lot of data basically driven by payment, not by health concerns.
I think we are going to continue to do that. Post hospital care or post-acute care is not a super precise term, I agree with you. There are going to be people who I think -- we are already developing these models in Minnesota, and I suspect other people are as well, get this care without ever getting into the system, mostly nuder managed care auspices, because that is where you are allowed to be more innovative to do it. So it won't capture everybody in exactly the same way.
But I think it is a pretty good way to begin. Most of the people who begin their careers in this area tend to begin them with a hospitalization. I think one of the key areas of problem that were have in this country right now is how we handle the end of hospitalization. I would suggest that one of the areas of broad endemic malpractice in this country is what we call discharge planning, or what we prefer to call launch coordination. It is the ability to count backwards from 10 that defines the prime characteristic of a discharge planner.
I think that we need to begin to focus some attention on that problem. I think defining it that way makes it a potentially manageable problem. It also introduces a certain amount of artifact into that, because it triggers it off of the hospital, and we will lose that portion of people who don't come through the hospital as the way of getting into the system.
But I think we could build a very good system through the hospital, and make adaptations to it to begin to accommodate other people who look like those people. I think if we try and start the other way, we are going to have a very amorphous system with no social structures to it, around which it will be difficult, however morally correct, to build a useful data system.
So I guess I am willing to make that tradeoff of saying, this is a somewhat arbitrary definition. But hell, the world is filled with arbitrary definitions. We compare hospital care. We know that there are people out in the community getting ambulatory care that looks a lot like hospital care. Gradually, we take the hospital care and we begin to say, okay, now can we have alternatives to hospital care, and we make the box a little bit bigger by bringing in those things.
It is the successive approximations that we tend to use in the way we collect data. I think we would make a giant step forward if we could just collect some decent data to answer some of these social problems, even if it was less than perfect.
DR. IEZZONI: A quick rejoinder, Barbara?
DR. STARFIELD: A quick rejoinder. You started off by saying how bad hospital data was, right? I wouldn't want to split it to other things. I would rather start new.
DR. KANE: However bad it is, it is a hell of a lot better than what we have now in these other areas.
DR. IEZZONI: Kathy, did you have a question? Then we'll close with one question from the audience, because I want to move on to Korbin.
DR. COLTIN: I just have one question. I didn't know, when you were talking about direct admissions to post-acute care, whether there are in fact data available on what percent of admissions to each of these various types of settings are direct admissions, whether they differ at all in terms of the clinical conditions that the patients have, and whether it has been changing over time. Is there information about that available, and is that something you see as --
DR. KANE: The question is actually more complicated than that, because what we have are a series of providers who provide a variety of services, not all of which are post-acute care.
Now, part of that is definition. For example, if you take the nursing home, even if for the moment we don't look at the front end, we look at the back end, I think most people would argue that if you have a demented person in a nursing home for eight years, that is probably -- they passed the post-acute period some time ago. We may not know exactly when. Under home health care, I guess if they were demented at home, they could still be getting post-acute care.
So what you have here is this much more difficult problem of providers of services who are providing a variety of services to a variety of populations. So defining what is a direct post-acute care admission as opposed to a direct admission, long term care admission, in the case of a nursing home, is difficult.
We have data on -- not systematic national data, but data from a number of sources now, on people who are coming into nursing homes from different places, so we can tell you what proportion of people come into nursing homes from different locations. Some of it is not great data, but it is pretty good. But that doesn't mean those are all post-acute care admissions. So it becomes harder to interpret.
Likewise, home care agencies, some of which are home health care agencies certified by Medicare, some of which are agencies which provide home health care services, but also provide a whole host of other services, sometimes using similar if not the same personnel to do both kinds of services, get their clients from different sources under different auspices. You could track the change in their business over time, but I'm not sure -- you would need a better definition first of what constitutes a post-acute care service to know what proportion of those were direct admissions.
In the Minnesota example I was citing, rehabilitation is probably -- I may be wrong, but I don't think there are too many direct admissions to rehabilitation, but there may be some. In the direct post-acute system I was talking about in Minnesota, this is being run under the auspices of a managed care company, and they are using the same vendors and basically admitting right out of the ER. But that is still innovative. I wouldn't say experimental, because that would give it a pejorative twist, but a developmental process that is being tested right now to see if that could be done. It has been described in the literature.
DR. IEZZONI: Let's just have a quick question and quick response from the audience. Could you please identify yourself? Then we'll move on to Korbin.
MR. MUSE: Don Muse, Muse and Associates. Just a quick question. Going back to your slide that fascinates me, who gets what care, it looks like the research has focused on the characteristics of the patient as predicting different types of post-acute care.
Are you aware of any research that would look at the characteristics of the provider, in terms of who gets care afterward? For example, one could hypothesize that a hospital that owns a nursing home is more likely to refer people to a nursing home than a hospital that doesn't have one, or home health or rehab hospital. Is there a body of literature out there on that?
DR. KANE: There are some appendages. I'm not sure they quite add up to a body yet. We actually did a study that looked at that very question. We believed going into this that any intelligent hospital would obviously cream off the good cases for themselves. So we were following a couple of thousand post-acute care cases, and actually went back and re-analyzed it, got data from hospitals on their ownership patterns, and discovered to our amazement that it had profoundly little effect on what the hospitals did.
Now, this only adds credence to the belief that hospitals are incompetent to manage post-acute care, or at least it hadn't been a big enough issue on their radar screen at that point in time to act that way. One would certainly anticipate selection bias. We couldn't find it, and we looked for it pretty hard. There have been about three or four other studies with mixed results in terms of how strong that selection is.
You tend to see a lot of it in the literature among post-acute care providers, that they are being biased against by the hospitals. But in terms of what we call disinterested research studies, I'm not aware of that data that really suggests that what would make sense to expect has manifested itself in any significant way yet.
DR. IEZZONI: Great. Bob, can you stick around?
DR. KANE: Sure.
DR. IEZZONI: Great. Thank you, that was very informative. Korbin, let's hear from you.
DR. LIU: As Bob was going through the 65 questions about what we don't know about post-acute care, I thought it would be a really nice presentation if the rest of us could answer those questions this afternoon and the rest of this morning.
Unfortunately, we can't, as you might expect. We probably won't be able to for the next 25 years, even if they increase HCFA's research budget by 2000 percent. What I would like to do is highlight some of the points that Bob made this morning with some data that we actually do have.
I'd like to refer you to the handout with a number of tables in them. Like other self-respecting researchers, I stole the data from other people whenever I could, but these are good numbers, and they are mostly from claims data. We can thank Propac and Medpac for a lot of these numbers.
Let me refer you first to the first table, which is Medicare SNF spending and use. As you can see, one of the reasons why some of us are here today is because of that first chart. The expenditures for Medicare SNFs rose dramatically since 1990.
For example, in 1990, Medicare spent three billion dollars for skilled nursing facility care; by 1996 it is up to about $12 billion. So that is a four fold, 400 percent increase in a very short period of time.
The table underneath gives you some information on the utilization of the SNF services. What is interesting is that although the length of stay -- again, looking for example between 1990 and 1996, the length of stay in SNFs under Medicare are about the same. On the other hand, the expenditures have gone up 400 percent.
What basically has been happening is that there has been some increase in utilization. More people are using the SNF benefit. The length of stay is about the same, but the cost per day has been going up dramatically. I think this is due to increased use of therapy and other ancillary services.
The next table gives similar information on home health benefit. Here again, it increased from about four billion in 1990 to $18 billion in 1996. In the case of the home health benefit, the use rate doubled from about 57 per 1,000 in 1990 to 98 per 1,000 in 1996.
In addition, the visits per person increased dramatically over that time period. It increased from 36 to 77 visits per person. Hence, unlike the SNF, the expenditure increase in Medicare home health is due both to utilization and use per person.
What happens in Washington is, when the policy makers see numbers like this, or when they see curves for expenditures rising at that rate, they have to do something. So typically, expenditure increases like this trigger and stimulate a positive response. The most recent policy response of these numbers was the call for prospective payment systems that were mandated in the '97 Balanced Budget Act, or Better Budget Act, as some people think of it.
So that was the immediate kneejerk response. The Balanced Budget Act mandated prospective payment systems for skilled nursing facilities, home health agencies, rehab facilities and a study on long term care hospitals.
The Balanced Budget Act did not respond to a lot of other issues though related to post-acute care. Because there were provider specific payment system responses, it did not address some of the issues that Bob raised, like transfers from one provider to another.
It also did not call for an examination of the overlap patients between the various providers. This is an issue that Bob had raised earlier: who does go where and who are these post-acute care patients.
Let me refer you to the next table. This is a table we actually constructed. This is changes in Medicare expenditures for various time periods. I think the most interesting rows are four and five. These are percentage of Medicare spending growth attributable to in row four hospital in-patient care and row five, skilled nursing facility and home health care.
So in the pre-PPS period, 1974 to '83, in-patient hospital care accounted for 63.5 percent of the growth in Medicare expenditures. For that same time period, SNF and home health accounted for 3.4 percent.
If you look at the last column, 1990 to 1995, hospital care accounted for 38.5 percent of increase in Medicare spending, and SNF and home health accounted for 30 percent. So there was a dramatic shift in the distribution of Medicare spending growth.
Now, purely coincidentally, for the 1974 to '83 period, in-patient, hospital and SNF home health accounted for two-thirds of Medicare spending growth. In 1990-95, those three categories accounted for about two-thirds of Medicare spending growth.
We recognize the fact that the system was a lot more complicated than the coincidence suggests. It would be simplistic and nice to say that there was an offset between post-acute care and in-patient care, but they both accounted for about the same amount of growth in the two time periods.
I think it just simply leaves us with a beacon for further research on the potential for substitution. It raises the question though, how much of this expanded growth in post-acute care recently is due to inappropriate use, how much of it is a response to need that resulted from the implementation of hospital PPS. Again, it is a question we do not have an answer for.
The next table highlights the current use of post-acute care. I think the key number is the 23 percent in the first row, which says that about a quarter of all hospital discharges under Medicare result in use of one of the multiple post-acute care providers. So that is a very high incidence rate of post-acute care.
Most of the discharges are to SNFs and home health and discharges occurring within one day of hospital discharge. About 11 percent are admissions to SNFs and about nine percent are admissions to home health agencies.
The lower incidence of admissions to rehab and long term care hospitals may simply reflect the fact that there are not many of them, relative to SNF and home health. I think currently, there are about 15,000 Medicare certified SNFs in the country, about 10,000 Medicare home health agencies in the country, approximately 1,000 rehab facilities and approximately 200 long term care hospitals.
The next two tables look at the transfer between post-acute care providers. Table 6, these are from the Medpac report of 1998. The top one gives the proportion of post-acute care users who go to a second post-acute care provider within an episode. Of the people who go to a post-acute care provider, approximately 18 percent go to a second one within the episode.
Then it breaks it down by the major post-acute care providers. Rehab facilities tend to have the highest use of a second provider. Almost all of the cases that are sent to rehab facilities end up going to a second post-acute care provider. If you are admitted to a home health agency after a hospital stay, you are very likely to end your episode with home health care.
Then Table 6-5 gives a breakdown. It gives patterns of --
DR. STARFIELD: When you say the percent followed by a second, does that mean a different type of, or could it be the same?
DR. LIU: Different type of. It could be, but it is most likely a different type of. Table 6-5 gives some clarification about that. It gives the patterns, at least for the first two transitions.
For example, the SNF to home health transition occurs in 15 percent of the cases, and the rehab to home health occurs in 37 percent of the cases. As Bob pointed out, the SNF to SNF does occur, but that is only 6.5 percent.
DR. IEZZONI: Aaron, did you have a quick question?
DR. HANDLER: Would it help at all to tease out the rehabilitation stays that are done because of motor vehicle accidents or falling off a horse, spinal cord injury, as opposed to stroke or illnesses associated with old age? Would that make a good example?
DR. LIU: I think that would be very interesting.
DR. HANDLER: These are Medicare data, right?
DR. LIU: Yes, these are Medicare data, but I think you're on the right track.
DR. IEZZONI: Fifteen percent of Medicare eligibles are disabled. So were these just 65 and older Medicare?
DR. LIU: I think everybody was included.
DR. IEZZONI: So there might be some folks with spinal cord injury and so on in there.
DR. KANE: But it would be their initial injury, but still wouldn't be their thing that made them --
DR. HANDLER: I think the perfect solution is if we had national health care.
DR. IEZZONI: Thank you, Aaron. I like these political comments. Korbin?
DR. LIU: Despite the VBA changes, observers still feel that a more rationale post-acute care system is needed. One view is the need for a more consistent payment across the Medicare post-acute care providers. They are reimbursed in different ways, they have different current eligibility policies, they govern their services, they have different certification requirements.
I think one of the points that Bob made was the notion of a patient system where the payment goes with the type of patient, rather than to a particular type of provider per se.
I guess before proceeding rapidly in any direction, it is probably worth asking the question, how broken is the post-acute care system now. The first question there is, how much patient overlap is there between SNF and home health and rehab? Then in cases where there is -- let's assume overlap in patient types, what types and amounts of services -- and this is the content that Bob was referring to -- produce the best or an acceptable outcome.
Then, how do we incorporate this information after we do all this research into the Medicare payment system? These are more questions. At the present, there are not too many systematic answers to these questions.
On patient overlap, we can get some sense by looking at DRGs. Let me refer you to the last table, which is the small table with 400 numbers on it. This is one my colleague, Barbara Gage, produced with Medicare claims data.
What we did here was to create episodes of care where we linked post-acute care events, and the events stopped whenever there was a gap of approximately 30 days of non-service use.
DR. STARFIELD: Can I just clarify again? These are the hospital diagnoses, is that right?
DR. LIU: These are hospital DRGs, yes.
DR. STARFIELD: The diagnoses that come from the hospitalization.
DR. LIU: Right, right What we did was to select the top 20 DRGs for each of the post-acute care providers, and then we combined those top DRGs for each of the providers and ended up with essentially 32 DRGs, which are the ones in this table. So the DRGs in this table make up a large number of the post-acute care cases for SNF, home health and rehabilitations.
DR. MOR: Korbin, the denominator here is just the -- you had some post-acute care in one of them?
DR. LIU: That's right. The first thing about this table is the fact that there is a concentration of DRGs that result in post-acute care use.
Let me point out a few things that come to mind. One is, the rehab cases tend to require therapy. The rehab here refers to rehab facilities. Stroke is a good example. About 11 percent of the hospital DRG cases that result in any post-acute care go to rehab facilities.
Again, for stroke, about 32 percent go to SNFs only, and 28 percent go to home health agencies. For stroke DRGs, about 30 percent go to multiple post-acute care providers.
It is also worth noting that for rehab facilities, some of the DRGs are just non-existent. A lot of the heart disease-lung disease related DRGs are not seen in rehab facilities. For chronic obstructive pulmonary disease, this is DRG 88, which you guys know very well, there is just a marginal number of cases that end up in rehab facilities. They tend to be treated in the home health situation.
Also, in general we have a group of cases that had multiple post-acute care providers after the hospital stay. They tend to be more common among the rehab DRGs as opposed to the medical DRGs.
Now, this is interesting, and it gives you a notion that there is some overlap. The question is, how good are DRGs as indicators. The table would suggest that there is overlap, but I think some of the other factors which Bob pointed out where very important in terms of who gets what where and how much you need, are not available in the claims data. So the informal care, functional disability and cognitive impairment factors are not in here, plus there is a lot of heterogeneity within the DRGs.
So whereas it gives you an indicator, there might be overlap. Again, it does not clearly answer the question, how much overlap in specific patient types do you find across the various post-acute care providers.
From a bottom line data perspective, we really need a lot more information, particularly information that allows us to compare across post-acute care providers. This applies to the patients. Are they similar enough that we can case mix adjust or are they different enough that we need to create different categories. The content question, we have no clue who gets what, where. Finally, there is very little information on outcomes of patient care. I think Andy Kramer, Bob and a handful of people have done research in this area, but there is really a paucity of data.
Fortunately, -- I'd like to end on a positive note -- our department has multiple projects that they have initiated to begin to answer some of these questions for us.
Thank you.
DR. STARFIELD: Liza, can I ask a question about the tables before we go on to the discussion?
DR. IEZZONI: Yes.
DR. STARFIELD: I know the last tables are all post hospital, but I'm not sure about the first three pages of the tables.
DR. LIU: For SNF, SNFs are post-acute. For home health, I think the home health includes home health care.
DR. STARFIELD: I have trouble with post-acute. Is it post hospital? That is really what I --
DR. LIU: The SNF is post hospital. The home health refers to all home health, some of which occur after a hospital stay and some of which do not. So there is no requirement under Medicare that home health services have to be preceded by a prior hospital stay.
DR. STARFIELD: So the first three pages of data at least with regard to home health and rehab are not necessarily post hospital?
DR. LIU: That's right.
DR. STARFIELD: Thanks.
DR. IEZZONI: David?
DR. TAKEUCHI: Korbin, I was wondering. About a quarter get some kind of care after discharge. Is there a sense as to the characteristics of the folks who are getting, in terms of men, women, race, ethnicity and those kinds of characteristics?
DR. LIU: Yes. There is some basic information that is available through the Medicare claims. These tables heavily use Medicare claims data, age, gender, possible Medicaid status are all available through the administrative records system.
As I recall, the older you are, the more likely you are to use post-acute care. The age may be in part a proxy for functional status or something else. Women I think are more likely to use nursing home care. So there is some literature along those lines.
The other source of information are essentially the other type of data, where there is survey data, with a lot of explanatory variables and very few cases. They do provide the kind of breakdowns you are referring to. Functional status comes out extremely powerful as a predictor of post-acute care. It is linear. We have looked at IEDL versus ADL-1 to ADL-34, and it goes up. The more disabled you are, the more likely you use post-acute care.
DR. IEZZONI: Other committee members, are there questions? Barbara?
DR. STARFIELD: Clarification again. Does HCFA have information on the rehab and home health users, on whether they are post hospital or not? Someone is shaking their head no back there.
DR. IEZZONI: You can probably tell by linking claims.
DR. RIMES: Linking claims is not a universal system, either.
DR. IEZZONI: Exactly. Kathy?
DR. COLTIN: For the conditions that you have listed on Table 6, all of the DRGs, do you have a companion table that isn't included here that would indicate what percent of the hospital discharges for these diagnoses got any post-acute care?
DR. LIU: Yes, we can.
DR. IEZZONI: That would be a good column to have.
DR. COLTIN: And also, what year are they for? Is this 1996 also?
DR. LIU: These are '95.
DR. GAGE: As Korbin mentioned, we have looked at some of these issues. The top five DRGs that went on to post-acute care in here, we're talking about DRG 127 and 209, 210, some of the back and neck procedures. A large proportion of those in the hospital then go on to use post-acute care.
Going back to one of the earlier questions that came up about the distribution across providers, again, it is only Medicare related data. But in the Medicare population, about 90 percent of the rehab admissions go on to post-acute care. So there is a pretty large proportion that is following a hospital stay.
DR. KANE: Barbara, explain. When you say rehab admission, do you mean the DRG for rehabilitation in the hospital?
DR. GAGE: No, I mean they go from a hospital -- in looking at claims, they are discharged from a PPS hospital into a rehab hospital within 30 days, correct. So it is provider based. And similarly with the home health agencies; about 60 percent of the cases are post-acute care under Medicare. Now, as has been mentioned, --
DR. STARFIELD: Post acute meaning post hospital?
DR. GAGE: Post hospital, I'm sorry. So you have about 40 percent of the Medicare home health users who are not coming within 30 days from an acute care hospital stay.
As mentioned earlier about the differences in the patients, if you look at the probability of admission to the rehab facility versus the SNF versus the home health, you can see that the rehab facility users tend to be, as the literature has shown, men, younger, more likely to have a better outcome, presumably, although that is not in the claims. There is no measure of function in the claim, so it is very hard to get at that. That is only a hypothesis that can be driven. As mentioned, the SNF users are the women, the older people. The home health users are either the younger disabled or the very old, but they also are within the 65 to 84 year old group, too. But again, that probably varies by whether they were discharged from a hospital or not.
So without function, it is very hard to really distinguish among these patients at any measure of function.
DR. IEZZONI: That is a clear message. I was going to ask Korbin about data gaps. We have heard repeatedly about the function, which is one of the things that we are going to be hopefully focusing on. Paul, you have a question?
DR. NEWACHECK: Korbin, you identified several different areas where we could use additional data or information. I'm wondering if in thinking about those whether or not you feel we can address those by modifying our existing data systems and tools that we have already out in the field, or do we need new data systems, for example, new surveys to be able to address some of these important questions about post-acute care, the continuum of care?
DR. LIU: My opinion is that the new surveys probably -- we probably can't afford new surveys that can give us the Ns that we would really like to have. Whether we can incorporate these into the claims, these data items into claims, I'm not sure.
There are other -- the assessment instruments that are being developed or have been developed seem to be a really good source of information on disability, because they are built into those instruments. That may be ultimately where we go with it. It is not the claims data, but it is administrative data.
DR. NEWACHECK: To follow up, if we did have more resources, what kinds of additional -- like you mentioned, you would like to have surveys that would have larger Ns. Would it be enhancing existing surveys or new kinds of things?
DR. LIU: The assessment instruments for the skilled nursing facilities, for example. The nursing homes are now reporting information, so I think right now, you have got a very rich data set source that is being developed.
The surveys, again, I'm not sure how much we can -- I think the MCBS is one that we have been using a lot, provides an excellent opportunity to look at changes in Medicare at a national level. IT has been continuous since '91, so you had a natural pre-Balanced Budget Act change baseline. My understanding is that it will continue to be collected, so that you will be able to see the implementation of the various prospective payment systems and the impact on specific subgroups of individuals. So I think that is a very concrete thing. I would certainly support continuation of that data collection.
DR. IEZZONI: Bob?
DR. KANE: The NCVHS is a wonderful tool. I think we all derive a lot of very useful information from it. But to study post hospital care, none of those surveys will get it. To study post hospital care, you need a data system which captures real-time events. The NCBS captures a cross-sectional measure of peoples' cross-functional status at some time during the year, and then you can relate that to their subsequent use of services. But that is not the same thing.
Peoples' functional status changes over time. So in order to get at the questions that we are interested in today, I think we need to understand that probably no survey would be around often enough to get that kind of information. What we need to do is to build data collection devices or enrich data collection devices into an ongoing system.
DR. IEZZONI: That is a helpful comment. Vince?
DR. MOR: Korbin, on Tables 6-4 and 6-5, the role of mortality or survival in these two sets in terms of what the influence of the denominator is, you've got this competing problem of -- you might not be around to have your next transition, and that likelihood of dying is substantially different as a function of which one of these options you go to first and/or second.
Could you just comment, or do you know anything more about that?
DR. LIU: As far as these tables?
DR. MOR: Yes.
DR. LIU: No, I don't know anything more about it. I bring these up primarily because they were illustrative, and from a global perspective they give you a nice picture of what is going on. From a hazard model perspective, no, I don't think that is built into this kind of a table.
DR. KANE: We did run some hazard models on -- this is Medicare data; we ran it off of the set of data that we collected. We couldn't find significantly different risks of dying across the different sites, once you case mix corrected for them.
So you get different case mix, or your point is correct that for example, people going into nursing homes are more at risk of dying than other people. But if you case mix correct, you don't find differences in mortality, but they would have less chance of making the second transition than would the other people. The people in rehab have the least chance of dying, and so they have the highest chance of surviving to make a second transition.
DR. MOR: I think that is the point I really wanted to make, is that this intervening differential risk of mortality as a function of what place you went to would affect the interpretation of those transition rates.
DR. IEZZONI: That is a very good point. Korbin, you stick around too, can you, for the next hour?
DR. LIU: I will.
DR. IEZZONI: We would like to conclude the morning by our two final speakers, Andy Kramer and Arthur Webb. Andy, do you want to kick off?
DR. KRAMER: Sure. There is a handout coming around. What I would like to do is to begin by talking a little bit about some of the background issues on post-acute care, although many of those have been covered already, so I won't dwell on those, move into some of the more specific issues related to information needs, and finally, get to this issue that was recently raised about data needs for particularly outcome measurement and quality measurement.
The perspective that we have taken on this is, we are starting from a clean blank slate on a study funded by ASBI, and saying, if we were to measure outcomes as opposed to acute care based on a definition of post-acute care that I will discuss here, what are the indicators of quality one would need and how would one collect those indicators of quality. Going from there back into, what do we have available to us in order to address that. I think you will see a very different perspective on it than the approach of trying to build it up from what we have been doing to date.
The key post-acute care issues that I just wanted to mention from the research are listed here in this first page of this handout. First of all, we and others have found considerable geographic variation in the settings that are used. That variation is definitely across the country. Also, there are community differences. This has been documented in the percent of people in the different communities that get rehab hospital care versus skilled nursing facility care. There has also been shown to be a somewhat inverse relationship in some communities between -- across communities in the amount of rehab hospital use relative to the amount of skilled nursing facility use. Interestingly, home health use often follows in line closer to rehab hospital use, but as Korbin pointed out, those are often used in conjunction.
Managed care penetration we have also found drives even the fee for service end, as to how much skilled nursing facility use there is. The communities that are very high in managed care penetration, it seems that that drives the standard of care towards skilled nursing facility as opposed to in-patient rehab hospital care.
Because managed care organizations don't use rehab hospital care for Medicare beneficiaries by and large in the work that we have seen, except in exceptional situations, that really is something that needs to be taken into consideration whenever we report these statistics that have been reported by Korbin and Barbara. Those are rates that are not obviously including managed care use and the alternatives that they are using.
The next point is that service intensity and volume of service over a period of time, as Bob was talking about, vary both within and across provider settings. That is, rehabilitation hospitals have the three hour requirement. Somebody has to be able to tolerate three hours of physical therapy services or they are not admitted to a rehabilitation hospital. Such a requirement does not exist in skilled nursing facilities. So you automatically get this distinction across provider types.
But it is not even that clean. Within skilled nursing facilities we have done some analyses where there was a vast range within diagnosis, and after adjusting for function and all sorts of other risk characteristics, in the amount of physical therapy, occupational therapy that a given individual will receive. So there is not real clear standards.
In those same models, the factors that tend to drive, be most strongly associated with how much therapy, are geographic factors, and some organizational factors.
There is also variation in how much physician care. Again, you need to think in terms of rehab hospitals. Physician payment is rolled into the payment. There is a medical director who is there, half time for a unit, full time for a separate distinct hospital. Individuals are rounded on daily or at least several times a week. Skilled nursing facilities, as you know, the minimum requirement is once a month for a skilled nursing facility. Many patients are seen much more than that if they are more subacute, but that all requires justification.
Furthermore, the mix of staff. Rehabilitation hospitals have a much higher proportion of licensed physical therapists as opposed to PT aides, much higher proportion of RNs and RNs trained in rehabilitation, in contrast to nurses' aides, that are much more common in skilled nursing facilities.
Thirdly, case mix differing across settings when overlap exists. A lot of the early work we did when we were trying to compare our settings for outcomes was to make sure we would be able to control for outcome risk factor differences, and Bob has run into this same thing.
Certainly diagnosis was not enough. We did it within diagnostic categories. Even still, function is not enough. You find some interesting issues related to function. For example, the rehabilitation preferred population doesn't tend to be the least functionally impaired or the most functionally impaired. It is this middle distribution, the middle set of function impaired. That may be the most appropriate group for them to treat, but on the other hand, that is the group that is targeted in rehabilitation hospitals. But on the other hand, that is the group that is targeted in rehabilitation hospitals, as opposed to skilled nursing facilities that might take either extreme. So those things are important as well.
Fourth, the post-acute setting frequently involves more than one provider, or the episode frequently involves more than one provider setting. Korbin and Bob have both talked about that. I want to throw in one other little piece on that. That is, when you are dealing with post-acute care by disease, you also have out patient physician care involved there. People who go into home health care are seeing a physician as well, and the physician is not seeing them as part of the home health agency; the physician is seeing them as an out patient physician. They will go see specialists following a stroke. They will go see primary care physicians following an acute event. These physicians at times will go into the facilities. At times, patients will be brought to the provider.
So within that episode, there is also the physician care that is provided that is part of post-acute care, in the sense of, it is part of care. If you are not going to define it exclusively by what the nursing home does and what the home health agency does, and you want a similar package to what is done in the rehab hospital versus what is done in the home health agency, the rehab hospital includes the physician. The home health agency, the physician is out in the community.
Point five is, we have found that settings are equally effective for some conditions while not for other conditions. This is very important. These conditions in this case were defined by diagnosis. Most of our work has been in hip fracture and stroke. We are doing some other work with medical surgical conditions after hospital stays. But the idea that it is not going to be one size fits all -- we found that there really were considerable added outcome benefits for individuals with stroke three months and six months down the line that may well warrant this higher intensity care. Whereas, for hip fracture, although you are getting higher intensity care, you may not have added benefits.
There is a range of conditions, and across all conditions you are going to find some variability. So we have to deal with this in a condition specific manner.
My sixth point -- this was before today's discussion -- hospitalization is not necessarily required for post-acute care. We have been doing a study in managed care organizations over the past several years, and have found a number of them that go emergency room to skilled nursing facility, and some that even skip the emergency room.
One of the questions is how we should deal with them, whether we ought to begin as Bob has suggested with people who have gone to the hospital may be the best way to go, but the dilemma is when you are comparing then HMO and fee for service, you are looking at two different populations, because the HMOs may not send the equivalent individual to the hospital. We have seen that in a number of those settings.
The other question is how you even deal with somebody who is in the nursing home, in the skilled nursing facility, and gets a severe nursing home acquired pneumonia that is very acute, but the physician and nursing home agree to treat there. That is a situation and you say, maybe we don't want to count those, but it is an acute event. Somebody could have gone to an ER or a hospital and then be treated.
Implications for information needs. First of all, I think unequivocally we require uniform data across settings. When we did our studies, we could not at that point -- and this work began in the '90s; we are not comfortable taking the items that had been collected in the various settings to follow people over time, for several reasons which I will get into. But it was too difficult to equate those items, and comparisons require this.
One of the major issues is this whole issue of being defined by episodes and by admission-discharge periods, and the fact that everybody had different admission-discharge periods. As I will talk about later, you really need to follow people over fixed time intervals. All of a sudden you don't have fixed time intervals as you cut across settings. So at that point, we weren't comfortable trying to combine existing data.
Furthermore, one of the things that we found and that you will see later is the idea that self report information was considered by us and by others to be critical. None of the existing systems are really individual report information, along the lines of the sickness impact profile and the SF-36 and some of these others.
That is not saying I am advocating some of those instruments for post-acute care, but as you will see, our panels also raised a great many issues about how quality of care information requires that source of data, as well as some in-facility information.
Data must be adequate to take risk factor differences into consideration. As I mentioned, there are differences across settings, although there is considerable overlap. We use various methods to define that overlap. We use propensity scores to find out which patients really didn't have a propensity to be treated in both settings, and target on a group.
On our study, because of the concern about the resistance to the results by the community, we went so far as to look at CT and MRI scans for stroke victims and pre and post fracture films for hip fractures. I guarantee you don't have to do that, and probably many of you would have known you didn't have to do that, but in order to try to adjust for risk factors in a way that would hold up among the specialty communities, we looked at those kinds of -- we went that far in terms of the acute findings.
However, within diagnosis, there are some factors beyond just function that drive outcomes. There are disease specific signs after strokes that are very predictive of outcome that aren't captured when you look at a generic risk factor model. Things like cognition and depression have to be measured very, very carefully, because those are very important predictors. Caregiver support can be key on some of these things.
Finally, there is a whole area of pre-morbid function that we found to be a very big predictor of final outcomes. As Bob has talked about, the greatest predictor of where somebody is going to be is where they were, even if they have had an acute event in the interim.
The episode really needs to be defined by a fixed interval, not admission to discharge from a setting. Effects of some of these post-acute care are not always felt at time of discharge. Somebody leaves a rehab hospital after 21 days or leaves a home health agency after 30 days, they may still be on the upward trajectory, in part because of lifestyle changes, exercises, other teachings of these settings that have led to these issues. They may be issues like whether they reside in the community at three months or six months are the kinds of outcomes that are the final benefits of rehabilitation.
So again, those are key issues. I think the last two points here are pretty well what we talked about, conditions studied separately using relevant measures and defining post-acute care by acute disease.
For this current project we were doing, we defined post-acute care and then had this hypothesis that quality indicators for post-acute care would be a blend of acute care quality indicators and the traditional long term care quality indicators. So what we did is to enumerate the four conditions, stroke, pneumonia, CHF and back problems, a set of potential quality indicators, drawing from the acute care literature, the out patient medical literature and the long term care literature, to the extent that it existed, and put these on lists of 120 or more indicators together for panels composed of a blend of people who come from very strong long term care background, representing home health, skilled nursing facilities, rehab hospitals, but also people who worked on the ports. For example, we have people for the pneumonia ports, somebody from the stroke port and so on.
So we had this real range of participants, using a modified delphi approach, where they rated them and we discussed them and rated them again.
What I am going to show you is what some of the highest rated indicators are on the next two pages for two of the conditions. I don't necessarily want to talk through these indicators now. I'm not trying to sell you on a set of indicators.
We are in the process of working on developing measures related to these indicators, seeing what can be pulled from existing instruments, but recognize that people were thinking of this as covering all post-acute care settings, measuring over fixed time intervals, because that was the theme that they agreed you needed to do to cover these episodes, and pulling from whatever data sources were necessary. As you can see, a number of these issues, quality of life satisfaction, depression, even some of the issues related to functional measurement, they believe should come from self report, and that that is the optimal appropriate data source.
Oh, and symptoms as well. As you will see for pneumonia, fatigue, dysthnea, cough, some of those symptoms, they all fell very strongly -- for back, you can well imagine, pain came in. Again, there were scales that people talked about, but self report was believed to be critical.
In addition, as you can see, the bolded ones are some of the more traditional global measures. Mortality could fall in that category. But people didn't say mortality is unimportant for post-acute care. It was key for a number of these conditions, although not all.
Some areas, they chose to argue for process measurement, saying that to look at changes in cognition can be very difficult, although we want it as a risk factor, but let's make sure they are doing what they ought to be doing to assess cognition. Because people couldn't set standards for appropriate therapy, occupational therapy and other therapy services, they said, let's make sure the evaluation is done and that it is adequate. Again, fear in some settings that those may not be conducted.
So you can see the array of the kinds of indicators that these panels came up with, this blend of acute and more traditional long term care measures that we are now working on further development.
DR. IEZZONI: Andy, thank you. That is really interesting. I'd like to lead off with a question. ASPI is funding this. Therefore, we expect that there might be some idea that what you develop might be disseminated, implemented or thought about more widely. But we also know that HCFA right now has developed OASIS. There are other methods for looking at what happens in nursing homes and rehabilitation facilities.
To what extent do you think the data systems that are being created by HCFA right now for these different settings of care really would allow you to do what you panels and you are coming up with as a comprehensive quality measurement strategy?
DR. KRAMER: This is a tricky question.
DR. IEZZONI: No, there are all sorts of euphemistic ways to raise this.
DR. KRAMER: I'm not known for being in politics, so I'm just going to come right out and say what I -- no. The reality here is, I think we could draw in certain areas from the existing instruments. There is no question. The elements that are here, the first place one would go is the existing instruments, and take what one could from those.
However, I think when you look at how you want to do this over an extended time period, where you measure intervals and you get self report information, I think there is a significant portion of it that is not incorporated in existing instruments. I don't know quite how it is going to be, given the current strategies, because those are not self report over fixed intervals; they are within provider. That is a real problem.
Eventually, the two things could come together, and I would hope that that is what might happen.
The other thing is, some of this pulls from multiple sources. That is the other dilemma; the hospitalization and utilization data would come barrelling in from claims potentially and the like, or some of it is being tracked longitudinally. But you have to make sure all of that fits together.
Our hope would be to -- in developing this instrument, is to make it the kind of instrument that could be developed along the lines of an SF-36 type instrument for post-acute care. That is not to say it would b the SF-36; it is quite different. The kinds of questions that the SF-36 asks, many of our panelists said -- in our studies of post-acute care for pneumonia, we use the SF-36. When we asked people about golfing, they say, we don't know what you are talking about. So there is no question that there are major -- it is a different world.
So we're not saying that that is what we take, but an instrument that had several of the major dimensions, had self report form, and pulled in the elements to the extent that they exist from existing instruments. Then ultimately hopefully these things could be married, so that one could proceed.
DR. IEZZONI: Can I just ask you, is your ultimate goal to specify what the instrumentation would be, so your project will end with a specific recommendation for instrumentation?
DR. KRAMER: Yes, that is what we are working on at this point, although how far it goes and how well tested it is, is still sort of not clear.
DR. IEZZONI: And how much longer is your project?
DR. KRAMER: This phase of it ends in October, and by that point, we probably won't have the whole instrument fully developed. But we are looking at ways of continuing that work.
DR. IEZZONI: Barbara?
DR. STARFIELD: I take it it is not limited to those over 65, is that right?
DR. KRAMER: No, not exactly. It is limited to people with those conditions, is basically the way we designed it.
DR. STARFIELD: I have a question for everybody. Can I ask a general question?
DR. IEZZONI: Yes.
DR. STARFIELD: Part of the work of this committee has to do with the standards in general. We have spent a fair amount of time over the last few months talking about, for example, the 2010 goals for the nation and the impetus toward equity, or reducing disparities.
DR. KRAMER: Right.
DR. STARFIELD: To what extent do the data systems you deal with focus on looking at disparities across race, ethnicities, social class? And do you see this as part of your work, any of you?
DR. KRAMER: We have looked at some of those issues in some of our data sets. Certainly, the primary data that we have collected enable us to look at those issues. I guess the biggest dilemma is, we haven't stratified on some of those issues, like ethnicity, to the point where we have large enough samples within ethnic groups to determine whether there is outcome differences. We are seeing use differences by different ethnic groups, but we don't have large enough numbers to look at outcomes within ethnic groups.
DR. STARFIELD: But it is part of your data set?
DR. KRAMER: Yes, it is part of our data set.
DR. STARFIELD: And what about all the HCFA data sets that you guys have been using?
DR. KANE: We actually use both primary data and HCFA data. I think the point that Andy is making, which is that you can detect differences in access or utilization by these characteristics, but up to now there have been no sampling strategies to look at those particular ethnic groups in sufficient numbers, because they tend to be under represented, either because they proportionately use less, or because they are proportionately less common in the population.
So if you are just taking a cohort of people who are using a particular service to try and look at the outcomes of that service, then those people by definition will be singly or doubly under represented, and you would need to do specific sampling stratification strategies to collect adequate numbers to look for those differences.
We have certainly included those variables in our analyses, and have generally not found them -- when other things are adjusted for -- to be particularly important in terms of affecting the outcomes of the studies that we have done on post-acute care. However, there is an a priori bias that they may not have had equal opportunities to get that care.
In our study, for example, where we traced people from the hospital forward, we didn't find that race -- and ours is a fairly crude measure -- that race measured as black, Hispanic, other or white, was a significant variable in determining what kind of post-acute care people got or whether they got any post-acute care, nor did we find that if you adjusted for other factors -- and here, we had this Richard Davis set with all of these functional variables -- that it was a predictor outcome.
But I can tell you, it may not have been important, and particularly for the second analysis, because there were relatively small numbers of those people, and this was not a national sample, it was in cities which may not have had equal opportunities to get large ethnic populations, that we wouldn't have found if we had gone looking for it. The studies were not set up to detect it.
DR. STARFIELD: That is what I was striving at. It wasn't so much findings, it was whether the data were available to do this sort of thing. How do you get race in Medicare?
DR. LIU: It's not in the administrative record. It is in the surveys. It is in the Medicare beneficiary survey.
DR. STARFIELD: And do you link those?
DR. LIU: No, you can't.
DR. IEZZONI: There is race data in Medicare.
DR. STARFIELD: Where in Medicare? Not in the claims data.
(Simultaneous discussion.)
DR. IEZZONI: It is in the claim, yes.
PARTICIPANT: It is, but it is not always reliable.
DR. IEZZONI: It is not reliable. Can we move on, because I want to make sure that Arthur has an opportunity to talk here. But I do think, Andy, that one of the issues that Barbara's question raises is that especially if you begin to look at things such as depression, you are going to have to be very culturally sensitive in thinking about how some of these self reported concepts are going to be viewed by different cultural groups, even things like mobility impairments.
I have been doing research on this recently. Different cultural groups have very different attitudes about mobility, even though it might seem like a very objective thing to measure. When you begin to talk about self reports and attitudes about it, there are cultural groups that view it very, very differently.
So obviously, we don't mean to pile on, because I'm sure that you are very sensitive to these issues. But I think that this is something that our subcommittee is keeping a watch on as we think about data systems that are going to be implemented nationwide.
Arthur Webb, you are standing between us and lunch, but the lunch that we are going to offer from the penthouse might not be something that -- anyway. So why don't we hear from you now, and then hopefully we might have a few minutes.
DR. WEBB: Thank you very much. I am Arthur Webb. I am the president and CEO of Village Care of New York, which is a community based not for profit organization in the city of New York, headquartered in West Village.
We serve about 900 AIDS patients every day in seven different settings. We are probably one of the largest providers of AIDS services, clinic care services for people living with AIDS in the country. We also serve close to 300 individuals who are frail and geriatric in a variety of settings.
We have a geriatric nursing home, a day programs home care community outreach, meal support, a whole variety of things. We are in the midst of having St. Luke's Hospital -- we are right in the environment where St. Luke's Hospital, St. Vincent's Hospital, NYU are converging together into our market.
It is good actually to be here today, because I thought I was the only one frustrated by this issue; obviously not. But on the other side of this, nothing is new. This has been going on for a long time. In fact, if you look at the (word lost) report on subacute care, I think it was '95 or '96 perhaps, that is essentially what he said, and that study put a little bit of a damper on this huge initiative, at least in New York, where hospitals and others, big large for-profit companies coming in, try and establish this level of care called subacute, and hopefully drive some reimbursement to do that. That has been put on hold, and there is a whole task force on this whole issue of post acute.
Our goal is really kind of frustrated by what has already been said. In fact, I was going to ask Bob if I could borrow his slides for my presentation, because indeed, all of what he said and Andy and Korbin have said is exactly what we are faced with.
There is no system. The service delivery system is terribly fragmented from a data point of view and a service, and at best it is maybe a handoff, like in football you don't fumble. The hospital data we get on individuals being discharged is pitiful. It is really shameful. You can't use it at all. We actually have to send over a very skilled clinician to actually do case review in order to understand what is going on. We have a whole series of issues here, comorbidity and the living situation and all that, and hospitals just don't know. They have no clue how to provide care after the hospital after someone is rolled out the door.
However, on the other hand, there is incredible competition for those post-acute care folks, hospitals, nursing homes, doctors, specialty groups coming in, and everybody is trying to pick this off.
Managed care has really disrupted any kind of system development. They have not added any value at all. At best, they are arguing and managing against price, certainly not service integration. So consequently in my mind, there is no rhyme or reason to anything that is going on, and that is true for us.
We have broken in a very crude way -- and I am going to speak in shorthand, if you don't mind, I know this is being recorded and all those folks in the Internet hopefully are not turning red. But anyway, crudely, we have three groups of individuals, and I'll tell you which ones we serve. We have folks coming in the hospital, in and out, regardless of what age, who almost disappear from the screen. Then there are folks who go into the hospital and oops, a complication happens. All of a sudden, other issues start to compound, length of stays increase, and the acute issue has disappeared. Now they are into some other kind of medical management, and it really complicates matters when you are trying to do a post-acute review.
Then there is a whole group of individuals who are chronically ill anyway, due to heart, cancer or stroke. If you did a course of an illness, this might be just one of the blips in this kind of periodicity of acute care, and we can manage them anyway over time.
Then thee is a whole series of things where they might have gone in for a particular issue, and then a whole series of other things start to happen in terms of depression and it turns into a major asthma issue and it turns into a bronchial case when it wasn't a pulmonary case in the first place at all.
Then I think the question that Dr. Starfield asked about death, there is data on Medicare and hospice, and I'm surprised that is not built into any of this analysis. We believe death is a major predictor. In fact, much of what we do post acute quite honestly is managing death, and managing all the issues related to death. For some reason, we don't talk a whole lot about that stuff, either in data or in these discussions, but that is who we are.
Interestingly enough, in our two groups of individuals we serve, the geriatric and our AIDS populations, in many ways they are very similar because we are dealing in many cases of chronic illness, but they are very different. The average age of the person we serve in our AIDS system -- and we built our own network because no one else is incentivized enough to do anything along those lines. In our network, the average age is 41. Many of them have had close to 20 years that in some cases they have carried the virus, only been symptomatic for the last four or five years. Seventy-five percent come in with a history of substance abuse. About 82 percent come in with a history of homelessness, and they are really sick, I've got to tell you.
However, miracles of miracles, and this country really hasn't taken the time to applaud the great research that has gone on, and the new drugs we are using, the new combination therapies are absolutely close to a miracle. We have people now who are walking out and going home and living. It adds new dimensions to our problems, but nevertheless it is some wonderful and fabulous news.
So we have had to change from -- we have been running a pretty acute care model, both from our home health and in our nursing home, and we have actually flown right past chronic care and now we're into trying to figure out how to manage on an ambulatory care basis, many, many individuals that we would not have served in the past. In fact, right now, death is only 16 percent cause of our discharge; two years ago it was 60 percent. It has radically changed the way we go about our business, and that is the good news.
It is very difficult to change a system once you have built it for one model. And again, there are no incentives for us to do any of that.
We have been trying to build a database system using the existing format and intent of both the MDS and OASIS. I would just suggest we don't add any more. It is hard enough. I can't even tell you how hard it is to take a system and build a database and begin to use it. We are not there yet. The MDS is robust, but we have no quality indicators that go with it, they are not standardized in any way, and so we have some real difficulties when we try to take all of this data and try to apply it to clinical outcomes.
There are virtually no protocols. Even if you look at the HETUS stuff, I think there is only one section just dealing with some of the outcome measures, and it is just a terrible appalling lack of development in terms of protocols and outcome efforts.
So what do we do? What do we rely upon in terms of what data do we use to try to help manage our quality and to try to play in this world of being a credible, effective competitor, because nursing homes are no longer -- you just build it and they come. That is not the way life is anymore. You have to be smart and a whole lot better than ever before.
So what do we do? We essentially use accepted professional standards, and that is just about it. If PT comes in and the doctors come in, we essentially accept the professional standards of each of those disciplines. That may not seem much, but it is a lot, because there are some accepted professional standards.
Another goal that we have is to get people home. Another goal we have, and we have some evidence to show that our interventions have stopped rehospitalizations -- and I haven't had a chance to look at all the data that was presented, but there is some analysis that shows that individuals -- Medicare and Medicare data, shows the number of rehospitalizations for those coming out on orthopedic, there is some high degree of evidence of, within nine months someone will be rehospitalized, frequently due to a variety of different things, but falls as an example is frequently a cause in this particular situation.
What we have tried to do is prevent some of those things that would cause rehospitalizations. Example, working with another community-based group, we are doing home repair, we are doing re-site, we are doing visiting neighbors -- simple little things that Medicare doesn't even come close to paying for and will never pay for, and maybe they shouldn't, I don't know. Medicaid may be a little bit.
Another goal is to improve functional status and independence, so that individuals can live pretty decent lives. So our goal is not to fill the beds at nursing homes, our goal is to try to get people home, and those are the -- when you put all that together, that is our effort.
We have some work going on in terms of using MDS in the nursing home side to help us manage some kinds of outcomes. That MDS data has a long way to go. It is very, very weak in the mental health piece, and really trying to help us think about depression.
Again, we find in our care across the board, when you are managing chronic illness pain management and depression, those things are frequently not diagnosed.
Patients are amazing. What they are able to absorb and their level of tolerance is remarkable. I guess they have never been told that you should complain a little bit. So the other part of what we are trying to do is bring, believe it or not, the patient back into the center of our approach to care, to see if we can't use some of that information to help us drive and think about not only issues of depression, but also pain management.
So our goal is to find ways to manage post-acute. We don't see again hospitals as the event that triggers our response, quite honestly. It just happens to be a blip in many cases, but nevertheless be able to manage across time, place and discipline. That is very, very hard. There are very few models in the country that are able to think about this and effectively structure the financial wherewithal to be able to manage the particular individual and help them manage their illness so they can manage their lives.
However, I've got to tell you, we have gotten pretty good at mixing and matching Medicare and Medicaid. Some cases I probably shouldn't tell you, but nothing is wrong with it; it is just that it gets petty confusing.
As an example, when we want to reduce the intensity of service but improve the quality of life aspects, we use our licensed agencies, who are highly trained individuals, our home health and housekeeper folks, who can actually help deal with some of the social aspects of much of this.
We look at the post-acute as about a 100 day period of time. Now, there is a reason for that in the hospital side, because that is what Medicare will pay. But nevertheless, many of these individuals who come to us already have used up their 100 days. It is not 100 days, it is 100 days over the life period. If you think of chronic illness and periodicity, many of these individuals have already taken a good deal of their days. So in some cases, they get in there -- the frustrating part is getting people eligible for Medicaid, and that is not easy these days.
So we have tried to patch together a system of care that allows us to vary the intensity of support, depending upon the needs of the individual. We continue to do this. We took the APS study, we took the work from VNCA, et cetera. We took the '99 DRGs that APT did in terms of post-acute care, those most likely to benefit from post-acute care, and we did our own model. My director of research if he were here could probably explain it. I can tell you from a management point of view what does this really mean.
It took seven specialties. We did some crosswalks with your work. It is interesting; it really matches a lot of what we do. In our seven specialties and 66 DRGs, we predicted with a high degree of predictability. In working with our ascending hospitals -- that is why I mentioned those in the first place, we have gone to them and shown them their own data. We have data on every hospital and every discharge, and Medicaid, Medicare and private pay in our database. We are a teeny little provider.
But anyway, we use our model, and we went to them and said, look, we have certain capabilities. We can't do everything, but we have certain capabilities for the geriatric population and we have certain capabilities for the AIDS populations. I can show you right away how you can save money and not be panelized against your DRG penalties. They couldn't believe it. We even gave them five days over their DRG average, and for instance when you look at the DRUG-210, which is your basic hip fracture, we in particular saved one hospital 300 days of penalties. That is a lot of money in that game.
So part of what we are doing in quality care is trying to make sure that we really understand our own capabilities in both our studies for different populations. That is the hardest thing to do for us. We beefed up our rehabilitation, we have done a variety of different things.
The most interesting thing that you will find in a post-acute are chronic care nursing homes residents -- let me start over. Our subacute persons coming in want to be called patients, not residents. They want to be on a separate floor, they don't want to eat with the clinic care folks. So we have had to totally remodel one of our floors. It is now a subacute floor. We have taken all of our nurses, we have had a massive culture change with our aides and with our nurses and our physical therapists. We have had to go through some very interesting training, some skill based, to be able to maximize the capability so we can get people out.
We have now moved from our short term, from about 90 days when we started, down to about 42 days, and our goal is to get down to about 25 days. That is very hard in the nursing homes that are traditionally chronic care entities. They just don't know how to deal with the concept of short term rehabilitation, in and out. Numerous complications in terms of service delivery and quality care management.
In the AIDS world, it is similar, but I'll give you an example. We have a highly specialized subacute, where we use critical care nurses from hospitals. We are serving maybe seven to eight hours of skilled nursing care a day per person. These are very, very complicated individuals. There, we do soup to nuts. We do room management to very intensive rehab.
So part of the post-acute is having our home care entity, our nursing home really understand what the limits of the capability are, and having some ability to feed back that what they are doing has some immediate and direct benefit for the individuals they are serving. That cuts across both the medical, social and psychological aspects of that care. That is what, at least from one provider, we are trying to do to be able to compete quite honestly in an incredibly charged, wild west in the city of New York.
The other part of it though is that hospitals still dominate the landscape. They still drive the definition of what quote health care is, which is really unfortunate, and everything they do is to maximize in-patient days, not necessarily systems of care.
We have gone to them and we are trying four or five different models all simultaneously. As an example, we are taking them some of the most complicated individuals coming from one hospital, and we are assigning a nurse practitioner to those to monitor and manage their care across each of the different systems in the care management model. We are paying for ourselves, and right now we are just looking at 10 individuals.
Some very interesting things are starting to happen. The education of the docs when they go back into the community. The primary care practitioners when they go back into the community have no clue, it is really sad. Not that they should, but many of the primary care practitioners are not very good at managing depression and pain management. So having this nurse practitioner now allows them to be much more engaged in some quality care kinds of things, both from a social and a medical point of view.
So it is a teeny little model; I don't know how far we're going to go. There has been some stuff in the literature that talks about some of these things.
From a provider perspective, you are onto I think one of the more important issues that Medicare needs to face, and get away from this idea that systems should be driven by setting, and data should be driven by patient needs and care costs. If you ever pull that off in our lifetime, that would be a remarkable achievement.
DR. IEZZONI: I cannot promise that we will, but I thank you for a refreshing and insightful and reality based presentation. Do any of the committee members have any quick questions? That was very helpful. Michael?
DR. MILLMAN: We really heard a lot about structural characteristics, but not much about the data availability, particularly staffing patterns. Has anybody given much thought to what is available, how good is the quality of that data, and what we ought to be focusing on in terms of the structural characteristics of the providers and the organizations and their relationship?
DR. WEBB: I'm involved in another effort to try to get to some of these issues about staffing and facility characteristics, whether or not it has any relationship to quality, outcome or access. I'm not a researcher, so I'll give you my layperson's opinion.
At least what I have been presented, there is no data that gives me any kind of confidence that indeed we know what those structural characteristics are and how they vary against particular clinical outcomes and how those ratio patterns might change, either at a staffing level or types of staffing or replacement of staffing or whatever comes, and how that would change in terms of managing care.
There is obviously some data that shows that some level of staffing has some relationship to quality. What that level is, I'm not really sure.
DR. IEZZONI: Bob, do you want to comment on that?
DR. KANE: From a methodological standpoint, the problem you have is using large systematic data. A, the data isn't there, but if it were there, you would have huge problems with confounding.
I think the question you are getting at is a very important question: how much of what we insist on and in some cases cause to be by regulation -- Andy referred to the three hour rule and some of the requirements for psyiatrists or medical directors and things like that, the use of registered nurses. There is a whole area of nurse delegation in nursing homes. There are a number of these issues that I think are very important to pursue.
They are not likely to be easily pursued by extant data sets. It is going to require some more innovative programs, either to get the waivers to allow them to happen in the first place, or to isolate those events and those things in the context of other things that are going on.
Rehabilitation is a very complex package. There is a lot of stuff. The number of variations you would want to test is probably larger than what you could do with a systematic data set.
So I think we need to know much more about that. I don't think from this committee's purview that is going to happen by mandating more reports on staffing or organizations.
DR. IEZZONI: Andy, do you want to comment?
DR. KRAMER: A really quick point. The biggest dilemma is the way the staffing data are collected across facilities. They are not easy to desegregate to the individual level. So HCFA has done quite a bit of time study work to set rates on nursing care and therapy care, and they have done quite a bit of time study work.
We have done some time study work as well, looking at each type of therapy and different RN levels. One of the things that we have found is some upside-down levelling off distributions of the kind you would expect. Up until a certain point, physical therapy seems to be associated with enhanced outcome. Then there is this levelling off point above which you no longer see any improvements in functioning.
But as Bob mentioned, the confounding is what is so difficult in terms of how you adjust for the fact that certain people are going to get more therapy because of their characteristics.
DR. IEZZONI: Vince?
DR. MOR: Art, you mentioned you were trying to deal with OASIS and MDS, the crosswalking, which was an interesting challenge. Andy, you mentioned that you are trying to deal with common items across settings. I've been worrying about that for some time. I'd like both of your comments, or maybe your practical experience and, Andy, how you have been thinking about it. Does the same data item mean the same thing when it is measured in two radically different settings?
DR. WEBB: Short answer, no, because in the same field you can do an interesting match. We are trying to do this ourselves, and we are working -- and I can't say who this is, but we are working with a major national company as a design partner to do this thing.
When you look at OASIS and MDS, similar kinds of words, similar kinds of fields. But how do you go about that in terms of standardization of that assessment capability? It is not standardized yet, and it should be, especially if you're going to build a system, which is what we're trying to do. But we are a long, long way away.
Even after you have done that and you aggregate it all together, a person in home care looking the same as a person in my nursing home and the same as in my day program, it is not the same.
DR. KRAMER: You understand the rub clearly. When you are looking for very gross -- the original CAD scale, which is independent-dependent, you can find those independent-dependent in 14 different ways. But the dilemma is, when you try to get enough precision that you actually are sensitive to outcome differences over time, we found that we couldn't just throw one scale in for another scale, that we needed consistency.
So the dilemma we are up against is, okay, if we start from the quality measure point of view, we go, these are the quality measures we think really ought to be there. Then the battle is, which data items. Obviously you first turn to all the extant data items, to the extent they are available, and try to figure out some way to pull things from those extant data items.
DR. MOR: But even forgetting any extant data items, if you had the ideal, whatever it is, would it mean the same thing in different settings?
DR. KRAMER: I see. If you took the exact same functional measure and you used it in home care and you use it -- yes.
DR. MOR: Does it mean the same thing (words lost) of administration?
DR. KRAMER: Yes, one of the dilemmas there is, you get the provider differences in how they assess these things. Part of the reason I come back to self report for some types of items -- and again, I would argue that not everything should be done self report. But for some of these kinds of things, yes, there are ethnic biases and issues of individual subjectivity, but you don't have the provider views that influence, as you point out, a scale where it almost has to be worded differently for some of these settings.
DR. WEBB: But on the other hand though, the two predictors of variability, of intensity -- I don't know about outcomes -- are one, the functional level, and the other one is depression, no matter what the presenting issue is. When we measure functionality in terms of ADL across settings, they look the same.
However, when you go to ambulation at home, can you ambulate enough to get your groceries? Nursing home, can you ambulate in terms of self care and in terms of personal hygiene? A very, very different set of --
DR. MOR: That is the demand context. That is sort of what I was getting at.
DR. KANE: This is a fundamental problem of measurement. Context changes measurement, order effects change measurement, respondents change measurement. These are not new problems. I think the issue is the noise-signal ratio question. You are going to get some differences.
Right now, we are sitting in the middle of a desert, if I could use the OASIS model. We are debating whether figs and dates have the same nutritive value. The answer is, nobody is eating anything right now. I'd like to take a first step, and then down the road we'll have an esoteric conference on what we are doing to distort those measures a little bit.
So the long story from your perspective, we should do something. I think we ought to be thoughtful in what we're doing. We need to have consistent measures. I would rather use a consistent measure. If we had time, I would come back to your whole issue about the ethnic bias of measurement and whether you adjust for that on the right-hand side or change it on the left-hand side. I would not change it on the left-hand side. I think those are things you adjust in your interpretation. But if you start trying to get measures that capture each nuance of each respondent, you wind up with the Tower of Babel in measurement.
I think these are issues we know how to deal with. I think we have much bigger problems. I think what Andy is saying is that there are important domains that are simply not addressed by the limitations we have placed on measurement today, that are fundamental characteristics both for risk adjustment as well as for outcomes. I think that is a big ticket issue, and we need to think about how to do that. We want to do that consistently across the various sites.
Then down the road, I am willing to go into the next level of, how do those change by the a priori sets of what people have, of what their expected value responses are going to be, and how does that change the response patterns. But that is a second order question. We have to get the first one done, I think.
DR. IEZZONI: Thank you. I really hate to stop conversation. I know that the gentleman back here has a question for Korbin. Maybe you could ask him privately afterwards, because I just really feel like we need to break now. I think I see people needing to get to lunch and to the restroom and so on.
The lunchroom is upstairs. It is called the Penthouse. If people would mind bringing their lunches back in here so we can start around 1:10 or so, so we can try to break on time, because I know people have flights leaving around 5 o'clock. Thank you.
(The meeting recessed for lunch at 12:42 p.m., to reconvene at 1:10 p.m.)
DR. IEZZONI: Because we had such a good time this morning, we want to make sure that we have all the time this afternoon to hear from the speakers.
We might go in a different order of the speakers, because our first speaker isn't here. So Sue Nonemaker?
PARTICIPANT: (Comments off mike.)
DR. IEZZONI: We wanted to have a comment though at the beginning of the afternoon, because many of you might have had a question as to why nursing homes costs skyrocketed in 1990. We have a gentleman here who has an explanation for us. Could you introduce yourself?
MR. CLARK: Yes, I'm Gene Clark. I am with Beverly Enterprises, in charge of quality management for the company. Prior to that, I was with HCFA for quite a few years, so an interesting background, I guess.
What I wanted to point out was pertinent to a chart that Korbin had used. That was the one that spoke to Medicare skilled nursing facility payments over the years 1986 to 1996. And basically, if you haven't got it in front of you, it goes along almost relatively flat with some escalation until 1990, and all of a sudden in 1990 almost takes off vertical.
I believe that one of the major reasons if not the reason why that happened was because of the implementation of the OBRA statute, which I'm sure you will remember went into effect in October of 1990. It specifically had a requirement that called for providers to meet the highest practicable level of functioning for the residents.
I will tell you from experience that that has about 7,000 different interpretations across the country by way of 7,000 surveyors who apply it, but it has had a major impact on the number and level of services that have been provided to nursing home residents since October of 1990.
One little anecdote and then I'll sit down or go back to my place. We had a facility wherein we had a resident whose care plan required that she ambulate 100 feet per day. During a survey, she happened to ambulate 120 feet on a given day. We were cited with failure to achieve substantial compliance, and the facility was terminated because we failed as a result of that particular day's ambulation to change the care plan and the highest practicable level of expectation to at least 120 feet. Actual case.
Thank you.
DR. IEZZONI: Well, that is a sobering note, but thank you for that insight. That is helpful to help us understand why those costs changed.
While we are waiting for -- oh good, here is Henry. Henry, sorry to rush you, but we have limited time. I have already heard from three committee members that they have flights early this afternoon. So I wanted to get started. So Dr. Henry Ireys from Johns Hopkins will talk to us next.
DR. IREYS: I have some overheads. Is there a facility?
DR. IEZZONI: Yes, there is, but Henry, you weren't here this morning. We are being broadcast on the Internet, so you have to be wired for sound the entire time.
DR. IREYS: Well, I am delighted to be here, although I must say, it feels akin to bringing coals to Newcastle, since you have in your midst Paul Newacheck, who probably knows more about the epidemiology of children who need long term care than anybody else in the country. So I will try to get on base, and he can hit cleanup and bring me home.
I want to talk in the relatively brief time -- I believe I was asked to speak for about 15 minutes, and then field some questions -- about a number of key issues. Most of you should have a handout which essentially replicates the points I'm going to make.
I'm going to just talk a little bit about the population definition of children of special health care needs, many of whom have long term care needs. I want to speak a little bit about current sources of epidemiologic data that are now available for information about these kids. I also want to talk a little bit about the characteristics of kids. It is very different from an adult population. I want to underscore some of those differences, because we really can't capture the sense of the data that we need unless we really understand some important characteristics about the population.
Then I am going to just comment on some of the critical data needs that I think are important, and then summarize a bit. These comments are designed to be informal. Please raise your hands, interrupt, bang on the table. I'll be happy to stop and we can have a conversation about any of these.
Let me talk first a little bit about how this population has been defined. This has really been quite an important issue in the field and among the various folks who are concerned about children with disabilities, children with long term care needs. I just want to highlight some of the key concepts in defining this population for epidemiologic purposes.
There have been a couple of traditional approaches that have been used. One of them is to define the population by using a list based approach, based on diagnoses or ICD-9 codes. So for example you can say, we are going to define this population by the conditions it has, cerebral palsy, spina bifida, a whole lot of different inborn errors of metabolism, congenital defects. There is a whole list of diagnoses that yield long term care needs.
Well, that makes a lot of sense and is easily understood, but there is some drawbacks to that approach because there are over 200 different chronic ongoing serious health conditions of childhood, and it is difficult to encompass all 200 in a particular list approach to gathering data. You can't interview a lot of people in all of those 200 different conditions. Some of the most rare are also the most expensive and have the most implications of the service system and for long term care needs.
Coming at it from a diagnostic point of view, it also means that you are counting conditions, not children. There are many children who have multiple conditions, and in fact they are often the ones who are the most severely compromised. And ultimately we want to count kids and understand their needs, not necessarily conditions.
Another traditional approach has been to focus on children with activity limitations, kids who can't dress themselves or feed themselves or in some cases, less severe activity limitations, can't necessarily participate in school in the same way that other kids can. This has been frequently used approach to defining the population, so we have some trend data, items related to this definition have been included in previous and repeated NHIS surveys, especially when there have been some child health supplements, so we have some trend data.
The problem with this is that it is really quite a narrow focus. There are many children who have long term care needs who are not necessarily limited in their activities. For example, children who have diabetes or who have hemophilia or sickle cell anemia certainly have long term -- those conditions are going to be with them for the rest of their lives. They are not necessarily limited in their activity. So that approach also has some drawbacks.
Recently, the field has really come together in defining the population from a non-categorical functional approach, saying what is most important is to look at not only whether there is a chronic ongoing condition, whether it is physical, developmental, behavioral or emotional condition, but also whether that condition has important consequences, important for the family or society or for the health care system.
So based on a lot of work that the Bureau of Maternal and Child Health funded and supported and also work done at Albert Einstein led by Ruth Stein and work that Paul has been involved in right from the get-go, there has been a focus on defining the population as children who have a condition that is either behavioral, developmental, emotional or physical, which also increases the need for more than the usual amount of services, so it heightens the need for services above age appropriate norms, that limit function, picking up on the notion of functionality, and that leads to dependency on compensatory mechanisms.
So for example, children who may need insulin shots or an insulin pump function fine. They would never be picked up as having necessarily a problem. They may not even have a heightened need for service, if their pediatrician can manage the condition on a regular basis once a year. But they are dependent on that insulin pump. If they don't have it, they are in big trouble. It is really important to get that element of the population in, because it certainly speaks to the long term needs that they will have.
Let me just mention the major sources of epidemiologic data that are now contributing to the knowledge base about this population. I suspect most of you or at least many of you are familiar with the 1994-95 childhood disability supplement of the national health interview survey, certainly a product of many peoples' involvement. It really has provided for the first time in this country a base of information about individuals with disability, including children with disability.
Rather than going into the nature and extent of that, we can deal with that through questions. But it is certainly a very, very important source of data, rapidly aging, however, in light of the many changes in the health care system since '94 and '95.
Anyway, in 1998, NHIS had a childhood supplement which has been the source of a lot of analyses and a lot of information about this population. There have been a number of local area child health surveys. Rochester had a very famous one in the '70s, Ontario had a survey in the '80s that really yielded important information about prevalence rates and needs for services.
But again, there are some limitations in those kinds of surveys. There are many program based surveys available. There was a comprehensive assessment of health needs of children in special education in the early '80s. SSI programmed a recent publication on patterns of service use and costs in SSI. Then there is of course encounter data that is being collected by both Medicaid and also managed care organizations now, whether it is in the commercial or public sector.
All of these have their strengths. All of them also have many weaknesses, whether it is an attenuated range of variables, inappropriate sampling, unrepresentative sampling or systematic ignoring of key populations.
So while we have some good data, there is an awful lot that is incomplete and we need a lot more, which I am going to get to in just a minute.
Let me talk about what we do know very briefly about the population characteristics of this group. About 15 to 18 percent of all children have some kind of special health need or chronic ongoing condition. About six to seven percent have some kind of activity limitation in age appropriate activity, and about two or three percent have functionally severe conditions, depending on how you define severity.
I mention -- these data really come from recent analyses of the national health interview survey disability supplement. The Maternal and Child Health Bureau is funding a chart book that will hopefully be ready in 12 months, if we keep our fingers crossed. All are very much involved in this project. That will produce for wide distribution the data on this population from the NHIS disability supplement.
We know already that there are higher rates for children in poverty in most of these prevalence estimations. We also know that -- I just want to underscore the importance of this diagnostic heterogeneity. If you think about adults or elderly who have long term care needs, you really consider about 10 different conditions that contribute to that. There is Alzheimer's and diabetes, the orthopedic problems of the elderly. There is really only about -- and cancer. There is really only about 10 different conditions, most of which have relatively high prevalence rates.
In children, it is a little bit reversed. There are about 200 more different chronic health conditions, and with the exception of asthma, which is relatively highly prevalent, most of these chronic conditions are really relatively rare.
So I think it is important to keep some of the distinctions in mind in terms of the differences between adults and children in the kinds of conditions that make up the population.
A couple of other points I want to make, really focusing on developmental characteristics. Again, this is a population of children who are growing, so they have a lifelong risk. Again, an adult or an elderly person who develops a chronic health problem may live for five, 10, 15, 20 years. A child who has a chronic health problem will live his or her lifetime with that, in many instances.
Also, children are naturally dependent on family members and communities. The presence of a long term health condition is influenced by the health status of other family members. So we really have to think of these children in that context.
Also, we have very little information about the complex interactions between growth, development and health status. For a simple example, a child who is in a wheelchair at age five, by age six, even though the wheelchair may be functioning fine, that child cannot use that wheelchair anymore because he or she has grown out of it. Yet, it many instances, the insurers will not pay for another wheelchair because companies often only pay for a wheelchair every three years, in case they break down. But it is the realities of development and growth that often power this long term health care need of children who have chronic health care problems. It is an important factor to keep in mind.
A couple of other general characteristics. I'm going to skip that, that is listed in your handout. For the sake of the conversation, I'll move on.
I want to just talk about the data needs as I see them, at least some of them. First of all, picking up on the developmental issues, one of the most glaring absences of information is longitudinal data. Very few of those epidemiological sources I listed before have been replicated in a way that we can really track individuals over time in terms of how their needs change.
We have relatively little knowledge about how development itself affects health status. So longitudinal data of repeated surveys over time would be very important.
Also, we have very little data on how families respond to and are influenced by a long term health condition. I don't think we will really understand the impact of children's chronic conditions or long term health care needs if we don't understand also how the family responds to that and is influenced by that condition.
Also, virtually all of those epidemiological data sources that I mentioned before are based on non-institutionalized populations. So we have virtually no data on the various institutionalized subgroups, and there are a couple of prominent ones. For example, infants who are born at very low birth rates and who may be for other congenital reasons dependent on medical technology, may need to be in nursing homes if families or communities are unable to support them. We don't really know much about that group, and yet they are very costly, and presumably are at high risk for lousy outcomes.
We also don't know much about teenagers who have serious emotional disorders, many of whom are also institutionalized. Again, I think that is a glaring data need.
Let me just conclude by getting on the soapbox. I think it is really important to have better data for a number of different reasons. One of them is -- not surprisingly, I'm sure, to most of the folks in here -- if we have good data, we reduce the risk for poor outcomes. If we know where the problems are, we can begin to intervene or create service systems that really address them.
Also, I think at the moment, the service system has really been driven by advocacy efforts, focused on a particular problem or a particular subgroup. That is not bad. We have had some good outcomes as a result of those advocacy efforts, but it has yielded a service system that is quite uncoordinated and piecemeal. If we have better data, I think we can work with the advocacy organizations to argue for a much more coherent and logical service system.
Finally, I think ultimately, children who have long term care needs are a very vulnerable group of children. In fact, almost by definition, children are supposed to be healthy and happy and active and have a long life of health in front of them. These kids violate all of those assumptions. I think in many cases, it makes people quite uncomfortable thinking about how to respond in a compassionate and effective way to these kids. So I think if we have better data on the nature of this population and their outcomes, we will be able to contribute to the debate about how society wants to support and value these children in a way that hasn't been possible.
I'm sure there are some questions.
DR. IEZZONI: Henry, why don't you take a seat? Thank you.
DR. IREYS: Sure.
DR. IEZZONI: Paul?
DR. NEWACHECK: I had a question. Thank you, Henry, that was a very useful and interesting presentation. I wanted to ask you about prospective payment for these kids in managed care. More and more of them are going into managed care, although in the past they have been able to stay out. But it does seem to be the trend now both in the public sector and in the private sector.
I am wondering if you can comment, because I know you have done some work in this area, about where we stand with regard to developing risk adjustment techniques for this population, whether we need to do more, or whether we need other data in that area, to insure that these children are being capitated at the appropriate rates.
DR. IREYS: Right. Well, it's a good question. As I'm sure you all know, the issue of capitation for populations that ar at risk for long term care and who are likely to be very expensive is a pretty critical one. Many of the existing risk adjustment strategies are based on adult populations. There has been with one or two exceptions relatively few of the existing risk adjustment methods that have actively incorporated children with disabilities.
One of the strategies that we are doing at Hopkins, as you know, is trying to evaluate the different risk adjustment methods out there in terms of how well they do in predicting costs for this particular population. We are coming to an end on that study, and the results should be available pretty soon.
In addition, there has been other work. The National Association of Children's Hospitals and related institutions, NACHRI, has developed essentially a risk adjustment system that is probably within the next 12 months going to be actively marketed. They feel that that is a very good system.
So I think overall, we are at the beginning of understanding how to do risk adjustment better for these children, but in terms of getting stuff working in the field and figuring out how it is going to work, getting managed care organizations to adopt some of these special procedures for this population, I think we have a long way to go.
DR. IEZZONI: Barbara?
DR. STARFIELD: Henry, thanks, that was nice. While I understand why you would adopt the perspective that children are different and that we look at things differently because children are different, I think from the overall point of view of data strategy, it is probably useful to try to find commonalities with adults, because a lot of the allies of children have to be adults.
So there are a couple of things that I can think of. First of all, this notion of comorbidity. It isn't really true that adults have less heterogeneity, and they have at least as much comorbidity. So that is an area of commonality, how you deal with that.
The second thing is the mandate of the continuum of care. We are not focusing on particular settings or sites, but focusing on the needs and how we can get data that cross different kinds of facilities. Do you have any insights on how we might form allegiances with the adults?
DR. IREYS: Well, in terms of getting data across service sites? Well, I'm not sure I can speak to the similarities between kids and adults, although obviously there are some, and those are important. But I think that the challenge of gathering data from multiple service systems and sites is a major one. Kids illustrate that, so do adults.
And managed care organizations for example are gathering data about kids with disabilities, but they are not really interested in linking any of that data to the school system, where the school system deals with a lot of kids who have long term care needs. In fact, it is in the context of the school system where a lot of services for those long term care needs are provided.
So I think one of the major strategies is trying to figure out how to bring those two different data sources together in kids. i know there are equivalent problems for adults and elderly as well.
DR. IEZZONI: I'd like to build off of Barbara's question, to ask you whether you think ICD-9CM adequately reflects the diagnoses of children. Then the second question, as we were talking this morning -- Henry, you were unfortunately unable to hear the discussion about measuring functional status, and the fact that we were talking primarily about adults this morning probably, and that functional status measures measured in a nursing home setting might be different than might be different than measured in a home health setting because of contextual issues.
So my second question for you is, might an ADL type of scale for a child look very different than for an adult?
DR. IREYS: Right.
DR. IEZZONI: So if we wanted to have a functional measure placeholder in a data set, would that have to look very different for a kid, and how would you define a child in that kind of context? Because there are probably a variety of different age cutoffs that you have to look at in that.
DR. IREYS: I think that is where the developmental issue comes in, because what you would expect for a two year old is really very different than what you would expect for a 10 year old. The items that you would want to include in a data set to assess functionality -- and again, Paul has struggled with this for a long time -- would have to be different, because your basis of what is appropriate is different.
It is different also across different areas of the child's development. Physical development is separate from language development, and there are different norms that also have to be taken into account in terms of the different areas of development that are going on.
So I think that it is a formidable task. There is so much debate in the field about how do you assess functionality of children at different ages, because we really don't know quite yet how to do that. In certain areas, I think we have more progress than in other areas, but as a whole I think we have a long way to go.
The national health interview survey disability supplement probably represents fairly good state of the art, and even that has problems. Many people feel that that really didn't capture all the items well enough.
DR. IEZZONI: How about the ICD-9CM question?
DR. IREYS: Right. I was just yesterday talking to a pediatric neurologist who was in a developmental clinic, the Kennedy Krieger Institute in Baltimore. He takes care of a lot of kids with congenital problems. He was pointing out that for many of the rare congenital problems, there isn't even a diagnosis made for an extended period of time, sometimes two or three years, before the diagnosis really can be made.
Now, that is not a lot of kids, but the fact is that there are some rare developmental genetic problems that we are now first of all detecting better than we have in the past, and second of all, don't necessarily have a good label for. So I think that illustrates --
DR. IEZZONI: What does he use for his billing code?
DR. IREYS: I think they do two things. They try to approximate it as much as possible, and they get on the phone and start haggling with the state Medicaid, or now with the managed care organization. Every kid in some ways is like an auction from their point of view, in terms of trying to figure out who is going to pay for what, and under what billing code or procedure code.
So we can get into a lot more about the service system and who pays for what, but I think that is an example of how the ICD-9 codes are not wholly -- don't really wholly capture this population.
DR. IEZZONI: That is helpful to know. I just wanted to mention something. I was glad that you brought up what you called compensatory services, for example, the insulin pump and various other assisted technologies. We haven't talked about that at all this morning. But I think especially when you think about a population with disabilities, you do need to think in terms of the assistive devices and technologies. We don't really have a great way to capture a lot of that information, I don't think, currently, and how that affects peoples' ability to function in the day to day world, not just whether the muscle mass is firing equally or has equal strength.
Aaron, did you have a question?
DR. HANDLER: Yes. There is a Dr. George Brenneman who is on the staff of the American Indian Health Project at Johns Hopkins University, who was formerly the maternal and child health officer for the Indian Health Service. I don't know if you have been dealing with him or have met him?
DR. IREYS: No, I know George, yes.
DR. HANDLER: You might coordinate activities with him to some extent to try to get longitudinal data on American Indian children with disabilities, using our patient management system that the Indian Health Service runs. He probably would like to help you with that and work with our people. He knows all of the inside people that could help you with that, as well as comprehensive family based data. We see all members of the family, not just the children.
That is a good data source to find out what you can from our health data and possibly use that as an example to go to other populations. We have a database that can be used for that purpose. George would be the contact person. He also works very closely with the former Indian Health Service director, Dr. Everett Rhodes. He is in Oklahoma, but every now and then he comes down here as well.
DR. IREYS: I think that is a great idea. In fact, probably two or three years ago, George and I crossed paths and sat down and said, we have a lot of overlapping interests, let's try to do something.
DR. HANDLER: He keeps me busier now that he is retired than when he used to be in his office.
DR. IREYS: That is why I haven't seen him.
DR. IEZZONI: Are there other questions for Henry? Henry, if you would care to stick around, that would be great. I think Dr. David Zimmerman is next, unless Susan Nonemaker wants to go. Either one.
DR. NONEMAKER: Okay. Liza, since we are behind on the agenda, how much time do the two of us have together?
DR. IEZZONI: I think we have plenty of time. The last panel has a huge amount of time allotted to it that it will not use.
DR. NONEMAKER: I didn't want to be insensitive to them, so I wanted to clarify what we have.
DR. IEZZONI: You spend your 15 minutes talking, and then we'll do --
DR. NONEMAKER: Fifteen minutes?
DR. IEZZONI: Ten, 15 minutes each.
DR. NONEMAKER: I'll try to be careful not to take three hours. Basically, I have been thinking, as I was invited and came in today, it has been a long time since I have been here. I really do appreciate the opportunity to be with you and talk about what we are doing at NCFA.
I used to come before the subcommittee on long term care, the precursor to this group, probably about twice a year to give a status report on what we were doing with various initiatives. So I feel like I have a lot to cover this afternoon, and I'll try to do the best that I can in 15 minutes or so.
Basically, I am from the Office of Clinical Standards and Quality, where we have had the lead on developing basically clinical measures or the clinical information systems and data sets that have been introduced in post-acute care and long term care.
I guess the thing that is striking me is that there isn't the vast wasteland that was somewhat described this morning, or that I think people may believe exists. Bob did talk about a desert, which I think was probably a very nice analogy, in terms of data collection for this environment, but we really are much further ahead in terms of standardizing clinical information than many of the post-acute environments. We have had rich claims information in those systems, but we haven't seen a lot generalized. There is a lot that is patient specific, but not a tremendous amount that is generalized as we have been trying to do more with post-acute and long term care.
I also want to underscore what Bob was promoting, and I think he said it incredibly well. I want to remember your analogy, in terms of the desert situation. I think to the extent that the perfect is the enemy of the good, to the extent that we try to create the perfect measures or the perfect systems will be paralyzed if we spend years trying to come up with the perfect recommendations. I think the most important thing is to move ahead and to build systems so that we can begin to start to understand the patient characteristic piece and the care need piece in a standardized way, so that we can answer many of the questions that were raised this morning in terms of what is post-acute care, who needs it, what does it involve, resource intensity, and all that type of thing.
I have some transparencies I'd like to run through. What I wanted to start with was talking about the conceptual framework that we have been working with in the past several years related to post-acute care. We really are trying to build on many of the themes that were amplified this morning, in terms of using common data elements, looking more at patient centered care as opposed to what I consider more the artificial characteristics of the care provider or the environment.
So what we are trying to do with our quality assurance and improvement agenda is to really look at building standardized data elements that would help us to measure in a consistent way a beneficiary's clinical characteristics and care needs.
In terms of building those measures from the quality side, we are looking at being able to use those in a regulatory quality monitoring system in terms of giving providers feedback about how they would rate in terms of key performance measures as compared to their peers, and also looking at being able to provide better information to consumers about post-acute care choices and quality issues.
On the payment side, we are looking at being able to have payment systems in the future that become more beneficiary centered, looking at trying to focus more on measuring care needs, as opposed to having those systems all be very discrete and being focused primarily on the characteristics of the care provider.
I can't emphasize enough how much I as a clinician agree with what was said this morning about the variation, the tremendous variation that you see across providers and the kinds of settings in which care can be received for a particular kind of clinical condition. We need to have ways to level the playing field, or be able to compare apples to apples, as opposed to putting different measurement systems into place that would be used across different settings along a continuum.
So what we are trying to do in the short term, largely as a result of the Balanced Budget Act of '97 that was referenced this morning by Korbin, is to put in place provider specific payment systems, which I'll go through in a minute, and then as we get those systems in place, work towards more of a long term goal. I really see this as at least a 10, if not a 15 or 20 year plan that would get us towards more of an integrative post-acute care system.
The bottom line is that we have to in some way cap costs, but unless we are able to do it in such a way that we are not just shifting from one setting to the other as we are maxing out in one setting and then being able to go on to the next provider, we are never going to really get a handle on costs within the program. So we need to have a new way of being able to pay for care in such a way that it makes sense.
What we are trying to go towards is building a core data set or some kind of common assessment instrument or a common data element that will be used across settings, that would help to inform the building of this payment system.
So the way I see it is that we are going to need to collect data for a period of years to even be able to understand how that system would be configured, how those rates will be derived, and everything else. It is really a matter of starting with getting a better information system related to clinical characteristics and care needs first.
Where we are has been referenced this morning, but I put it up on a transparency just to highlight. We have got the SNF perspective payment system in place that began on July 1, which is in the process of being phased in over the next three years. We are moving forward on the home care front. This has been slipped to the year 2000, and we will be getting ready to implement the OASIS data set. I can see them smiling now; I'm willing to take questions or have things thrown at me if you like, but I can only say that I personally reinforce the need for a common data element across MDS and OASIS.
Then finally on the rehab front, we are at this point looking to use an MDS based approach. There haven't been final decisions made, but in terms of the instrument that has been used traditionally in the rehab field, that being the functional independence measure or FIM, we did not want to create a payment system that was basically built on apples when we had very similar care being provided in many skilled nursing facilities using a system that was built in oranges, if you understand the analogy that I am trying to make. We wanted to have the same kinds of measures in place regarding clinical characteristics and care needs across these two data sets that will be the foundation for the payment system as well as quality monitoring systems for the future.
I want to talk about MDS as a data structure for a minute, before I get more into the post-acute care data system that we are now building as an instrument. In terms of automation, the model that we set out basically has information coming in from providers, this being an administrative data set, information coming in from the provider, going through states. That could be a state survey agency, a Medicaid agency or someone that is under contract at the state level, then that information being passed on to HCFA, so that we build state data systems that are then capable of going out and doing a lot of quality regulatory related work. We then have information being passed on to a national system for both payment and quality purposes, as well as to support all sorts of research.
On the MDS system, what we have as a model is having information coming in for all of those residents who are in Medicare or Medicaid certified facilities. So we have got probably about 1.6 million residents that we have, or patients or whatever you want to call them, that we have data coming in on periodically throughout the course of the year, in terms of having a longitudinal record established for an individual.
We have basically built an information infrastructure that has put a standardized hardware platform in place within each of the states. We are in the process now of getting ready to deploy some applications that states will use to be able to generate quality indicators in a consistent way.
So basically, we have got this big infrastructure that we are starting to build, in terms of how data is collected, how it is analyzed and can be used at the state level, and then brought on up to the national level. We would expect to use that same infrastructure that we have built to be able to collect information from other types of providers, be it home care agencies, hospitals or whatever, is the model that we are working under.
We have had the regulations related to MDS transmission in place since this past summer. They went into effect June 22, and at this point we are in the process of building a national repository.
In terms of the handout that you were given on MDS for post-acute care, I wanted to go on to talk briefly about the data set that we are building for the rehab community, that will eventually transition our SNFs over to, provided that this all goes forward. It would become the common assessment instrument that would be used for the high end nursing home population, as well as the rehab hospital population.
We had originally started this as a clinical instrument, given that there was a lot of concern expressed by the industry, I think very rightly so that the long term care MDS that we had developed and put into place was not the best assessment clinically for use with the so-called subacute population within nursing homes.
So we started developing another instrument. About that same time, the Balanced Budget Act of '97 was passed, and we began to look at using this for rehab hospitals as well potentially long term care hospitals.
We have been developing this instrument for over a year. We are just in the beginning of field testing it. You have a draft in front of you. I believe in the last meeting that you had, I was unable to be here, but I think Carolyn had brought a draft at that point and had done a brief summary a long time ago. So this is a different variation than what you had; I'm glad I brought it along with me.
What I wanted to talk about, because it really builds on what was said this morning, in terms of one of the phases of the instrument development study that we are currently conducting, is that we have been making a lot of assumptions about substitabiltiy of care, believing that in many instances -- and this came out of a focus group discussion that we held with the industry -- you can have the same individual. Say you have somebody that has a fractured hip, just to use a very basic example. That person could go to a rehab hospital, they could go to a regular nursing home, they could go to a specialized nursing home, a so-called subacute specializing in rehabilitation, they could go to home care.
Given the tremendous variation in costs as well as potential outcomes, you need to have a measurement system in place in order to be able to understand the individuals that are going into those types of care.
So we had done some primary data collection, looking at collecting data in rehab hospitals as well as long term care hospitals, given that they have not been using the MDS that has been used in nursing homes. In that, we tried to collect about 1800 patients. I think we actually got data on about 1200 patients. Not everyone who said they would participate did. But we now have information using the same kinds of measures for these three populations.
Just to give you a thumbnail sketch, rehab and SNFs don't look all that tremendously different in terms of the characteristics of the patients that are in those two types of settings.
There is also some similarity with the long term care hospitals, but not nearly as much as what there is between rehab and SNF. I would say that people in the rehabilitation hospital setting are receiving more therapy services, but it is hard to say whether that is because of the setting they are in. I think we had talked this morning about three hours of therapy being a pre-condition to even being admitted in that type of setting. So it is really difficult to say whether the care is being driven by the setting or because of what they actually needed.
Anyway, we are just beginning to look at the data, but it is really the first time this kind of comparison has actually been made.
We are also -- after we finished the MDS PAC, which we are hoping to finish up this spring, hoping to begin work on a third iteration of the MDS that will pick up some of the items from the MDS PAC. We are also hoping to do work on some special populations within the long term care institutionalized population.
One of the areas that I think has not been well addressed is the whole area of palliative and terminal care. So we are hoping to build some measures for that population in particular. But ideally, it will be able to used across program lines, being tailored more towards the characteristics of the patient regardless of what setting they are actually in.
I want to talk a few minutes about some of the projects that we have underway related to quality measurement. We have just started a project to develop and test indicators for quality of life in nursing homes. Rosalie Kane is doing that at the University of Minnesota. That is basically about a two and a half year project.
We are just starting as well another project with a very long name that I will go through and explain what it actually means. We are going to develop and validate measures and indicators for quality and appropriateness of services in post-acute care and long term care.
We have as principal investigators for that John Morris at HRCA, as well as Vince, who is a member of your committee. We are really excited about the work that we are going to be able to do under this project. It is being headed up at ACT. We have a four year period of performance on that.
In terms of some of the specific tasks, we are going to look at the existing measures that exist in the literature, that have been developed by individual corporations, what exists on websites; we are doing a very exhaustive search, in terms of what is already out there. Then we are going to look at the science that is being conducted to support those measures, whether there has been enough validation that has actually occurred, and make recommendations to HCFA, hopefully by this May, about some measures that we really think have been significantly substantiated and are ready for use within our programs.
We then want to come up with a design to go on to test those measures that have been insufficiently validated, as well as do development work for areas in which there are gaps or holes, areas that weren't really well addressed by existing measures.
One of the major tasks under the contract is to go on to develop markers for the post-acute population. I would see this as being less condition specific than the work that Andy talked about this morning. These would be broader measures, as opposed to the more disease specific situations that he was describing. We will then go on to do testing, as well as hopefully do some development work for special populations.
Two that I have picked out as areas that we will probably pursue are the areas of palliative terminal care and individuals with dementia, trying to come up with more specialized measures for those areas.
How do we expect to be able to use this information? We have another contract underway that David and Andy are working with us on, in terms of being able to integrate quality indicators into our long term care survey process. I would see us moving much more significantly in terms of using data in a very objective, standardized way in our regulatory process over the next five years or so.
We also as I said are in the process of being able to distribute information to providers that they can use for their own quality assessment and improvement activities. We are going to begin to work more on distributing information to consumers, and then there is the whole notion of value based purchasing, or trying to have a better understanding of what it is that we are actually buying as a program through Medicare. I think that the measures we will be developing related to clinical characteristics and care needs will move us much further in the direction of being able to understand what it is that we are actually buying related to services.
Two other points that I wanted to make. That is that the plan with these quality indicators and markers is that we would apply them according to the characteristics, the patient and population, rather than the care setting. That would mean for indicators related to rehabilitation, we would expect to see those measures used for patients that were in SNF, patients that were in home care, patients that were in rehab hospitals. They would be tailored more to the patient population as opposed to the setting in which they were actually receiving care.
The other thing is that we would expect to develop markers or indicators derived from data obtained from a variety of data sources. As I have been talking about, we have been using provider assessment instruments to develop administrative claims bases, believing that that will give us a very rich source of information, as well as being a methodology that is very feasible and very much something that we can afford to do.
While I certainly understand the importance of conducting surveys, and I very much understand and would agree with the importance of having self reported data, you need to be able to think through what would be the data collection mechanism, how would we operationalize that as a program, thinking about Medicare, how can we collect this information in a consistent manner over the long term.
So what we are going to do through the contract that Vince is working on is, as we are developing markers, to be thinking about where can we get this information, how can we get it reliable, how can we afford to collect it over the long term, and how can we operationalize this in the Medicare program.
So that to me is the fun of it. It is not just the scientists developing the measures, but can we actually make something that we can operationalize and make work.
So that is the short term plan, Liza. I'll turn things over to David, or do you want to take questions first?
DR. IEZZONI: You have a very full plate.
DR. NONEMAKER: Yes, we're kind of busy.
DR. IEZZONI: Yes, I think so. Sue, why don't you sit down for a moment and let me see whether any of the committee members have questions at this point?
DR. NEWACHECK: I have a question about children. There are about 100,000 children in institutions now, and the majority of those are SSI recipients and eligible for Medicaid, so they come under HCFA's domain. I am wondering, does any of the work that you describe, which sounds very fascinating and interesting, apply to children, or will it be applied to children?
DR. NONEMAKER: Right now, if you have a child that is in a certified long term care facility, the facilities are required to do the MDS. There has been debate since we first implemented in 1990 about whether this is a good tool.
The questions that you got in terms of, can you do like an adult functional status assessment and how does that need to be modified to make it most workable for children, have really come up in that venue. We have been having those providers use the MDS. it is a condition of their participating in Medicare.
But what we are in the process of starting is an analysis of the under-65 population within nursing homes, and expect to be able to look at pediatrics under 18 as a group in and of themselves. What we are hoping to do, both through work that we are doing now and as we begin on version three of the MDS, is to come up with more of a pediatric version, to the extent that there appears to be a real need for it.
There has been some question as to how many children are in facilities and number of facilities. It seems to be very pocketed within the country.
DR. NEWACHECK: That is one of the problems that Henry alluded to. We have no real data to even tell us how many kids.
DR. NONEMAKER: Yes. I was kind of excited, because I was thinking, we were going to do this analysis, and then I got a phone call in November from some people in New York, which has quite a few facilities. They were complaining about needing to transmit NDS data. They said that they had previously had a waiver from that state requirement, but now that it was a federal requirement, could they continue to have a waiver, and we said no. I started talking about this analysis that we were going to do and how we were going to do all this development work. But then it occurred to me, gee, we have no data to work from if they are not going to transmit data to us.
There is no standardized system in place to help us even understand this population. That has been an issue since 1990, frankly. I am real if anything sensitive to wanting to do something that makes clinical sense. I have asked them, who are these kids, what do they look like, and they don't know. Everybody tells their own little anecdotes. That is what we had to work from.
So we are going to try to work on it, Paul.
DR. NEWACHECK: Great.
DR. COLTIN: I believe you said that the MDS data that are being transmitted to the states and to the federal government from the Medicare certified facilities are on all of their patients, not just their Medicaid --
DR. NONEMAKER: It is on all of their patients, right. I talked extremely hurriedly, and I apologize for that. But it is on everyone within the facility.
In terms of how our quality regulations work, we look at care provided to everyone within those facilities, not just to the Medicare patients. So we argued that we have the authority by nature of our statutory responsibility, to look at care for all patients or residents within a Medicare certified facility to collect information on all of the individuals. So it is private pay, it is Medicare, Medicaid, third party, whatever.
DR. MOR: Managed care.
DR. NONEMAKER: Managed care, thank you, Vince.
DR. COLTIN: I have a followup question. Given that I understood you correctly, you said that for the post-acute patients that you were going to be measuring it five days and 14 days, is that right?
DR. NONEMAKER: The current system for Medicare right now looks at five days and 14 days. Under our MDS post-acute care, we are looking at using basically a three-day observation period, so it would be done on day four. Then we are looking at having the next observation be about seven days thereafter. So we are looking at testing day four and day 11, and then having another assessment done at the point of discharge. So if somebody for example has an assessment done on day seven, but they actually got on on day eight or nine, they would have some data elements repeated just so you would have a final set of outcome measures.
DR. COLTIN: And if they go out on day three?
DR. NONEMAKER: They would have to have the entire assessment completed once. The problem is that you have no point of comparison. We were talking about this as we broke for lunch. We are not thinking that we are going to be able to get an assessment done at the point of admission as the patient rolls in the door. From talking to clinicians, it just didn't seem that that was something that was feasible for them to do. A, they need a certain period of time clinically to complete it, but for many of the functional status measures, you want to have a period of observation so that they can see what that person is really like over a period of time to score them more accurately.
DR. COLTIN: I know that in our data, and I'm talking about managed care data now in our data set, we have about -- between 15 and 20 percent of our patients would be discharged before five days. We had a lot of concerns about that, not having information. Some of our contracts with the nursing facilities, they actually come and start the evaluation while the patient is in the hospital, before they even get to the nursing home, and then complete it within two days of being admitted, so they have a little bit longer observation period. But they are better able to assess the patient's needs if they actually come into the hospital to plan.
DR. NONEMAKER: Sure. So it all argues towards a more seamless approach that uses common data elements across care settings.
Oh, David, if I could have two more minutes, one thing that I forgot to mention was the uniform needs assessment initiative. That is one of the things that I used to come before the committee and talk about. That is something that was a part of OBRA 86. It was actually the reason that I was first hired to come to HCFA, when I first became the assessment cheerleader at HCFA, was to develop a data set that could be used across care settings to assess individuals receiving post-acute or long term care.
So the Congress had viewed this instrument as something that could be used by hospital discharge planners, but they had also looked at it being used at different points along the continuum.
To the extent that we as an agency say that we are going to work towards more commonality between OASIS and MDS, that is the vehicle that we are going to try to use, in terms of doing some synthesis work, looking at reliabilities, and then moving towards hopefully more commonalty as we create the next iteration of all of these instruments.
DR. IEZZONI: Can I just make one comment before we are going to have to move on, because I want to be sure to give David time.
You raised the issue of liability, and I just want to touch on that briefly. I think for us to be able to convince people that they should add specific data elements to other data sets outside of the context in which they were developed, we are going to have to assure them that data would be reliable and credible.
DR. NONEMAKER: I agree.
DR. IEZZONI: Having first cut my teeth in my research on DRGs and DRG creep and ICD-9 coding creep, just looking at this instrument, there are lots of opportunities for subjectivity.
For example, there are a whole bunch of questions about whether performance is more impaired now than it was a few days ago or at a particular time. Given that this information is going to be recorded by the nursing home that is getting paid, and given that their payment is going to be based on this information, it seems to me that there are real opportunities here for creep.
So how are you as an agency going to be convincing yourself that the data are accurate, reliable, valid?
DR. NONEMAKER: A couple of points. First of all, the specific item you referenced is something we just developed. It was something that clinicians wanted. We are just starting on liability testing. So we don't know how it is going to work.
To the extent that it doesn't work well, we are going to try to improve it. If we can't get it to be reliable enough, we will probably just discard it. It wouldn't be on the form.
We have had good levels, if not very good levels, of reliability reported for MDS data. But I have to say that the degree of reliability that you get is very much associated with how the information is being used. So if facilities know that they are getting paid off of it, if they know that they are being monitored off of it, they have a tremendous incentive in which to collect accurate information, and they do so.
What we are in the process of beginning to study is to come up with a system to develop a system to look at the accuracy of MDS data. Then we would use those same kind of principles with our other data systems. Andy Kramer is one of the principal investigators for that. We will be basically putting in a system that will help us to target the cases that will be reviewed on site. There would be more of a pattern analysis of large volumes of data to determine the records that we want to zero in on, the kinds of homes, the kinds of patients and that type of thing.
It is going to cost some money, is the bottom line, and it is going to take a process that doesn't yet exist that we are going to have to implement.
DR. IEZZONI: In fact, the standard way of testing reliability isn't going to really work in this context. The standard way of testing reliability is going back and re-abstracting retroactively, retrospectively, medical records once again. The only way you an really do this is to re-examine patients, to make sure that the assessments are correct.
So I think that it is going to be challenging. But this is a new instrument and one that you are going to be --
DR. NONEMAKER: Oh, yes, I just want to emphasize that this is in the process of becoming, so this is not a done deal. What I was talking about with the MDS is, we are trying to come up with a menu of approaches that we can put into place, some of which won't rely on an on-site review of records, but rather simply pattern analysis, different kinds of ways of reviewing the information to see then where we want to spend our efforts on site.
So I would see that work being done over the next 18 months. Then we need to put in place that system to actually operationalize it. I'll tell you, I actually have some concern about the fact that we don't have a big enough system in place right now.
DR. IEZZONI: Thank you, Sue, for coming back and talking to us. That was very helpful.
DR. HANDLER: Liza, could I ask one question?
DR. IEZZONI: Quickly.
DR. HANDLER: Quickly. Item 11, Section A, I suggest that you use the multiple race item that the year 2000 census is going to use. There is an interagency task force that is meeting now to determine how to tabulate and publish data using the multiple race item. That is an OMB directive. That is not what this is.
DR. NONEMAKER: I'm sorry, sir. My understanding was that we had the latest -- we'll fix it.
DR. IEZZONI: That is a good point.
DR. NONEMAKER: Yes, thank you for telling us that. I thought we had done that some time ago.
DR. HANDLER: Not here. This was the prior one, not as of last summer.
DR. NONEMAKER: As of a year ago last summer. It will be done. Thanks.
DR. IEZZONI: David, we are going to have to make sure that you have got the lavaliere mike on, which I guess our trusty audio person has given you.
DR. ZIMMERMAN: I'm wired, and I'll assume you will tell me if I'm not really electrified.
I thought that it would be the most useful thing to do to explain to you in a little bit more detail, actually following on what Sue was saying, what is going on now with respect to the MDS use in quality assurance, since we are at a propitious time to discuss it. It is going to be going into effect within the next six months. Talk about what is going to happen with respect to the use of the MDS with quality indicators in the nursing home setting.
I'm not sure that this is such a great model, but it is a model, and it is going ahead, and the train has left the track, is moving down the track pretty quickly. So if people want to make a change in how this is going to happen for other settings, I think the challenge is to do it pretty quickly.
So I'd like to do that, and then I'd like to come back and talk about two other general issues pertaining to the implications of what we are doing in the nursing home area for post-acute care.
One of them is measures, and I'll actually finish up with that. But the other one is the quality issues, since we are supposed to be talking about quality issues as opposed to data issues. Certainly one has implications for the other, but I'd like to talk about what some of the issues are that we have faced in terms of implementing a set of indicators for a variety of purposes.
First of all, just some very basic background. I'm sure that many of you are quite familiar with the indicators that have been developed, at least that we developed at our center. We have 12 domains of indicators. Now, this covers what somebody called in a presentation they were making yesterday the traditional nursing home population, as opposed to the post-acute or subacute nursing home population. At any rate, there are 12 domains of care, only 11 of which can be defined from the now most popular version of the MDS, so we only have 11 domains.
We have 24 quality indicators that cover those 11 domains. I'm not going to go into all of the individual indicators. I'd be happy to talk to you about them if you would like, but I think that wouldn't be the most effective use of our time.
I will say with respect to domains that the major issue that has been raised with respect to the quality indicators is the extent to which they cover what has come to be known as quality of life. I don't like the term, but at any rate, there is certainly very limited coverage of the MDS and therefore the quality indicators, because our indicators are based entirely on the MDS at this point.
There is limited coverage in terms of what has been called quality of life. Let's leave that alone, and if people want to address it, we can talk about that.
At any rate, we have developed a set of initially well over 100 indicators through clinical panel review. We have seven clinical panels representing the primary disciplines in long term care, medicine, nursing, the various therapies, pharmacy, dietetics, et cetera. And we narrowed it down to 30 and then to 24, which are available in the most frequently used version of the MDS, as I said.
We have outcome indicators such as the prevalence of pressure sores or the prevalence of bladder or bowel incontinence. There are indicators relating to processes of care or the absence of process in some cases. For example, the prevalence of physical restraints or the prevalence of antipsychotic medication use, et cetera. There are some indicators, a few of them, that actually could be considered a combination of outcome and process, in the sense that they reflect both an outcome or the status of a resident at least, and the process or care, or the lack of that process,for example, depression with no treatment is an example and incontinence with no toileting programming, things of that nature.
Rather than dwell on the indicators themselves, what I would like to do is follow up on what Sue talked about, and show how they are going to be used and how they already are being used in both internal quality improvement functions by providers and then external quality assurance by a variety of people, the survey process, the joint commission, et cetera, in its certification process.
There are two levels of indicator reports that are typically available now to providers participating in various projects, and they will be available to all providers within the next six months. There is a set of facility reports, which include both facility characteristics, basic demographic characteristics about the case mix in that facility, the residents of that facility, et cetera, and then a series of quality indicator profiles that are available, in which the facility is compared against a peer group. Usually the peer group is all over facilities in the state, but it doesn't have to be, and there will be flexibility in the system in terms of how an individual facility or the survey agency can define the particular peer group.
Then there is a series of resident reports, only one of which exists now. This is a report showing the QIs that are flagged for each resident.
Now, what you see in front of you, which you can't see very well -- and what I will do is send back to you examples of the report. I apologize for this; the only way I could show the entire report on a screen was to do so with relatively small print, and I expected that we would have a little bit larger screen to work with, but we don't. So I will tell you that along the left side of this report are all the indicators, or each one of the indicators. Then there is information about the number of residents of the nursing home that actually has the condition, an incidence of a new fracture, prevalence of falls, whether or not they have behavior symptoms, whether they have symptoms of depression. There is also as you saw earlier an indicator on symptoms of depression without antidepressant therapy, so there are some symptoms of functional depression with no evidence that they are receive antidepressant drugs.
At any rate, the information will be available in such a way that they can look at their own percentage and then the peer group percentage. Then it will also provide information on the percentile ranking in which the facility finds itself. That is, if one arrays it from the most prevalent to the least prevalent, where do they fall within that distribution across the peer group.
There are also possibilities for actually having graphic flags that they can use to determine whether or not they are above or below some threshold. That whole issue of threshold is one of the four or five major issues that I think has yet to be addressed adequately in terms of use of this type of information for quality purposes.
At any rate, that is an example of a facility report.
There is also a resident report, in which for each resident, which is a line item here, there is an indication of which conditions they have, which conditions they flag on. Then you can total that information across all of the residents and across each resident as well.
Now, I show you this because this is very similar to the kind of information that within the next six months is going to be available both to all survey agencies in the United States, each state survey agency, and every facility also in the United States. They will be available through software that we have developed and are now alpha testing. We are just finishing that aspect of it. We will go into beta testing next month in one or two states, and implementation is planned for the middle of this year.
What will happen when it is implemented is that each survey agency will be able to pull down information like this on every facility, and every facility will be able to go into the same system and pull down the same measures and the same reports from the Internet. So the same information will be available to both providers and to the survey agencies. This information is now not currently available.
What will the system look like? Well, the software development recently has been completed and alpha tested. We have demonstrated the software recently in Baltimore, and people seem to think that it worked well for the purposes for which it was intended. Obviously, people wanted more flexibility, but in general they seemed to think it would do the job. We are going to begin beta testing, and the deployment is scheduled now for the middle of '99; whether or not we will make the first half is anybody's guess. But I think probably it is more important to tell you that if we don't make the first half, we will make the summer, and the summer will be 1999, which is actually than a lot of other initiatives that one might have seen in which the initiative has actually hit the streets two or three years later.
So it will happen. It will happen this year. What is going on now is that training is being planned for both survey agencies and the core facilities to use this information.
Now, it is not that facilities or providers don't have any experience with this at all. There are several different organizations that offer programs using MDS data that can provide facilities with this information for their own internal quality improvement reports. For those of you who are familiar with the joint commission's ORIX initiative, there are many vendors who are providing information like this as part of that initiative. So it is not as if they don't have the information yet. At least some of the providers have this information, some of them do not. In fact, the majority do not.
The reports will be identical for survey agencies and facilities. At least, that is the intention. As far as our input is concerned with HCFA, we have tried to insist that that was the case. We have strongly encouraged identical information available for both surveys and season facilities.
They will use the same access mechanism, survey agencies and facilities, in order to draw down the information that they need. The access will be via the Internet. The information is designed to be used both in the survey process, to drive the survey process in a more structured way. I will tell you that we are working with Andy Kramer and his group at the University of Colorado on combining these indicators with the more structured approach to the survey process or the quality assurance process on site, so that it will be a more structured process in general, and it will be coordinated between the offsite look and the onsite protocols.
I wanted to talk a little bit about a couple of issues that have come up repeatedly in terms of the work that we have done in the nursing home area. I think frankly that the issues in using this information for quality purposes are probably even more similar between settings or across settings, that is, post-acute versus nursing home, just to use those two examples, since that seems to be the focal point of this session.
I think that there is more similarity and comparability in the data in the quality methodological issues, policy and methodological issues, even than there is in terms of the potential applicability of measures from one setting to another. Yet, I think that there has been precious little attention paid to some of these issues in terms of how this information is going to be used.
One of the biggest issues frankly is the difference between how the data are going to be used and how the indicators would be used. There is a major distinction to be made here between whether indicators like this are going to be used as targeting mechanisms to guide the future inquiry, to guide more intensive inquiry or review in the next phase of that quality assurance process on the one hand. That is, as targeting mechanisms to help determine more effectively and efficiently where we should be looking next, That is for both external and internal quality assurance.
A big distinction between that and using as the basis for decisions -- that is, using the reports themselves as the basis for decisions without that subsequent review. The survey process -- many providers may not agree with this, but at least in concept or theory, the survey process is indeed supposed to be an example of the former. That is, it is supposed to be an example in which you would use information like this to then target more efficiently the followup activity.
However, in some cases this information is being used as the basis for decisions without any followup activity. One of the most important and unfortunate examples of it is actually in the managed care field. I have received many now calls from managed care organizations saying that they are particularly proud that they are using information like ours or other ones to make decisions about whether or not a provider is part of their network or not part of their network, about whether or not they are going to get some sort of a reimbursement bonus for high quality, or a sanction for low quality. They are using this information not as a targeting mechanism for the next step to follow up, but as a basis for decisions now. It is a big issue, and it is one that goes right to the heart of both the reliability issue that you mentioned, Liza, and the validity issue, if you define predictive validity in terms of the ability of a measure of an item to accurately predict that there is a problem. So we have to be careful about it.
I don't think the answer is that you simply shouldn't do any of this, because right now they are using managed care organizations and other organizations are using information that is far less reliable than this, is actually far more subjective than this to make decisions on their own. So when you look at it as a compared to what problem, you really don't come to the conclusion that you simply can adopt the head in the sand approach and say let's just not offer the information. In addition to which, it is critically valuable for care planning, which is frankly its original purpose.
I wanted to talk also about several policy and methods issues. We have heard several references today to the issue of risk adjustment and how well we can adjust for risk, how well we can make appropriate comparisons.
I think that here is an area where many of the same types of issues that we are facing in long term care measures we are going to face in post-acute care measures. There is very little difference in this respect. Frankly, there should be much less interest in the way in which one risks adjusts, because the way in which one risk adjusts is actually probably more conditional on what the purpose of that risk adjustment is than whether or not we are actually capturing valid and fair adjustors of risk.
Just to offer somewhat of a contrary view to the view that is offered in most cases, I think that we run the risk of over adjusting for risk because of the fact that in many cases, I fear we are going to adjust out actual factors that are reflective of quality of care. They are not reflective of the stock of the condition that the resident brings with him or her to that situation. So I think we have got to be careful about over adjusting in that regard.
The issue is not simply whether or not somebody is at higher risk, but are they at higher risk and not at higher risk because of previous care issues in that facility or for that provider in general. Secondly, are they at higher risk and are they at higher risk with virtually no intervention potential on the part of the facility or on the part of the provider. In many cases, it is frankly the responsibility of the provider to assess that risk and act accordingly to reduce the probability that the risk ends up increasing the probability of the event. Just because somebody is at higher risk does not mean that nothing can be done to interfere with that relationship between the higher risk and the probability of the condition.
Now, having said that, I think that if we don't risk adjust and we don't risk adjust carefully, we run the risk of employing unfair quality indicator variables and nothing will bring down the system quicker than legitimate concern about those risk factors on the part of providers. So we have to be very careful about it.
The second issue is the issue of attribution. I think there are really two aspects of this issue, and several times people will only capture one or none.
First of all, did the condition develop in the facility? If you want to transfer that to post-acute providers, whether or not they are institutional providers or home care providers, the setting really does not matter. Can we logically say that the condition that we are now measuring is in fact one that is likely to have occurred on the watch of the provider? If the answer is no, we had better be very careful about making sure that we don't make that inference that they are responsible for the care that is provided.
Even if it did occur in the facility or on the watch of the provider, is the condition associated with poor care? I actually don't like the term outcome either, because outcome connotes that it actually is -- not only that there is a status of a resident or a patient, but that that status somehow inferentially is a result of a particular set of interventions.
We don't know that yet. We may be able to appeal to the law of large numbers over the course of a lot of epidemiological research, but I don't think we are there yet. So we have to be careful about that attribution issue.
Thirdly, the use of standard and threshold levels. The standard and threshold levels that we are using, that are being used in general, are oftentimes selected with very little consideration to whether or not they are appropriate or what their purpose is. You can see that for example in the survey process, that it is not clear whether or not the survey agency is going to use an 80 percent criterion, whether they are going to use a two standard deviations or three standard deviations. Frankly, it is a concept that is foreign to them and is simply not going to be of that much practical use.
Just to give you an example, I was giving a presentation yesterday to a state directors of nursing association, and three people in a row asked me the same question. Now, are the survey agencies and is everybody else who is going to be judging us going to be using what kind of threshold levels, what kind of standards are they going to be subjecting us to. So we have to deal with that very carefully.
The last one that I will mention is the whole concept of aggregation and the relationship between quality indicators or any other quality measure. I think there is currently a rush to aggregation. There is a rush to want to adopt that single index, that single score, that single quantitative value that we can use to assess whether or not there is high or low quality.
Yet, if you take a look at nursing home care -- and I would argue that the same thing is true in post-acute -- if you take a look at nursing home care, it is such a multidimensional concept, that any attempt to rate across those dimensions or even within those dimensions has got to be done very carefully. We are using incredibly unsophisticated methods to rate or priortize these measures, and we are not looking at all at whether or not they are in fact conditional or contingent on the status of the individual. So I think we need to be careful about that.
What I'd like to do at this point is simply say that -- I'd like to talk a couple of minutes about the measures. I think that the measures that we have in the long term care area probably do have limited applicability to post-acute care. We have to be careful about whether or not we can simply wholesale these measures.
But one area that I think is probably under addressed is, in this rush to develop post-acute, whether or not it is condition specific or even some of the global measures, I think we are forgetting that in terms of the long term care population, we have a group that actually has chronic care problems and has chronic impairment levels that then are exacerbated, sometimes dramatically so, by the fact that they suffer a traumatic event.
When they suffer that traumatic event, I think we have the very real danger of forgetting that they still have these chronic care problems, and jump into a level of assessment or a technique of measurement that simply ignores those chronic care impairments and treats them as if they are just like any other post-acute population. I think that is problematic as well.
DR. STARFIELD: Can i clarify, what do you mean by post-acute? Do you mean post hospital? You are distinguishing long term care from post-acute? What is post-acute?
DR. ZIMMERMAN: I don't necessarily mean post hospital, and I don't think we should get in the habit of using post hospital, because as nursing homes become much more sophisticated in their ability to deal with acute care problems, it is not necessarily a post hospital event. Nursing homes now as they become more adept at dealing with acute care problems actually handle a lot more acute care problems within their own setting. So you may not technically have a post hospital situation.
But I do mean the period immediately after fairly intensive care for that acute problem. That is the way I am defining post-acute care. I don't know if it is right, but when people talk about post-acute care, I presume that they are using the term for a reason. So it is the care that comes out of the more acute intensive care, period. I just don't think that is necessarily going to be hospital based as much as it has been in the past.
DR. IEZZONI: Bob Kane this morning in the initial talk made the argument that we should be quite clear when we use the term post hospital. Barbara, we'll work this out. We're not going to solve it right now.
DR. STARFIELD: I just want to understand what people are talking about.
DR. IEZZONI: I was going to say, it is a good thing for us to hear from everybody on the panel, how they are using the word.
DR. ZIMMERMAN: But having said that, I do think that if you take a look at the measures that are being bandied about under the title post-acute, you have a lot of measures that aren't just post intensive acute care or post hospital. You have a lot of measures that in fact reflect something that is a very heavy rehab orientation, something that may not simply come out of -- not all rehab is based on simply the post intensive intervention coming out of a traumatic event, not at all.
So I use post-acute simply because it seems to me that that is what people mean when they say post-acute. But I do think that the whole concept of rehabilitation, for example, really poses a major problem in terms of confusing what post-acute means. I wish I was the bearer of better news than that, but I'm not.
DR. IEZZONI: Thank you. I just wanted to underscore your first step on the attribution decision. Knowing the timing of events is going to be really, really critical. So whatever data set we put up, we would have to have a very clear indication of whether something was pre-existing or occurred after a certain intervention, and having a sense of the timing of events is extremely necessary.
DR. ZIMMERMAN: I think the problem is probably even worse than that.
DR. IEZZONI: Tell us how.
DR. ZIMMERMAN: Well, it certainly is a mighty effort on the part of those who have been engaged in the development of post-acute instruments, and nobody has worked harder at it than the woman sitting next to me. In that process, they have really tried to capture the moment as quickly after it happens as possible by moving to five and 14 day.
But if you take a look at the statistics on the discharge rates, before 14 days you really do lose, depending on the statistics, depending on who is arguing it, you lose at least half, if not 80 percent of the individuals who then only have one.
I think Sue referred to it. I think the conundrum here is that it may be very difficult to get same day assessment. But having said that, i guess I don't buy the argument that we can't do same day assessment. It seems to me that whoever is providing that care actually has to do same day assessment. The question is whether or not there is a reasonable documentation strategy on first day assessment, that is, the day that they begin the post-acute care, that would be able to A, give them the kind of care planning help they need, and secondly, would provide the kind of information we need to have the baseline much more quickly than five days.
DR. IEZZONI: I think this feeds directly into our electronic medical record issue that our subcommittee as a whole is going to be focusing on. Once you have your daily care documentation in electronic form, the ability to extract the information that you need for quality monitoring measurement becomes easier. Pipe dream still, but it is something that you may ironically be closer to in long term care than they are in many of the acute care settings.
DR. ZIMMERMAN: Oh, I think that is true. I was going to suggest -- probably Vince would agree with this. I think we are a lot closer to it in long term care.
DR. MOR: I'd like for you to emphasize again that right now, there are 1.6 or 2 million of these records in some database someplace that people are extracting, and within six months they are going to be person specific and facility specific feedback in the process. That is a really big, big difference than anything we have ever looked at before as a committee in terms of real time information turnaround on administrative data sets that have clinical information embedded in them, with all of the problems associated.
DR. IEZZONI: Can I just ask you one mor question? This morning, Andy was talking about developing his quality measures. One of his quality measures was quality of life. I noticed that you said that you don't like the term quality of life. I know that that is a catchy phrase that a lot of people use, and I just wondered if you could talk to us a little bit more about how you feel about it and why you made the statement that you made.
DR. ZIMMERMAN: I don't like the term quality of life because when it was adopted, there came with it a fatal flaw. When it was adopted, people immediately made the distinction between quality of life and quality of care. I find the distinction to be the really destructive phenomenon.
I think that quality of life is what all of us think about, whether or not we live in nursing homes. The systematic difference between those of us and people who live in nursing homes is that because of their impairment level, more of the quality of life is affected by the quality of the care they receive, because the care they receive has a bigger impact on the quality of their life.
So I would prefer, instead of thinking about quality of life as something separate from quality of care, to think of quality of life actually as the overarching thing we are talking about here. That quality of life at any point in time can be more impacted by the clinical or medical status of the individual, their functional status, which can be broken down into physical functioning, cognitive functioning, emotional functioning, and what I would call social aspects of their well-being. That is, the interaction between themselves and other residents, the interaction between themselves and staff, et cetera.
I think that kind of framework is actually more useful than simply talking about quality of life. When people talk about quality of life, they are not talking about things like fecal impaction, but they ought to be, and they are not talking about things like whether or not has a pressure sore, but they ought to be. What they are talking about is whether or not an individual has a good interaction with the caregiver, which is very important, critically important, whether there is abuse -- not just at that level, but neglect, et cetera.
So I think it would be better if we tried to capture the dimensions that seem to make a difference to quality of life regardless of whether or not they are social or clinical or medical. I think that we don't lose anything by that.
If you are worried about the interaction between staff, et cetera, fine. You can include that very easily, just as easily as you can by the use of the word quality of life. So, that was soapbox.
DR. IEZZONI: Thank you. That was helpful. Kathy?
DR. COLTIN: I wanted to make a comment about your point on how the data get used and whether they are being used to narrow a network, for instance, as opposed to stimulating quality improvement.
I think that the observation that you made about how some of the managed care companies are using it to narrow their network is really not to be interpreted necessarily as a bad thing in the short term. It would certainly be a bad thing in the long term. The reason I say that is that we are learning increasingly -- and part of the reason that we are interested in looking at the continuum of care across these settings is that the areas where problems most frequently occur are in transitions from one setting to another or from one provider to another.
What managed care organizations have learned is that nurse care managers can be a very effective tool for helping patients through those transitions. But in our case for example, we had somewhere around 1200 or 1400 patients discharged in the first half of 1997 from over 200 different facilities. You can't implement a nurse case management program in 200 facilities when you have one patient every two months in that facility.
So there is an effort to narrow the network. Now, you can do that with information or without information. I think what you are seeing now is that plans, as they try to implement case management, are trying to make those decisions with information rather than just using price.
Now, once they know the network and they build relationships with particular facilities to be long term partners in caring for it, then I think it would be in appropriate to use that information to decide to drop a facility and go to another one. You would want to use it in that partnership mode for a quality improvement purpose.
DR. ZIMMERMAN: Right, I understand the distinction, and I would agree with it. I think if you make that distinction, then I would somewhat hesitantly agree with your point about the fact that it could be used correctly.
But let me give you an example of why I think you have to be very careful about how you would use information like that without the ability to subsequently review. When we did valuation studies of the indicators that we have developed, one of the things we looked at was whether or not the predictive validity of a particular item was higher, depending on the threshold level that was set for a distinction about whether or not somebody is likely to be giving bad care or good care.
Unfortunately, we weren't able to do enough validation work on large enough samples to be able to provide statistically powerful inferences here. I think that is very troublesome. We just don't do enough of it. It's not just ours, there are a lot of them.
But at any rate, the point is that we found that this validity -- and we defined it by having people go in and actually do a very intensive investigation of whether or not there was likely to be care problems, et cetera.
We found that this validity was highly contingent on the threshold level that was being used. That is to say that as you moved up into more of an outlier status, the items were highly valid. They were very good predictors of whether or not there were problems with care. We even used the facility's own staff to determine whether or not they agreed with us. In some cases, frankly I think they were a little tougher than we were.
But at any rate, the point is that if that is the case, then you have to be careful about how you use that information. The thing that was troubling to me was listening to individuals tell me that they were prepared to drop somebody if they weren't in the top half, where if they were below the median, then that is just not good enough for us.
Frankly, it is not clear that those kind of measures -- that we are at a level yet where -- of that precision, I think what Andy talked about and Vince talked about. The level of precision that we have right now is simply not ready for that kind of -- for those marginal distinctions. That is what I meant.
DR. COLTIN: I'd like to follow that up to say also, you are looking at one or two dimensions of the care process.
DR. ZIMMERMAN: Yes.
DR. COLTIN: For an organization to rely only on these quality indicators to make those decisions, even though you have multiple indicators, not just one, I would tend to agree with you, I think that what is fortunately occurring now in the field is, there are more sources of information. There are the standardized results of the surveys that are being done in the states. There is your data. I know Lee is going to talk about some patient reported data as well.
So when you are looking at all of these data, if collectively they all point to problems, I think you are on a little firmer ground to make those kind of decisions.
DR. ZIMMERMAN: That's true.
DR. IEZZONI: Thank you, that was really informative, Sue and David, and Henry, who has just left. Thank you.
Committee, I know that a number of you are going to begin dropping off. So what we will probably have to do is go into a working group, Marjorie, because we may no longer have a quorum. I just wanted to make sure with you that that is okay.
We have an hour, well, a little bit less than an hour -- sorry, guys -- because we are going to end at 4 o'clock. But I know that some people will be leaving before then. I just wanted to say before the last panel something that I should have said at the outset, which is to thank Carolyn for putting this together. I think it has been a really great day, and I really thank you for all the hard work that you have done on putting this together.
Our last two speakers, 15 minutes if you will, and then we'll have a little discussion. But, whatever, we're anxious to hear from you. Karl Kilgore is going to start.
DR. KILGORE: I'm going to if I can remain seated, except for when I stand up two times tops to just slap an overhead up there, if that's okay.
DR. IEZZONI: Patrice, can you help Karl put the overheads up?
DR. KILGORE: This will be the first one, and there will be a pause between there.
I have kind of been changing my talk along the course of the discussion, so I don't cover something that someone else has already covered a little bit more eloquently or proficiently than I.
I'm going to give a quick overview as to what I'm doing here, at least as I understand it. Integrated Health Services -- my employer would think me remiss if I didn't correct the spelling that is shown there -- is first of all, just by way of getting everybody's attention, when Carolyn asked me a couple of years ago now that maybe I'd like to come and talk a little bit about what I've been working on at Integrated, I was somewhat surprised. I was taken a little bit aback, but I said, I would consider it an honor and a privilege. I'd really like to do that.
But Carolyn, you work for HCFA. I work for a publicly traded for-profit health care company. Isn't that kind of like the Vatican calling a cardinals' enclave and then saying, and now for the opposing view, here's Satan? So that is one thing that you really don't need to know about the company.
Why might I have something that the committee might find of some value? Three hundred and about eighty long term care facilities and growing, a slightly smaller number of home care locations and shrinking, about a dozen hospices all over the country, and growing, about 18 PPS exempt long term acute care hospitals and growing, and one big data system that kind of tries to integrate a lot of stuff from there.
There are today about 32,000 patients in an IHS bed. This month, IHS will see about 22,000, 23,000 unique patients in home care. You don't have to have a real dense record to get a lot of data fairly quickly.
It is a good thing. In hearing Sue's talk, a four or five year time frame would be great. The long term in the for-profit side of the business is the next quarterly earnings report.
I want to make a couple of points I want to hit, a concept, by way of encouragement that I think might help, jump to maybe a couple of things that we have done, although I might let that go until the end, that might be of interest and maybe you can throw some questions at me. Also, a couple of recommendations that I would like to get to for what they are worth, which may add to or elaborate on what some of the other speakers have said.
Let me have the first slide. You can draw a picture of the patient that does not have any -- that does not say anything about the setting of care in which the patient -- this is not the model that will -- this is just a model that a data model falls out of. So this is just a model. The little bubbles that you see are data elements and clinical indicators, are elements that the patient brings to their interaction with the therapeutic environment, the distinction -- we don't want to draw too many distinctions here. Demographics are the things David was talking about, things you really can't do anything about to speak of, but they are risk factors, the genetic history, if you will, sex, age, a few other things. Clinical indicators are measures that maybe there is something you want to do about.
A treatment plan is generated from an examination of those two kinds of factors. You come up with a treatment plan, you carry out the treatments, and what comes out of that -- well, treatments cost things and they produce some outcomes.
The areas that connect these things are not meant to be any kind of a causal connection, as Bob Kane always says and as David also said. We don't really know in whatever this non-hospital delivery model we're talking about what causes those outcomes. It could just be the passage of time, do we treat the patient at all, I don't know, we have to see.
Anyway, all of this that you see up on the slide occurs within a social and regulatory backdrop. We need to be able to measure pretty much all this stuff. They are all very difficult to measure, even demographics. We had a lot of talk about what outcomes were, what do we mean by diagnosis, MDS gets at a lot of this, as OASIS does, measuring treatments, very difficult. If you are looking at -- the folks that do it can tell you an awful lot about the results of the various alternatives you have for your cardiac surgery, your drugs, your bypass, your balloon angioplasty or whatever. In some cases, you can find treatments very clearly. Bob Kane made that point quite clearly.
They are hard to measure. Regulatory is up there. The regulatory environments as some of the speakers have alluded to do determine, like it or not, that what we are talking about here is a Medicare certified and state licensed long term care skilled nursing facility. Same way with home care. There are regulations out there that determine what is going on here. What kind of patients can you see? Got to have a two day hospital stay before they get into SNF. What kind of patients can you see, and what kinds of treatments are you going to provide?
I'm not going to jump into the water and then curse it because I can't breathe. That is how it is. We need to know that thing.
A quick sideline, and I won't get sidetracked. Remember, I said we had 380-some nursing facilities and growing? That means added up, every day there is a HCFA annual survey starting at IHS. You don't think that contributes to cost? It does, and so we need to know.
The social environment is one of those things we don't know how to measure. At least, I don't. So you won't see a lot of that.
I started to say next slide and you don't have it. Here is a data repository. Again, it is not the data repository, it is a data repository that tries to implement the view of the patient that you saw before. You know what? This exists somewhere in some form or another. Again, I want to call to your attention, no balloon there that says, this is what comes in from home care, and this is what comes in for long term care. It is possible.
Anyway, demographic and clinical, you got an ATD system, a clinical system, hopefully you will be pulling diagnoses and the results of the H&P. I don't want to get into the details of the data element, but rather the concept.
Quality of care. I'm going to make a distinction here that helps because when you are somebody like me, you just stand back and pull existing data elements from somewhere else. You've got to do this with the existing balloons, if you will.
I'm thinking here very traditional quality of care measures in health care. We are talking about falls, acquired pressure ulcers, nosocomial rate. In the non-hospital based care, an unplanned or return to the hospital. If we are pitching ourselves as an alternative to hospital care, I think it is an important outcome of quality of care indicator to measure that.
At IHS, we have a proprietary system that gets us ORIX for free, that is really nice. But we are switching over to a UI based system along with the rest of the industry, and you have heard about that from David.
The financial balloon, that is a little data warehouse all its own. To act like a business guy rather thana numbers geek, which is really what I am; four years ago I bought a suit and a pair of shoes and moved east, and so now I have to sometimes speak in generalities. A patient level profit and loss system is not impossible. It ain't easy, but you can do it at the patient level. We are pulling right now -- and this is a fairly self contained system, albeit large -- you've got charges coming in from the building system, you've got revenue coming in from the AR, you've got costs coming in. In our case, you've got costs coming in from a Medicare cost report system, and then you do all kinds of adjustments to take data off of cost reports and allocate them down to the patient level, based on all kinds of things we can talk about if there are any questions. A nice activity-based cost system would be better, but we don't have that.
Just continuing on in a clockwise fashion, patient satisfaction. This is -- Andy was talking a lot about the necessity for self report, patient family, loved one. This is a social structure that we want to measure, at least their satisfaction with care. Andy is not here, I don't think. I would agree with him and elaborate in one sense by saying that we need self report data, but we can't ignore the other report data. We need both.
There is an old outcomes guy saying, that is, many a satisfied patient has died needlessly. So we need more than what the patients say. It sounds catchy, I know, but actually it is true.
Continuing on around, health status. This is the kind of stuff we have also heard a lot about. We are just talking here about, yes, mortality and morbidity, symptom reduction, that kind of stuff, functional status, independence, mobility, ABLs, cognitive status, adjustment to disability, these are the things that tend to be diagnosed as specific. You get them.
IHS until recently got them from a proprietary and relatively expensive system. An important lesson in building data warehouses: don't inscribe your data structures in stone, anybody, because you're going to have to change them. If you need to rebuild the whole thing and do a data dump and restructure and reload, -- because we're going from a proprietary outcome to FIM and a couple of other things that yo have heard about -- in the existing PPS and regulatory environment, we can't afford to run two systems that measure health status, and also quality of care, I might add, and a lot of demographics.
We've got OASIS on the home care side, we've got MDS. We recently changed gears, and I'll tell you a little bit about that, to bring that kind of stuff in. Regulatory compliance. We measure surveys. We can roll most of these indicators at an occupational level up to a facility level and use this information to actually improve regulatory compliance as a part of our Key Y program.
What I would like to do is, I'd like to jump ahead and talk about the kinds of things that we have done with this. I'll mention only one of them at any length and then just measure a couple of others by name and see if anything is of interest to you. We can leave that up there.
A couple of months ago, I got a phone call from our lobbyist. We are in Owings, Maryland, right up the street from HCFA, that's pretty good, also very close to D.C.
I hate getting a call from our lobbyist. I'm a two person department plus a secretary. I hate getting a call from the lobbyist, because typically the question is, I'm heading up to Capitol Hill now, I'm getting ready to meet with Senator So-and-so, what data do we have that says that subacute care is better than sex? So I hate to get the call.
The guy fortunately gave me a little bit of time. Listen, there are some discussions going on about PPS. The industry is bugging Congress and the appropriate regulatory bodies because we don't think non-therapy ancillaries are picked up by RUGS. Do we have any data that says right or wrong -- that can eliminate this? I said, I don't know, let me call you back.
Two bubbles up there, one of them says financial, the other says health status. Remember, I said we were shifting over from a proprietary system for measuring that health status thing to an MDS based system? I was prototyping that when he was calling. So I looked it up and I called him back and I said, well, -- I won't call him by name -- I said, we are prototyping something that might be able to help. I can categorize direct non-payor ancillary costs for our patients and I've got some MDS data, and RUGS. I said, the problem is, I don't have it for a lot of patients. He said, how many you got? I said, 8,000. He said, get to work, I need an answer to the question in a week.
This is not something I was tooled up to do. I did. This is a little something we did on quality. When I found out that I was going to be on a conference call with Bob Dee, the economist for ACA, the president of IHS, the lobbyist, the head of government relations and a couple of other big shots, I actually got to work on this.
Just so you know, RUGS in our database is 8,000 patients from 19 facilities, predicted a little bit less than five percent to the variability in non-therapy ancillary costs, direct costs. I felt pretty good about pulling that out in a week.
Another week went by, and I got another call from these guys again that said, hey, they might do something about it; what should we do? We talked to my clinical friends, we wrote a little diagnostic thing, all based on MDS data, produced the diagnostic criteria that explained 24 percent of non-therapy ancillary costs. You put both RUGS and this little diagnostic group that we worked up in a week. You know what? There is predictability in MDS. I'll return to that point a little bit later.
A couple of other things we've done. We did a model, we used this thing to do a model of the --
DR. IEZZONI: Karl, not too much later. Don't return to that point too much later. Return to it soon if you can. We're running a little shy.
DR. KILGORE: Of course. We did a qualifying study, we did a study of the 10-D RG transfer rule, if anybody is interested in that. We did a little study of relating quality indicators to the actual survey process, fascinating stuff. I don't really want to talk about that here, but we could if somebody is interested. Patients who did not have a hospital stay prior to coming to this post something. I can talk about some of that stuff.
A couple of recommendations from my experience, if I may. Number one. Remember, I said from that first slide about demographics might not be as easy to measure as I think? When you start pulling data from all these systems, particularly when you start trying to relate those patients who went from one level of care to another, you got a demographic thing that is like, who is this guy, who is this individual.
I've got to tell you, we have done dumb algorithms, we have tried to bring in some artificial intelligence and do smart algorithms. I will say that the genius who made the change from OASIS-B to OASIS-B1 said, let's add middle initial, that individual deserves a raise because that helps a lot. But the point is, we need a number. I know the issue with confidentiality. There is a lot of Juanita Gonzales's in Southern Florida, and I can't sort them out without a number.
By the way, middle initial doesn't get you all of it. There is one Karl with a K Kilgore who lives in Denver, Colorado. We've got to deal with this confidentiality issue somehow. Give me a number, you can keep their name. The number really helps when you start doing this.
Measuring patients across, another side issue. Title 18 says patients have choice. It is tough to initiate a continuum of care when there is a regulatory discontinuity where -- I have to tell people about all the other options they have to home care. I know that sounds crass, I don't have the time to get into it. But when the legal entities are distinct and the discharge planner has to make all the options known.
Timing in building a data warehouse. Timing is critical. A patient goes from RUGS group to RUGS group as they progress to our system. When it happens -- an admission occurs when it happens, a discharge occurs when it happens. Both come in on a monthly billing cycle. I don't know what to do about that, but it is one of the biggest obstacles to developing an integrated setting (word lost). I raise it because it is a huge problem.
For us to compare cost effectiveness or outcomes from different RUGS groups, we need to be able to know when the treatments occurred during the stay, and the RUGS categories don't change along with the monthly billing cycle.
Two more and I'm done, except for questions, of course. Episode of care. Andy said episode of care should be defined by a fixed interval, not admission to discharge. May I say, in addition to admission and discharge, it is possible -- it is theoretically possible, some of us in previous lives did this study, to match patients in one setting of care with those in another setting of care, pretty much statistically match them, and figure out -- and do a cost benefit analysis. But the period of time is going to be different, and we need some kind of end points to actually do it. I think we need to do both.
Methodological issues. This is geek stuff. I'm a psychometrician by training. That is a measurement guy. The issue of -- to concur with what I think the group says, with a little technical slant, the consistency -- to have uniform measures of those various bubbles that I talked about does not necessarily mean having the identical instrument. Uniform, yes, but it doesn't have to be identical. If we get the methodology down, you use the appropriate measure instrument for the task. If you are measuring something that is -- if the temperature is very cold in Kelvin and something else in Fahrenheit, that's okay, you can translate.
A former colleague of mine wrote an article once whose title was, why does our ability to measure outcomes depend on who holds the copyright of the instrument? It doesn't have to.
Did I actually make a point here? I tried to go fast.
DR. IEZZONI: Thank you. I think before we open up for questions, I'd like to give Lee an opportunity to give his presentation, and then we'll see whether there are any questions. Lee, we do have all your slides.
DR. HARGRAVES: So you do have all my slides, and you can read those on your respective planes or your cars or on the Metro.
If I could make a point, I would start at the end of my talk. One of my problems is, I don't have a wristwatch and there is no clock on the wall. So I would just like to say, doing studies of patients -- what I do is, I'm a sociologist and I talk to patients. I try and take that information and translate it and give it back to people so that they can know what happens to patients when they are under their care.
I am working on a lot of projects. One of them which we did last year, which was following patients who had had a heart attack, they had had an acute myocardial infarction. I think this is an exemplar of the project. It was done for the Foundation for Healthy Communities in New Hampshire. I was the person who was fortunate enough to -- I am a survey designer, and I designed surveys for them for that particular project.
Essentially, my message is, if you want to know what is important to patients, there is a very simple method for doing that. You go in and talk to patients. We use focus groups, we use intensive interviews and sit down and say, when you had your heart attack, what was important, what did you need and when did you need it.
That is my first point, talking to patients is key, and developing standardized survey instruments so that you can follow someone and you can make comparisons over time.
We actually surveyed in this project patients one month after they left the hospital. We asked them what happened while you were in the hospital, did they talk to you about this, did they do that. We asked them about, did they talk to you about what was going to happen to you after you left the hospital. We asked them, did they talk to you about home care services. I was actually surprised to find that a lot of folks after a heart attack receive home care.
I was simultaneously developing survey instruments for doing hip replacement, following hip replacement patients over time. I thought, that is very simple, actually. Home care for someone with a hip replacement makes sense. You go back home, you might need help in learning how to deal with life at home. But actually, a surprising number of patients who had had heart attacks -- we asked them about their home care, which was the major part of my presentation, was thinking about how we develop a home care survey.
We received -- we being folks at the Home Care Institute where I am located, folks at the New York DNA -- developed a survey instrument to measure home care experiences. We talked to patients about that. We actually followed them over time, and I think if you can't follow someone over time -- that is the way I live my life, over time. I don't live it in little snapshots, and I think that is the way that we should be measuring things.
So we follow them for one month, three months and 12 months. We ask them some questions which were very similar over time and some things which were different, depending upon where they are. It doesn't make sense 12 months out to say, by the way, could you recall that in-patient experience you had a year ago.
We asked them different questions. We asked them questions about, what is it like receiving care from multiple providers, what we like to think about as the continuum of care continuity. And of course, for people like me it is really fun, because we have a data set and we can make these linkages. We were also very fortunate because we asked the patients, do you mind if we take some of your administrative data and link that. So we have discharge data, we know a little bit about their health status when they left, what kind of care they received. For heart attack patients, you might want to know whether or not they had -- and I'm a sociologist, not a medical doctor, so if I don't pronounce angioplasty properly, forgive me, but you might ask them whether they had that or bypass graft or they had medical treatments. We can break people into different groups.
It makes a lot of sense to try and do subgroup analysis, because one of the difficulties in trying to talk about a concept -- and I pick up on the fact that there is some fuzziness about a concept, and a concept that has no edges is not a very useful concept.
One way to deal with that is by sampling and stratifying people into different groups and making comparisons among people who had a heart attack with other people who had a heart attack. I don't compare my hip replacement patients to my heart attack patients; it doesn't make sense for a number of reasons. Their projected illness is quite different. I like to say it is not survey science, but it is an in joke only among people like me.
And essentially, I think the other thing is, when I talk about measuring home care services, we actually ask a number of questions that relate to different parts of care. There are some things which -- I think measuring home care services is very difficult, because in-patient, it is all right there and it is captured.
I also work on a project called the consumer assessments of health plans project. While I was doing home care development, we were also working on that project, which was great, because I could borrow from one and use that in another place. We actually took our home care survey and tried to make it fit. We asked how often did these things happen, because that is the way the home care is provided. How often did your nurse come when he or she said she would? How often did the therapist tell you about your exercises? How often did they push you too much? We actually borrowed and tried to design survey instruments that both fit the care that the person was receiving and kept it within context.
I don't know if that is somewhat helpful. If you have questions, I could probably talk more. But I think our time is pretty much gone.
DR. IEZZONI: You have a few more minutes.
DR. HARGRAVES: It's tough to do this at the end of the day. I was sitting here thinking, when I teach in college I do 50 minutes three days a week. And now it's like, you folks have been here doing three credits in a day.
DR. IEZZONI: Well, Vince, do you want to hit Lee with a question?
DR. MOR: I've been waiting to ask you questions all day long. The person experiences the focused questions that you guy have been developing over time. As it applies to home care, how did you find that people were able to recall these fairly specific event year time things? I'm really curious about that.
DR. HARGRAVES: When we develop patient surveys, recall is a key factor. I think that is one criticism. People say, how can patients remember that?
One thing we did was, we designed the questions to not say -- you can't ask people specific questions. You can't ask them, how many times did they come 15 minutes late, 20 minutes late, 40 minutes late. You basically say, how often did they come late. So one thing is not to try and ask people questions which they cannot answer.
We always try to refer to the format. How often did they come when they said they were? And that basically removes that, trying to make people recall. Survey methodologists have known for years that people cannot deal very well with time related events. They do what is called telescoping.
For example, if I ask most people in this room, how long has it been since high school, you would give me a wrong answer. For me it was longer than five years ago. But in my mind's eye, it just seems like it was yesterday. So you don't ask them things which they can't tell you. And you ask them -- but then again, you ask them very specific questions about the nurse. We asked them, for example, how often did you see the same person, which we found is a very key part of home care services, seeing the same provider over time, having the same therapist come and not a different therapist.
I can answer that question, that is an easy thing. I know that it was a different person every time. Not only that, but that is very much related to my correlation with my satisfaction, which is one of the things we find. Sure enough, if you see somebody more, you feel comfortable with them.
Home care services -- in many ways, a patient is more vulnerable in the home than anywhere else. In the hospital you have a lot of people taking care of you, in the home -- I'm sitting here all day, waiting for someone to come, because I have this condition. So you ask them questions they care about and which they can report about.
I don't know if that is all of it. The recall aspect is also -- you ask them questions close to the event.
DR. MOR: Most satisfaction or anything like satisfaction questions have a high degree of response acquiescence. Are they doing about as well as they could, or everything you expected, et cetera. I know there is something about the way you guys phrase questions that somehow or another minimizes that. How was your experience in doing the home care survey?
DR. HARGRAVES: Well, basically people try and divide patient surveys into two groups. There is the people that do patient reports and there's the people that do patient satisfaction. Years ago, I would say satisfaction is too fuzzy and it doesn't mean anything. I've got a data display that shows, if you compare 15 home care agencies, everybody is satisfied, there is very little variation.
So one of the problems that you have with that is, it is too fuzzy. So you try and ask people things about -- tell me about what happened, and not how you feel about what happened. I think because both of those pieces of information are valuable. If I need to understand -- when someone leaves care and they say that that care was excellent, what you want to know is, what criteria did they use to judge excellence. If you have reports about what their experience was like, how often did you see the same provider, how often did they come when they told yo they would come, how often did they give you information in a way that you could understand, you could make correlations between those two, which is actually my little chart with the bubbles and arrows, what explains peoples' experiences in home care.
There used to be an advertisement in my local market area: it is the nurses. I didn't want to do a plug for nurses, but it is the nurses, and the whole composite of things about that. Trying to ask questions that remove someone's natural tendency to say everything was okay.
One of the difficulties in measuring health care from the patient's perspective is, people for the most part are very pleased with the care they get. To have someone come into your home and deliver health care is a wonderful thing, in some respects. It reminds me of -- in a nostalgic sense, I remember reading once that doctors actually used to come to peoples' homes.
DR. MOR: They can't get a parking space.
DR. HARGRAVES: I don't want to talk about parking in Boston, but now nurses come to your home, physical therapists come to your home. I have a relative who had a physical therapist come after hip replacement. So if you ask them, are you happy, are you satisfied, they will say yes. But then you say, tell me more, tell me, did they tell you how to get around your house, did they tell you what kinds of things you may want to change in your house because you are just coming home from hip replacement surgery. No one told me that. So there are certain aspects of care that are related to safety, to what I would think of as quality. You can say, rate your care, pick a scale, any scale. They are all skewed towards the positive end of the distribution. If you ask them about events, I think that is the key, is asking them about events. That is what we try to do.
DR. MOR: One last question, if I may. How do you deal with people who are not able to respond due to causative or other kinds of problems?
DR. HARGRAVES: Actually, my list of the difficulties of doing patient surveys. You need to develop a survey that works both as a mailed out/mailed back instrument and a telephone instrument. Not everyone likes to read instruments. They respond much better to the phone, particularly when you think about someone who is older. You definitely want to pick 11 point type, and maybe even then that doesn't work. You have to use the telephone. Not everyone reads English.
You might actually think about the fact that your population has a cognitive impairment. One of the difficulties that we found in surveying people for rehabilitation care after they have left the acute care setting and they are in rehab, we find that we might actually have to talk to someone else, what we would call a proxy respondent. There is a whole difficulty in proxy respondents.
As I said, patients are very happy with the care they get. Proxy respondents are standing on the outside looking in, and they are not as pleased. So I am comparing 10 hospitals and one hospital, 70 percent of the respondents were proxy respondents, the other hospital, 20 percent were proxy respondents. I don't want my institution or my home care agency to have to think bout that. But you have to do it, you have to allow proxy respondents. For long term care, if we ever got into that discussion, the proxy respondent is the respondent oftentimes.
So you have to allow for that. You have to make sure that you ask -- a frightening thing for me is, I have asked the questions for years and I never bothered to say who is filling out this questionnaire. It is a key thing to ask. I actually analyzed reliability, where you ask someone a question and a week later you ask them the question again. Then I got my reliability estimate. Ninety percent of the time you got the exact same answer, and for a couple of items, I was like, why did only 50 or 60 percent -- if you strategize your answers by those who are the patient and those who are the proxy, the proxy is actually much less reliable. Why? Because we didn't say, were you the same person I talked to last week.
I'm just thinking of the experiences I have had, where different people care for a person over time. In some settings, the family members and the loved ones and the neighbors and the friends are the care providers, and it is a triad. It is the patient, provider and someone else who -- and usually they have signup lists. I'm coming over to visit your mom on Tuesday, and we do that. God forbid, that was the day I called on the phone and asked about my survey, and then I call back a week later. So you need to know about proxy respondents.
DR. IEZZONI: Are there questions for Karl or any additional questions? Everybody is kind of packing up. It doesn't mean we aren't enthusiastic.
DR. HARGRAVES: I know where some of you are going.
DR. LIU: IHS is an integrated system, and your data system essentially covers the entire continuum of the services offered by IHS. Have you looked at this whole question of attribution as patients go through your system? Have you made any attempts to study what level care may be called for at station A before the person was transferred to station B, or was that sufficient, was it not resulting in the transfer? What kind of dynamic analysis of patient flow have you looked at?
DR. KILGORE: That is the $10,000 question. I'm just trying to decide if the -- we'll go with the short answer. We have -- in our scientific studies, no, or anything that we would call scientific studies. We have however in a couple of regions of the country developed intake centers essentially, that develop a relationship with discharge planners in hospitals, physicians' offices. They get the call without any preconceptions about whether this patient would best be served in a home care agency or in a SNF.
It is not scientific, and it is based in part on the conviction that referrals come about as a result of relationships, primarily. But we do know -- I think we know as an industry that the biggest predictor of where -- the setting of care to which a patient goes at the end of a treatment stay is -- the biggest predictor of whether they go home or not is if they have a home to go back to.
That is about the most concrete thing that I can say. We have not at IHS had the time to do it. Our outcome database is not as rich on the home care system yet to do it, and it probably is not going to be, unfortunately.
DR. IEZZONI: It looks like I am about to turn out the lights here. I don't know whether an N of one qualifies as a working group, but can one person be a group? Marjorie, as the designated federal official?
DR. GREENBERG: Working sessions are for gathering information.
DR. IEZZONI: Okay. Well, I think we will stop for the day then, and thank Karl. Thank you for coming over from Maryland and thank Lee for coming down from Boston, and thank Carolyn again for putting together a great day.
What we will do, the committee who is left, we will talk about this on February 2 and decide where we want to go. Thank you.
(Whereupon, the meeting was concluded at 3:50 p.m.)