Health Care Financing Administration
7500 Security
Blvd.
Baltimore, Maryland
Welcome and Opening Remarks -
Liza Iezzoni, MD, Chair, PSI Subcommittee
Barbara Starfield, MD,
Chair, HDNSS Subcommittee
Overview and Issues for Post Acute
Care: Health Care Financing Administration Panel -
Tom
Hoyer
Steven Clauser, PhD
Helene Fredeking, Health Care Financing
Administration
National Perspective on Post Acute
Care Panel
Marilyn Field, PhD, Institute of Medicine
Karen
Seitz, Veterans Affairs
Marc Freiman, Agency for Health Care Policy
and Research
Panel on Research and Demonstrations
Korbin Liu, ScD, Urban Institute
Chris Murtaugh, Center for Home
Care Policy and Research
David Rabin,MD, Georgetown University
Medical Center
Nancy Miller, PhD, Health Care Financing
Administration
Rehabilitation, Assisted Living and Performance
Measurement for Long-Term Care Panel
Bill Buczko, PhD, HCFA
Catherine Hawes, PhD., Meyers Research Institute
Panel on Current Collection and Data Standards
Issues
John Morris, PhD, Hebrew Rehab. Center for Aged
Ye-Fan Glavin, PhD, New Health Management
Nell Wood, Maryland
Hospital Association
Peter Shaughnessy, PhD
Kathryn Crisler, RN,
MS, Center for Health Services and Policy Research
Terry Moore, RN,
MPH, Abt Associates, Inc.
DR. IEZZONI: I think we would like to get started because as I said, we have a full afternoon, and we might as well get started. Dr. Buczko is here.
So, do you want to start?
DR. BUCZKO: Yes. There has been a demand that has been going for PPS for inpatient rehabilitation due to the upward growth in the number of rehabilitation facilities, number of patients using inpatient rehabilitation and dollars spent on inpatient rehabilitation.
Part of this is due to the weaknesses of the TEFLA(?) payment system which basically provides an average payment within certain specific incentive limits, and, also, is basically an average that is non-case-mix adjusted.
An alternative that has been specified, discussed as early as the mid-1980s and by an earlier study by Rand is to use a functional status measure as a case mix measure for inpatient rehabilitation and a possible base for prospective payment, and what we sought to examine in the research that HCFA has funded on this is to determine if we can use this as a feasible alternative for a prospective payment system.
One issue that was important here was that we did not mandate functional status data as something rehabilitation facilities were to send in to us. Therefore we, in effect, for any work that we would be doing on this had to outsource for our data and we had to set up an arrangement with people at UDSMR, at SUNY, Buffalo to obtain data from their archives in order to proceed with any study of inpatient rehabilitation, and one thing we ended up doing to this end was to let out a contract to the Rand Corporation to examine the feasibility of a prospective payment system using the FIM and using the FIM FRGs.
DR. IEZZONI: Could you just define the acronym FIM, please?
DR. BUCZKO: Functional independence measure and FIM FRGs are functional independent measure, functional related groups.
I have got my first overhead up there, and here the FIM contains 18 categories, each scored from 1 which equals maximum dependency to 7 which equals maximum independence. Thirteen of these items form a motor subscale. Five items form a cognitive subscale, and they incorporate ADL-type measures that are components of the Barthelle(?) index but it is more sensitive and more inclusive, and it is a measure of actual performance rather than potential performance, and the FIM database, first off the FIM is maintained and marketed by the UDSMR in Buffalo which maintains a database which extends pretty much throughout the 1990s with data from member hospitals, and the next slide shows the increase in data submissions for the UDSMR database.
On the slide I have here they have nearly 300,000 individual discharges in their USA comprehensive medical rehabilitation database. One alternative to the FIM when we were considering this project was the patient evaluation and conference system, PECS which is a much larger database, much larger instrument containing 115 categories scored on a similar range as the FIM, covering 16 sections.
MR. BLAIR: Evaluation and what system?
DR. BUCZKO: Patient evaluation and conference system.
DR. IEZZONI: Yes, the word "conference," I think is a little bit confusing to people.
DR. BUCZKO: It is. It is basically an instrument used to track patient status, set care goals and monitor outcomes. It has not been used as a summary measure that much. However, it covers a remarkably broad range of activities, and here we had basically a trade-off between on one hand data availability and conciseness versus a global measure that many rehab hospitals use, but there was no database that could be comparable with the database that UDSMR has.
So, on that note we focused in on the UDSMR as a database of choice for this project, and as our measure that we would be, a classification algorithm, we are going to be using the FIM-functional-related groups developed by Margaret Steinman at University of Pennsylvania.
First, I will do a quick description of the length of stay FIM FRGs. I don't know if folks are aware of Dr. Steinman's activities, but she has created a number of individual variants of FRGs for various purposes.
The length of stay FRGs which will be what we examine here and extend here in the Rand contract work has as dependent variable length of stay, and it is used for respective case management and payment, and they use the impairment category as a start point and patients were grouped using classification and regression tree or part methodology by admission FIM motor score, cognitive score and age, depending on the condition.
Version 1 which Dr. Steinman finished using 1990 data used 18 impairment categories to form 53 groups, and version 2 using 1992 data used 20 categories to form 67 groups.
Next?
Two more recent bits of work from Dr. Steinman are the gain FRGs which use FIM gain as a tool for case management and payment, and create 74 groups, and thirdly, she has, also, worked on the discharge motor FIM FRGs which has as its dependent variable FIM motor skill score at discharge, and she has a classification system that works with 139 groups on this, and she has suggested that the discharge motor FIM FRG and the length of stay of FIM FRGs can be used in concert to provide an overall quality and payment system that will deal with both the cost and patient management end of this and with managing outcomes.
I have as an example the impairment categories used. The slide I have has only 18 categories which represents the earlier version of the rehabilitation impairment categories. Dr. Steinman has added for her later work two classification codes for Guillian-Barre and has subdivided major multiple trauma No. 17 into two categories depending on whether it is a spinal cord injury involved or not.
Next?
Here is an example of how stroke is subdivided under the FIM FRGs where first the patients are divided by motor score and secondly they are divided by age. For some of the impairment groups cognitive status, also, is an important measure that is used in classification algorithm.
Next?
When we finalized the data use agreement with the UDSMR we went forward on a scope of work with the Rand Corporation who were our contractors for the solicitation we had for defining a prospective payment system for inpatient rehabilitation, and the work that we had here was kind of a period piece in that it mirrored what we would expect a rehab prospective payment system to look like if it were similar to what we had set up for inpatient rehabilitation, and as such a lot of the thinking we put into this project was basically prior to going forward with integration on a comprehensive long-term care system which dealt across sites of care.
The three components here are evaluating suitability of the FIM and FIM FRGs for use in prospective payment, replicating the construction of the FIM FRGs using linked UDS MEDPAR data for 1994, and simulating payments using an inpatient PPS-like model and obtaining payment to cost ratios to check our impact.
In our evaluation of the FIM and FIM-FRGs our panel gave overall support for the FIM as a functional assessment instrument and suggested possible improvements for the cognitive scale and this was the most pointed item that they had come up with.
They had suggested that a study of FIM reliability should be undertaken and they were generally supportive of the RIC and FRG structure in the FIM FRGs, and they felt that FRG splits within RIC should be limited to ensure adequate distance between the two groups as a break on one aspect of possible gaming of the system, and lastly, and what may be the most important in the long run for any post-acute care measurement the effect of comorbidities on classification and variance explanation should be examined.
For replication of the FIM-FRGs the FRG groups created by Rand explained 35 percent of the variation in length of stay and 32 percent of variation in Medicare costs, and this was fairly similar to what Margaret Steinman had obtained in her two studies where her explained variation length of stay was about 32 percent, and the groups were structurally similar. The FIM FRGs obtained did not capture comorbidity effect.
Thus, Rand decided to develop a separate adjuster for each impairment group category to augment the FIM-FRG classification, and in terms of representativeness of the data the analysis used data from 24 percent of hospitals treating Medicare rehabilitation patients, accounting for 40 percent of 1994 Medicare rehabilitation discharges.
If we were to do this study with say, 1996 data we would probably cover 50 percent of rehabilitation discharges, and in terms of representativeness our last slide will be payment simulation where we found that 50 percent of patient level costs and 60, 65 percent of hospital level cost variations were explained, and in general, general payment to cost ratios were pretty close. Most hospital groups' largest effect found for certification year, hospitals with recent base years were minus 10 percent. Hospitals with more distant base years were generally with positive effects. Payment factors of interest, area wage index and DSH were important but not large urban location or teaching.
Outliers found to be necessary for high-cost cases and payment reductions below very low cost case improved payment equity, and transfer payment per diem was recommended with an added half-day payment for the first day.
In short, we found that the PPS system was workable and that this provided a viable option for proceeding with prospective payment system for patient rehabilitation.
DR. IEZZONI: I am going to take the prerogative of the Chair and do a quick question for clarification because, Dr. Buczko, I am a little confused. This morning we heard clearly from the first panel that you were not, that HCFA was not advocating this particular system, that in fact for rehab prospective payment you are thinking about doing RUGS or some RUGS analog. So, I am just a little confused.
DR. BUCZKO: As I said, this was a period piece which represented where we had gone on this project, and what we found in our results.
As such it represents possibly an alternative that has been bypassed in the current debate. The issue here has been how is it best to move from say the RUGS down to rehab or employing FIM concepts if not FIM instruments in SNIFs or home health settings.
UDSMR has worked in that area, and we are currently doing work on the MDS to test its applicability in the rehab area.
DR. IEZZONI: So, just to summarize to make sure I understand, you have done the extensive work on the FIM FRGs but you are not going to use those. Instead you are going to try to develop something new that will be based on a RUGS type of model, and you hope that you will be able to implement that.
DR. BUCZKO: That from what I gather is the direction in which CHIPS is proceeding.
DR. IEZZONI: Okay. All right, we might return to that later, but I just wanted to clarify that because I was a little confused about the emphasis of the presentation given what we heard this morning.
DR. BUCZKO: As such, the work represented here has been put at least on the back burner.
DR. IEZZONI: The work has been put on the back burner? Okay.
Dr. Hawes, thank you for coming. We are looking forward to hearing from you.
DR. HAWES: I very carefully stayed up very late playing with a new toy called Power Point which I managed to then leave at home, and HCFA has kindly pulled it out of the fire by downloading it to their machine. I am going to talk about assisted living and residential care.
For those of you who don't know what assisted living is, you are in really good company because it varies from state to state and from facility to facility.
What we do know is that in 1991, in the national health provider inventory there were an estimated 34,000 licensed residential care facilities which included some assisted living facilities.
Residential care is known by more than 30 different names across the country. So, it is a little hard to get an accurate count. There are about 600,000 beds in these licensed facilities, and there are about between 12 and 20,000 or depending on whom you ask 40 or 50 thousand unlicensed homes.
The unlicensed homes I had always thought were small, mom and pop underground operations, but actually in many states there are large retirement communities that provide services or assisted living facilities that aren't required to be licensed.
So, unlicensed covers a broad spectrum, but it tends more to be the large retirement communities and assisted living facilities, and this is an area that is undergoing rapid growth which is one reason to be interested in it.
A second is that there are 500,000 frail elderly or more living in these kinds of facilities now. I am giving you data that we derived on the next overhead.
It shows some studies sponsored by the Assistant Secretary for Planning and Evaluation, Department of Health and Human Services with additional funding by NIA, AOA and HCFA.
We have two basic studies. One is a 10-state study of board and care homes in which we linked data on resident level health and functional status indicators to Medicare claims files, and we did the same for a 10-state sample of nursing home residents, and we compared their utilization of Medicare, and we are currently conducting for ASBI a national study of assisted living for the frail elderly.
It is a national probability sample of facilities. So, it will provide generalizable results, but it won't provide them for another few months.
In terms of resident characteristics we had a little trouble when we downloaded to Power Point 4.0, but basically the story that this tells is that in terms of age nursing home residents are much older. The group in the middle are licensed board and care homes. The group to the -- this is really bad for somebody who doesn't know left from right. The group to the right are unlicensed board and care homes and assisted living facilities.
Basically you have a much older population in nursing homes which is the thing to the far left, and you see the under 65 population is a little over 22 percent. That includes people with chronic mental illness and developmental disabilities who live in a few homes specially licensed for them but by and large in a mixed population with frail elderly.
Nursing home residents are older. They are more likely to be female, not surprisingly and somewhat less likely to be Caucasian.
In the next slide I think that we look at their activities of daily living and cognitive impairment.
More than two-thirds of the residents in nursing homes have moderate to severe cognitive impairment.
DR. COHN: Is that 10,000 or --
DR. HAWES: No, that is the percent of residents in facilities. It is an aggregate figure. So, almost 70 percent of residents in nursing homes have got moderate to severe cognitive impairment whereas in licensed and unlicensed board and care homes we are at about 40 percent, and that includes assisted living facilities.
The next sort of grouping over is people who receive assistance, hands-on physical assistance with zero to one activities of daily living, and basically what you see is that is practically no one in nursing homes. It is practically everyone in assisted living facilities, and it is, I have forgotten, I think about 68 percent of the residents in licensed board and care homes are only getting help with bathing or maybe dressing, and then you see sort of the inverse of that.
Sixty percent of the residents in nursing homes are receiving assistance with six to seven ADLs whereas 5 percent to 3 percent of residents in residential care. So, it is a much less functionally impaired population although there is significant cognitive impairment.
So, what would you then expect about Medicare claims which is the whole point of this which is do they use post-acute care, and in the next slide you will see the percent who use any Medicare services during a 12-month period.
What you find is that the highest users are the residents of licensed board and care homes and that is true whether it is inpatient care which is the left hand most grouping, SNIF care, home health, and it is slightly lower for physicians, but it is actually not statistically significant.
Much higher use of inpatient care, and one of the reasons is when you interview the operators, and you say, "What do you do if someone has the flu and needs temporary nursing care?" In one-third of the cases they discharge them to the hospital ER. If they need care for more than, a really cost effective way to take care of the flu. If they need care for more than 14 days, 80 percent of all the operators said that they would discharge the resident, and half of them said that they would send them to the hospital or to the VA.
So, I think part of what you are seeing is use of the hospital to provide really low-level nursing care. You, also, see though that they are picking up Medicare payment for home health services in residential care settings as you can see from the home health use.
So, the next slide then shows the expenditures by resident for those who had a Medicare claim, average expenditures and again, although the thing that is interesting here is while they are less likely to use services, people in unlicensed board and care homes and assisted living facilities have the highest for hospital inpatient care. Their median expenditure for inpatient care which includes ER in our lumping of things was $6376 compared to nursing home residents at $5700. Pretty much true across the board, people in residential care are heavier users and more expensive users of acute and post-acute care services.
So, what does that tell us about data needs? I think there are two things. One is the national health provider inventory is no longer going to include residential care in its listing of facilities, and I think that is really wrong.
It is unbelievably difficult and people need a lot of therapy after they have looked for licensed and unlicensed facilities, but I think it is really important to keep account of this growing, I don't want to call it a modality setting of care because the modality really differs a lot.
The second thing is I think you do need standardized data on resident status to see who is there and how it is changing. We saw that there was a significant increase in the level of impairment and acuity of needs among residents from 1983 to 1993, and I think that is really growing.
If you look at the next transparency, they have high use of acute and post-acute care. Virtually nothing is known about cost or quality of care. We sort of know something about the per diem payments but most people haven't linked it to all of the other costs like home health care and inpatient hospital care that is associated with the care of residents at these facilities.
There is, also, really rapid growth in the assisted living market and oversaturation of some markets. What that means in a proprietary industry is that in order to keep their occupancy up they will have to either accept people who are more impaired or let them age in place so that they become more impaired.
So, I think this is a setting where the acuity of residents is going to grow and then last but not least the state and data needs.
The states have really been asking, several of them for some kind of uniform resident assessment system, particularly the states that want to develop an integrated long-term care system so that they know who is getting home and community-based care, who is getting Medicaid home health, who is getting nursing home care, who is getting residential care. How do those residents or those clients compare in terms of health and functional status, cost and outcomes? They want, not only a uniform system of items, but they want consistency with the MDS because nursing home is still a huge chunk of their Medicaid budget, and they want to be able to compare to that population
There are several states that are using an MDS-compatible resident assessment instrument in residential care, New Jersey, North Carolina, Maine, Illinois, and we are told that Oklahoma is considering it, and some of the national chains are, also, moving to a uniform system so they can do benchmarking for outcomes and quality assessment.
In these states the data from the RAI for assisted living or adult care is being used in case management to identify clients who have heavy care needs and are going to get an enhanced Medicaid personal service payment. It is being used in service planning and to evaluate the adequacy of the service plan. It is being used in payment.
We have done two timed studies, and there will be a case mix payment system in at least one of the states, actually in two of the states.
It is being used in quality assurance. In Maine they are giving feedback on outcomes using the uniform data to facilities, and that is it.
DR. IEZZONI: Thank you, Dr. Hawes. That was revealing. I think some of us didn't know all of that.
Are there any questions from the Subcommittee? Barbara?
No? Okay, Elizabeth?
MS. WARD: This question is probably because I was reading and writing, and you probably said this. How many places is your instrument you described being used?
DR. HAWES: It is being used -- New Jersey developed its own MDS nursing home compatible assessment requirements. The RAI for assisted living in adult care which is desperately trying to keep up with demand is being used in North Carolina, Maine, Illinois and then I guess Oklahoma is meeting this week to try to decide.
MS. WARD: How is that happening? How are those particular states choosing to use this, and how is it getting distributed and sponsored?
DR. HAWES: Haphazardly, I am sort of embarrassed to say. We developed it for years in Maine and North Carolina. We didn't really think about demand in other states. We don't really have funding to do anything with it. So, basically people hear about it at a conference or they hear about it from somebody at HCFA, and they call up, but a lot of states are trying to move toward requirements about assessment and care planning simply because they are seeing a real shift in the acuity of the residents.
MS. WARD: Thank you.
DR. IEZZONI: Kathleen?
MS. FYFFE: A quick question. I might have not heard you correctly, but did you say something about home health care payments going to residential facilities?
DR. HAWES: Going to residents in residential care facilities.
MS. FYFFE: Okay, so, if you have a room and board facility with four people in it they could be getting --
DR. HAWES: They could be getting home health, Medicare home health or Medicaid home health. We found in a 2-week period 14 percent of the residents were using home health service provided outside. There are practically no nurses in these facilities. Only the really large ones have a full or part-time RN or LPN. So, if they need any kind of care or monitoring it is purchased.
DR. MOR: A data question, Catherine. I have never seen the data on claims match. Was that 60 percent of people in board and care facilities had at least in the last year a home health or what is the time period there?
DR. HAWES: No, that was over a 12-month period.
DR. MOR: That was a very high number. So, it is not necessarily while they were in the home although given their length of stay it is likely to be, yes or no?
DR. HAWES: Yes. I mean that is one of the problems with the data. We had a cross-section of residents. We had longitudinal claims data, and so you made certain happy assumptions about where people actually were over that full 12-month period. You were pretty confident about the 6 months before the midpoint and prayerful about the others.
(Laughter.)
DR. IEZZONI: That is interesting. We might want to adopt that position in our research.
DR. HAWES: It just shows you how much data is needed and how little is available that this is interesting at all.
DR. IEZZONI: Are there any questions over on the side?
May I return then to the rehab prospective payment system and specifically the question of data gathering burden for implementing a prospective payment system for rehabilitation? You said that the FIM isn't currently mandated to be required to be reported to HCFA.
DR. BUCZKO: Nor is anything else.
DR. IEZZONI: Nor is anything else. So, my question to you is, okay, if you are not going to use the FIM which sounds like it is kind of a single scale with, it is a single number that a place would report for a patient or a cognitive number and then a physical number, right?
DR. BUZCKO: Right.
DR. IEZZONI: What is required to report RUGS? To produce a RUGS-based system, how much data would the rehab facilities have to report?
DR. BUCZKO: They would probably have to, they would certainly have to report comparable cognitive and motor measures. They would have to, it would be useful to report comorbidities and some additional --
PARTICIPANT: He didn't understand your question.
DR. IEZZONI: Okay, my question is what is the data-gathering burden to produce a RUGS-based prospective payment system versus a FIM-FRG based prospective payment system?
DR. BUCZKO: In terms of actually what you need to get from where we are now to a fully operative system you would first have to begin data collection, and basically you look for 6 months of --
DR. IEZZONI: No, wait. What I want to know is are the data elements that are required by RUGS more or less in terms of the number and scope than the data elements that would be required by FIM FRGs.
DR. MOR: It is close to 100 data elements.
DR. IEZZONI: Right. See, that is what I was thinking. So, an MDS, it is basically kind of like, is the MDS required to produce RUGS?
DR. MOR: No.
DR. HAWES: Not all of it.
DR. IEZZONI: Not all of it, but a lot of it, 100 data elements?
DR. HAWES: Not even a lot of it.
DR. IEZZONI: Okay, my simple question is is the data-gathering burden that would be required for RUGS-based rehab prospective payment system greater than for the FIM- FRG based prospective payment system?
DR. BUCZKO: It would probably be greater, given the size of the instrument.
DR. HAWES: But, also, the amount of difference or variance in resource use that you would predict is greater because it does the same thing that we talked about in terms of --
DR. IEZZONI: Do you know that for sure?
DR. HAWES: -- catching comorbidities.
DR. IEZZONI: Because being able to explain 32 percent of the variation is a huge amount of variation that for a prospective payment system that is actually quite a bit to be explained, and so, I guess that is why I asked my question after Dr. Buczko gave his talk because I wanted to know if there was a comparable R squared that we could hear from the RUGS system yet, and it doesn't sound like there is.
DR. BUCZKO: I have not been involved with the RUGS development.
DR. IEZZONI: For rehab facilities.
DR. BUCZKO: For rehab facilities, but I don't think they have gotten that far on that yet.
DR. IEZZONI: Okay. I guess it is really confusing to me, but yes, Simon, you are wondering why I am confused?
DR. COHN: No, I am actually feeling similarly confused, I guess. I must say I apologize because I missed the very beginning of the presentation, and I guess I find myself confused. As I understand, Dr. Buczko, you are actually a representative of HCFA, and is that correct?
DR. BUCZKO: Yes. I was the project officer.
DR. COHN: Oh, I see, okay. I was trying to figure why that was being presented when they are talking about moving to --
DR. IEZZONI: That is what I was trying to clarify earlier because they were presenting something that has apparently been bypassed and decided that they are not going to do, but it sounds like something that has a great R squared and lesser data collection burden. So, I guess I am pretty confused.
DR. BUCZKO: Carolyn asked me to do it.
MS. RIMES: We wanted the presentation to include a discussion of FIM.
DR. IEZZONI: I am very glad that it did because it has been very instructive
MS. RIMES: I believe that even Tom Hoyer would say that this has not reached concrete shoes yet. I mean we have not buried it in the ocean. I think most reasonable people want to discuss the future and what will be most useful and most comfortable to use. There is good reason, and, Bill, I appreciate very much your doing this.
DR. IEZZONI: We are very happy to have heard it because it is very interesting. I mean as I said, I think that the --
MS. RIMES: I am sorry I missed that part. I didn't mean to be out of the room.
DR. IEZZONI: I think that it sounds like it was a great study that you were project officer on, and so, it is just kind of a shame that.
MS. RIMES: Actually it was exciting.
DR. IEZZONI: Yes, it was exciting. So, it is kind of a shame that it is not being considered anymore although again since I don't know all the details I may be speaking out of turn.
MS. RIMES: It is tough in here to say, "Not on anything."
DR. IEZZONI: Okay.
DR. MILLER: I was going to speak to this.
DR. IEZZONI: This is Nancy Miller, again, who was on the panel right before lunch.
DR. MILLER: Right. I mean this actually was something that we discussed as part of the post-acute care initiative, and we were looking at again kind of like our short-term strategies and the long-term strategy, and the long-term strategy, again, was to try to go to a more kind of unified or integrated system, and the discussion in those meetings, at least was that essentially if we went to like the FIM FRG system for rehabilitation hospitals in the short run we basically would be putting up three separate data systems, what we are doing in home health, and what we would be doing in rehab and what we would be doing in SNIF and that would take us kind of further away from any kind of unified approach, and the thinking was that if we built the rehab prospective payment system off of an existing data set we would be kind of hopefully working toward a more integrated rather than a separate system, but as Carolyn said, I mean those are discussions and kind of decisions now. They could be subject to change down the road.
DR. MOR: Just one of the other issues that I know is of interest from the point of view of comparing not only what Nancy said is three different systems, but you, also, have three different service providers with very substantial overlap in the characteristics of the patients that they serve. To the extent that you have different systems it becomes very difficult to say how much of an overlap there is. There is a lot of rationale for compounding.
DR. IEZZONI: I appreciate that. I am just thinking data-gathering burden and the fact that it seems that there are these short-term and long-term strategies, and you have got a system that has a low data-gathering burden that could be the short-term strategy, given that there seems to be a project that Dr. Buczko led that produced a prospective payment type of system. That could be the short-term strategy.
DR. MOR: You should know if you are interested in R squares, the R squared from the RUG system is substantially higher.
DR. IEZZONI: For rehab?
DR. MOR: For all populations including large populations of rehab patients.
DR. IEZZONI: I think it would be good to see a side by side of that actually, and judge it against the data-gathering burden. That would be a new burden imposed on the rehab hospitals.
Simon?
DR. COHN: I was just going to make sort of an editorial comment here. First of all, I would love to see more information about this stuff as well as the RUGS, but there was a comment made about overlapping things, and I think everything we have heard so far since we started this morning has been very overlapping with everything else. So, I think that is something we should just keep in mind.
DR. IEZZONI: And thinking towards what Nancy Miller is doing for the long term is where we really have to be focusing.
Bob May?
DR. MAY: I do want to point out that when we talk about data burdens there are both sides of this. It is one thing when you are talking about the burden on individual providers, but any system that we put into place, and this is really another reason for looking at integrating these, it is going to be a tremendous burden to operationalize these when you talk about the volume of data coming in on these systems, and to have to build three various separate systems and begin to pull all these together, that is really very significant. There is burden on both ends.
DR. IEZZONI: I must say that after the morning's session though, Bob, I have serious doubts about the time that you have available to build any system, given what Congress has mandated for starting prospective payment, and at least you have one system here that has had a lot of the up-front work done already although for the RUGS you do for nursing home as well. So, I mean it is all kind of very confusing.
DR. MAY: It is not the up front. It is actually building the computers that are actually going to pull all this stuff together through the states and everything.
DR. IEZZONI: All right, yes, well, this is going to be big.
A comment from the audience, and then we should probably move on.
DR. PHILLIPS: Phillips, from UDS.
DR. IEZZONI: UDS, by the way --
DR. PHILLIPS: Uniform data system for medical rehabilitation.
DR. IEZZONI: Okay.
DR. PHILIPPS: We have been hearing about the FIM, but let me tell you about an architecture that is already in place, and that is in the VA system, and I do understand the architecture that is going on, but there is a complete system which uses Web technology which is stationed in the Austin Automation Center in the VA collecting FIMs across the continuum of care. So, if you are interested in looking at something where you are looking at nursing homes and home care and CMRs or the high-tech rehab centers collecting the data in one pool using low-cost technology and than transferring the data back into their VIZINS(?) and other things I would suggest you look at the VA system that is in place.
We have given them our source code, and they have been in operation for about 18 months. I just visited one station in Texas, and it is very interesting to see how the quality assurance people and the financial people come together.
I think the rudiments of the arguments today are bringing other databases together in Web technologies, and what they have done is they have brought their durable medical goods databases together and their cost database together.
So, as they assign costs, and you are looking at function, and you are looking at what levels using the FRG system, they are able to really stratify cases across their whole continuum. So, I just give you an example that there is an operational unit and very low cost.
Thank you.
DR. IEZZONI: Interesting.
Okay, this has been unexpectedly stimulating for after lunch, and we thank our panelists for stimulating us. That was very interesting, and there is obviously a lot of work to do.
So, let us move on to the first of two panels on current collection and data standards issues. That is the tile of the next two final panels.
Dr. Morris appears to be the only person here. The other two are here as well? Okay.
John, do you want to begin?
DR. MORRIS: I am going to talk about the minimum data set for post-acute care. I am going to begin by giving some history of the minimum data set. I am going to talk about where we are in creating the minimum data set for post-acute care, and I am going to tell you of the steps that we will be taking in the coming months.
There is a set of terms. Seeing how, Lisa, you like to make sure of terms; you have been hearing them all day, but the minimum data set, the RAI, the RAI is an encompassing system that includes the minimum data set. The minimum data set is a series of standardized assessment items. Every item has a definition in back of it. It is a performance definition-driven system. We have examples only to clarify. We work from definitions. It is a clinical assessment system that includes instructions for integrating data from a variety of sources, from patient, from direct care staff, from family, from the physician, from the medical records.
The system includes specified time frames for every item. It includes instructions about what sources of data to use. So, in the RAI we have the minimum data set. That is the assessment component.
Because the RAI was designed for multiple purposes we have a number of tools that are built off of it. You have been hearing about the RUGS system, the RUGS-III system which is built off the minimum data set, and the MDS-PAC will include those items as does the current MDS 2.0, that is operational for the last 2 years.
The other components of that system include things like what are called RAPs, resident assessment protocols. Resident assessment protocols, and there are 18 of them in the current system specify certain types of problems, and there are certain items that specify them, and then we include guidelines and logic for how to think about those problems, the causal forces that might be leading to somebody falling might be leading to somebody having behavior problems, might indicate how we could prevent pressure ulcers, etc.
So, we have the minimum data set, the MDS. We have the RAPS. We have RUGS, and we have a series of quality indicators. The best known, of course, are the ones that have been developed in Wisconsin by David Zimmerman, but there is a series of others that other groups have created.
The system was created from the point of view of the resident. It is a resident-centered system. We brought together a lot of people in a process to begin to understand what it was you would want to know about that resident; what are the parameters that you would want to know and why? There are many disciplines that are involved with the resident in a nursing home. They are around the table, resident advocacy groups, around the table, HCFA, state, you name it, we brought around the table. We went through a process to create the original minimum data set and version 2, and we are replicating that process with the PAC.
What it involves is reaching out to people to find out what to include, making sure everybody agrees about what you would recommend for inclusion should be there and deciding when there is enough disagreement to exclude information, making sure that every item is well defined and understandable by clinical staff in facilities, then ensuring that there is a reliable assessment capability, given what you created.
We believe that as a foundation that every item in a system such as this should be replicable from one clinician to another clinician within a reasonable time frame.
In the minimum data set we include a few exceptions because there are certain clinical concepts, dehydration, delirium that have lower reliability. The reason they have lower reliability is because clinicians have more trouble identifying them, but most of what we have included in there has pretty decent reliability. Wherever possible it is performance based. We want to know what the individual did during the time period. We want to know who touched hands on them and supported them, but we try to define everything from the point of view of the patient.
As we have gone forward we are now into the world of the post-acute care continuum. When we designed the minimum data set, yes, there were skilled nursing facilities, and yes, there were people concerned about how the minimum data set would work in that environment, and questions were raised, but the environment has changed dramatically. The needs have changed dramatically, and so, HCFA is going through a process now of asking what should be in this new assessment instrument.
There was a handout at the front, and I just handed out in the Committee a document that explains a little bit about what we are doing. It includes a copy of the instrument and then includes a, for those in the Committee but not all back there, worksheet that we are asking people to complete about the assessment.
What we have done so far is the following. We have an expert panel that we have met with. It includes about 60 people from all industry segments. We have held several focus group meetings with industry segments, both nursing home, rehab facilities and specialty hospitals.
The current preliminary draft of the instrument is called draft six. What we have been doing is talking to people all around this country about what should be in this. We have a large consultant panel and staff from several locations who have been working on this with HCFA staff.
So, we have now a draft document, and if you just look at what, to clarify what we mean by a minimum data set, we got stuck with the term "minimal data set" by Congress in 1987, and as you know, the concept of MDS had been floating around in other governmental uses way before that, but this has become pretty well known as a minimal data set. It doesn't include everything, but it includes a lot.
One way to think of minimal might be 12 domains and one question for each domain. That is not what we mean. If you think that is what we mean, you are wrong.
We are trying to within each dimension capture information that clinicians think is relevant. Everything that we put together has to have a clinical utility to it. Even the items in the system such as RUGS has a clinical utility to it. We are building off of a very basic building block. So, feel free to flip through some of these pages and get a little bit of a flavor, and I will gladly comment on that instrument.
Where we are today. Well, the first thing we did is we took the minimum data set and had a couple of hundred folks who were interested look at version 2 and tell us what was in version 2 that you would or would not put forward into a post-acute instrument, and we had heard from a lot of in industry segments that there are things in here that you shouldn't have, you should exclude. There were all these quality of life sort of things, the social engagement sort of things, these activities. Why do we care about that for somebody who is going to be in here for 30 days or less?
Interesting, there was no item in the minimum data set for these roughly 200 to 250 responders in which there was a majority who would say to exclude it, and one of the realities about creating these types of instruments when you put constituencies together different people like different concepts, like different types of information, and when you sit around a table like this the problem gets that all of you want things in it, and we have a problem limiting what is included.
Where we are now. We have this draft. We have sent it out to about 700 people who have indicated an interest in commenting on it. We are asking them item by item to tell us what they would drop, what they would keep and what they would modify. We are asking them to comment on this as well, particularly if you are going to tell us to modify something. Tell us what you would do.
So, we sent 700 out. I wouldn't be surprised if we don't get 500 comments in the next couple of weeks back to us. We have a meeting scheduled with our technical expert panel in a couple of weeks, and they will comment back.
If it is like anything in the past, there is a multiplier of about five or six. So, those 700 get photocopied, and the first thing you know, the associations end up, and there will about 3000 of these things floating out there. I am not sure how many will come back into my office to process, but I hope it is not 3000.
We are going to take all of those comments, what the technical expert panel tells us, what we get from these comments by people from all the industry segments and create version 7 of the minimum data set post acute.
Now, we have already made a couple of important decisions. One is that, and we have asked the Technical Expert Panel and several other people have commented earlier to us, when should you do this assessment; how long into the stage that you first touch the patient, 7 days, 1 day, 1 hour, 14 days, 6 months? I mean when should you touch the patient? The median response was 48 hours. So, we are now going in, and you can see on the top of this a 48-hour evaluation. It includes a lot of questions about what the person was like 30 days ago. You are 48 hours into the stay, and there is a lot of reference for 30 days ago, because we want those anchors.
We are, also, asking questions, and we are continuing to ask questions in the 700 that we sent out as to how frequently should you do this; when do you do the next one? You do it 5 days, 7 days, 30 days?
Betty Cornelius in HCFA, when she did the case mix demonstration which is the demonstration that evaluated, constructed and looked at the RUGS-3 system put into place an assessment that goes 5 days, 14 days, 30 days, 60 days, 90 days, a series of these things. If you really want to get paranoid, and you are paying for this you could somebody who would say, "We will do it every day, you know, every day; we don't want to take any chances. We will miss something."
We are asking from a clinical perspective when we talk to people. I mean that is what we have always started this thing from. What do you think? How frequently should we do it? When do we touch the patient? How are they changing? When would it be important from a care planning process to look at it. We have always said that you can override. You can do an assessment out of cycle if the patient is changing or isn't achieving the types of objectives you would like them to achieve or you thought they would achieve, but how frequently should you do it?
We are going to come out with version 7 in the next month, and what are we going to do in version 7? We are going to do four data gatherings with version 7, and by the way, we are doing one big data gathering with version 2 of the MDS.
Let me step back to the first one, version 2 before I get into -- all right, so MDS 2.0, the existing MDS that is out in nursing homes. For nursing homes, for the post-acute care in nursing homes, we have a huge amount of computerized data that we can look at and understand who this patient population is. What we don't have is comparable data from either rehab facilities or specialty care hospitals.
So, we are going to go out into a reasonably-sized sample of both of those and gather some longitudinal data to understand exactly as best we can who is there. So, there are lots of statements that people are making about who is there, who they are serving.
You cannot really design an assessment system unless you know who you are designing it for.
DR. IEZZONI: We need to wrap up.
DR. MORRIS: Right. The final thing that we are now doing is four data-gathering efforts. We are going to do two small-scale efforts, finding out how the industry thinks about this, how long it takes, and we are going to do two large-scale efforts with the next version of this including reliability trials.
The goal to have this done will be by the end of the year, beginning of next year.
DR. IEZZONI: Okay, thank you.
DR. STARFIELD: Who is we?
DR. MORRIS: This is a contract that the Hebrew Rehabilitation Center has with HCFA. We have subcontractors with Myers Institute, with University of Michigan, and we have consultants from about five or six or seven other places that are actively involved in the design effort.
DR. COHN: A question of clarification. I know what MDS 2.0 is. You mentioned 7.0.
DR. MOORE: If I did, I slurred 2.0. There is MDS, MDS 2.0 and MDS-PAC. That is all there is..
DR. IEZZONI: I think what he was referring to was draft 7.
DR. MOORE: Draft 7. When we made MDS 1 we went up to draft 56. Catherine Hawes was our leader of the contract then at RCI. We got up to draft 56. We are only up to draft 6 now. We think by the time we are done we will be up to about draft 9.
DR. IEZZONI: Okay, great. Thank you for bringing a tangible thing for us to look at. It has been very helpful.
Okay, Dr. Glavin?
DR. GLAVIN: The first thing I will say is I am not a researcher. So, I don't have a lot of data. I would like to bring the information to you from a provider perspective, and I was involved in Northeast Ohio PAICE development and, also, I just finished 18 organizations long-term care from community service to skilled nursing, chronic care hospital, developed regional alliances. So, we have kind of extremes in various kinds of settings.
I prepared some of the transparencies. So, maybe we can start with that.
This is multiple information from various institutions, Benjamin Rose Institutes; there are 180 beds, skilled nursing facility with a 40-bed subacute unit called post-acute unit, adult day care community services and case management on the outpatient side, Grace Hospital which is a 52 licensed long-term care hospital and St. Michael's Hospital is a combination of a hospital with 50 beds of subacute unit and they have 100 beds that are hospital beds but with occupancy about 50 percent. So, it is really and equal size of subacute and hospital setting. Presbyterian Senior Care in Northeast Pennsylvania, and they have various nursing homes, residential care, CCRC, continued care, retirement community, adult day care and, also, they manage multiple subacute units for the hospitals.
Ohio, Pennsylvania Health Care Organization, they represent probably 60 providers of long-term care, and again, Ohio State Medicaid MDS initiative as well.
What I would like to talk about is the experience I share with you in terms of from the state policy perspective as well as from the provider and the operation perspective in terms of data gathering for the post-acute components, and what additional information do we collect currently in addition to MDS and how we as providers see the transition post-acute continue and then I will probably talk about some of our suggestions in terms of future data collection.
Ohio initiated the use of MDS originally in 1990, and actually implemented payment system in 1992. MDS plus is now being used. However, in addition to MDS looking at payment system there are other programs, also, happening in Ohio. PASPOR(?) which is Medicare home care and community waiver and pre-admission review of the Medicaid skilled nursing patient need to go to a nursing home has to be previewed before getting certified.
Next slide?
After implementation there are a few things we have found in the past. This is eye test. From 1990 to 1995, let me just kind of summarize some of the things. What we have found in Ohio is occupancy in the long-term care skilled nursing declined. The occupancy rate, you know, you can see is from 1991, 1990, and I cannot read it from here.
PARTICIPANT: Eight-eight.
DR. GLAVIN: Eighty-eight over the past few years, and you look at the Medicaid and Medicare certification, and you can see the decrease on the Medicaid and increase on the Medicare. So, one of the things is a lot of short-term acute patients now moving into existing skilled nursing facility.
Next slide, please?
This is an overview in terms of the changes in terms of the program on the policy level. As you can see there is skilled nursing and there is a facility, and we look at the utilization as well as the PASPOR which is the community waiver services, and we can see the increasing usage of PASPOR services especially in the 85, age 85 and over. What we have found is that the increase of the nursing home population and in a younger population which is 65 to 74, 75 to 84 pretty much contributes to the short-term acute right after the hospitalization.
The experience we have found in terms of MDS in the past 6 years is definitely we have found there is a better provider care planning and quality care after implementation of MDS, increased resident acuity and you know in terms of combination of the community and home waiver program and as well as more and more leaving now is taking more care of patients. Increase the facility's ability to budget planning and projection, it used to be they had to budget on a day-to-day basis. It depends on how many patients they have. Now, they pretty much can look at the MDS score and have some predictions.
The concern challenges this point, and the first one is assure the communication with the providers to minimize the system abuse, and we found acuity of the accuracy of the MDS score is still very low, and trend organizations, and their accuracy rate is pretty much high, up to 80, 90 percent. Some organizations will have 50 percent accuracy in terms of the staff rating on the MDS score. So, there is a big challenge in terms of Ohio. Is the trend then educating and communicating with the providers? I think Bob Coney from HCFA just shared with me the accuracy rate is about 83 percent at this point.
The other challenge is the Medicare-Medicaid MDS assessment and classification consistency. I think now Medicare has MDS and have a raw grouping and state has its own, and how to coordinate both of the grouping systems and make sure there is a consistency there, and, also, Medicare-Medicaid in terms of payment coordinations, and I think that is what we found from Ohio.
From the provider perspective and I want to use the majority of information from Benjamin Rose Institute and I mentioned before has a variety of services from skilled nursing, home care community service, education, research and, also, is the site of Northeast Ohio PAICE site and a member of long-term care continuum alliance organizations.
Over the past years the way we use MDS is to go back to some of the speakers talked about the multiple applications, and the payment system has a requirement from state because of the payment justification, and we use that for planning, and the first time I ever downloaded MDS on the PC which you can do, and it is for our planning of skilled nursing facility because I wanted to know what kind of resident mix and for future facilities. After downloading that information we realized there are so many things we can use this set of information, and we started looking into the opportunities.
Currently we are downloading MDS on a weekly basis and looking at that information. We have a weekly quality measure review committee to look at quality issues, and cases such as increase and decrease of falls and skin breakdowns, incontinence and we find it is helpful to be a reference to identify some quality issues.
Resource management I share with you some of the information we use. Pretty much you can look at MDS scores to decide about how many staff or what type of staff we need.
This is a typical weekly report. We look at overall census. We look at depression. What we have found out is depression is a big issue in terms of resource allocation in the skilled nursing facility versus physical needs. Rehabilitation, nursing rehabilitation, this is pretty much about a summary of the weekly report.
Next one?
This is a real eye test. Just to give you some overview and in the front is all the RUG grouping, and all the second column is a mixed score. Then across the whole line is we now only look at the whole facility in terms of case mix, and we divide them by the unit. You know, one of the things is that even in a skilled nursing facility they have different units, different floors and patients have different acuity scores, and so, we compare. We look at the total. We look at compared to each of the units and look at the mix.
Staffing, on each of the four we look at the staffing and pretty much use as you can see the last line is the case mix scores and the front is how many nurses we need, LPNs, RNs and nurse's aids and total staff configuration in terms of recommitting for that floor.
Key conclusions, development of common definition of language for multidisciplinary staff is a common tool. Definitely MDS is doing that. MDS, also, is the only objective in standardizing the payment resource allocation tool based on patient acuity. MDS-RAP and Dr. Morris just explained about different things. The plan of care development provides staff with standardized patient assessment tier management protocol, and if the score is accurate, there is 90 to 95 percent accuracy we feel in predicting resource allocation.
References are for staff or supply cost management, and we look at incontinence and we know how much supply we should put aside for that issue.
Ongoing quality management, I did mention about that. We review on a weekly basis to make sure we didn't lose any of the pictures. Effective communication for transition management. Not only patient transfer from setting to setting, also, from floor to floor, and I think MDS has served as a good tool in terms of communication during the transition management.
The issue of recommendation, there is some in the area. We think we probably need some more work to look at some accuracy and projection power. Decubitus care, the number of decubiti and various stages of severity which is a significant impact on resources; however, it is not dictated by the MDS.
Respiratory patients will need many suctions or one suction an hour or one suction a day and is the same score on the MDS score.
Patient with multiple services rated under most severe categories only rated under one category which is to capture all the resources needed. Again, short-term policy, acute medical and rehab patients and I think we already heard there is a demonstration going on with MDS-PAC.
DR. IEZZONI: We should try to wrap up.
DR. GLAVIN: The clinical applications is centralized data collections and how to collect the data effectively. They use a coordinator or use a decentralized caregiver, and the nurse is on the floor, and the short-term conditions such as the patient has 3 days pneumonia, recovered, never showed up in the differences and limited sensitivity in terms of the clinical intervention, and, Glenn maybe we can go to the next one?
Okay, Glenn, can you do the diagram from the circle? So, we skip some of this? The current post-acute care actually if you look at post-acute care pretty much we talk about post-hospital DRG payment system kind of program, and if you look at this all this is subacute care in terms of hospital based rehab long-term care and rehab services. If patient is discharged from a hospital they can pretty much go to -- every one of them depends on which program you have and with a potential of a financial incentive.
Next one?
The current post-acute care competes with each other, is provider driven, is highly regimented and lacks very clear clinical criteria, and there is no integrated infrastructure and no assessment admission and documentation discharge plan and, also, there is a lack of streamlined information system and management in terms of provider to provider or provider to caregivers.
The result is a waste of time and resources, increase of risk of error, frustration, dissatisfaction of patients, providers and caregivers and, also, poor tracking of the outcome and lack of provider accountability. I know my time is up. So, what I would like to do is can we do the last one?
As a provider we did put together a future post-acute care and integrated care system that I think we are going to hand to you with the handout. We agree with some of the Committee members' suggestion. We definitely have good definitions of what post-acute care is and develop a framework of post-acute or integrated care and should be established before the data, you know, to drive the data collection.
The other thing is that there are many, many demonstration projects, and our suggestion is the demonstration project should be linked together by this conceptual framework, and also, we like to encourage HCFA to consider long-term care solution not the short-term answers, and so we do need to look at the whole integration of services versus just looking at current box type of approach in terms of post-acute care.
DR. IEZZONI: Thank you.
We are a little bit behind schedule. I have just been informed that the cafeteria closes at three. I feel that there are some people who made a very special effort to get here for the 3 o'clock panel, and so, I really don't want to short change them.
So, what I would like to do is continue with Nell Wood and those of you who need to get coffee should go now and do so, and you will just be missing for a few minutes, but Ms. Wood was nice enough to give us a copy of her handout, and then we will take a stretch break briefly after this panel finishes before the next panel starts.
MS. WOOD: I was told I was going to have 15 minutes. So, I will just step it up and do it in 10, so that everybody can have a stretch. I am actually pretty good at tearing through material. I am actually often told I am overly fast. So, let me know if that ends up being a problem.
I was invited to share with you a little bit about what the quality indicator project has done in the area of developing some comparative performance indicators for long-term care. I would like to give you about a thumbnail version sketch of the quality indicator project and where we got our beginnings and that sort of thing so you understand where we are coming from in developing this indicator set.
Using the QI project as a template essentially refers to the fact that we began doing this about 12 years ago for acute care facilities, and we are operated by the Maryland Hospital Association, and we are part of the Maryland Hospital Association, and about 12 years ago which may have some curious timing we were approached by a handful of our members regarding opportunities to compare clinical performance across organizations, and we are governed by trustees from our member hospitals which makes us a little bit different than the average hospital association because most of them are governed by CEOs from their member hospitals.
What we were hearing from our trustees was essentially this. We have got plenty in the way of financial data and plenty in the way of utilization of data available at the trustee level, but you know when someone calls me and says, "Comment on the quality of care provided by your facility," I cannot and as luck would have it, I am, also, legally responsible for the quality of care provided by my facility.
So, without an opportunity to see this kind of information floating up to the board level at a regular basis I feel like I am not doing my job. This is what we were hearing from our trustees. So, that was the assignment, and we managed to talk about 7 acute care facilities in Maryland into giving up some of their most sensitive data and for purposes of comparing and trending on a regular basis, and within a short period of time we actually had all of the hospitals in Maryland participating in the project.
We received a grant from the Robert Wood Johnson Foundation about a year and one-half later to conduct a research and evaluation project for the QI project. Three years we were funded, and in that period of time we grew to about 150 acute care facilities.
All that brings us to today where right now we have about 1100 acute care facilities participating in the quality indicator project.
We are virtually governed by this group of facilities which has its good points and its bad points. One of the good points, I think, in being governed by your clients or your participants is the fact that when they are actually the users they are pretty much right on the mark when it comes to what they need next and what is going to be useful to them, and the development of the quality indicator project I think reflects the extension of acute care facilities along the continuum of care.
We began doing this for acute care 12 years ago. It has proved itself to be useful and on target, and I think again that is thanks to the participants who tell us what it is we need to be doing.
We expanded into psychiatric care about, well, we began our pilot test about 5 years ago and actually launched the indicator set 2 years ago, and following the foray into psychiatric care we launched an effort into long-term care, and again, this is at the direction of the participants who say to us that we are moving into this area.
If this methodology you have provided to us and continue to update for acute care setting proves useful, we would like something that sort of matches the approach and the philosophy that we can take into the long-term care setting.
Regarding the philosophy of the project it is essentially parsimony, and this is something that we embraced during the indicator development process as well, and when we brought together our expert panel to brainstorm the long-term care indicators I think we came away from our first meeting with some 100-odd measures, and it was quite a group we had around the table.
We had physicians and quality assurance professionals and RNs, and we had coders and everybody you could think of who would either be involved in getting the data or analyzing the data or looking at feedback and implementing it, and the real task of the moderators of the expert panel is to separate the nice-to-know from the need-to-know, and I bet a lot of people in this room have been set to that task previously.
So, as you could see, 100 indicators, and we actually ended up implementing seven indicator domains, each of which has a couple of different breakdown rates. So, that philosophy really ends up being very important because our whole feeling is that there is no reason to collect data if you are not going to do something with it, if it doesn't present to you an opportunity to better understand the quality of care that you are providing and identify opportunities for improvement, and when it comes to sticking to our knitting, that is what we believe is our knitting is establishing indicators that can help people actually do something.
The modus operandi is actually sort of curious, but I think it is one of the things that helps us keep our finger on the pulse of what is going on in the industry.
We, at the Maryland Hospital Association provide the QI project directly to facilities in Maryland to the acute care facilities and any other components that they have, whether they have free-standing psychiatric units or hospital-based units and, also, free-standing long-term care or skilled nursing facilities, as well as hospital-based units.
When we provide it to the organizations outside of Maryland what we actually do is establish partnerships with multi-hospital organizations around the country and then they work with their hospitals. They may be state hospital associations. They may be multi-hospital corporate entities. So, that is how we work, and they essentially are responsible for keeping in touch with their facilities that they provide the project to so that we can kind of figure out which directions we need to be pursuing.
Next slide?
When we first met with the expert panel this is essentially what we said to them. These are the things that you need to be considering when it comes to developing an indicator and even with this task list that we put in front of them we still did come away from that first meeting with about 100-some-odd measures that they wanted to try before we actually managed to ratchet it down into a smaller group.
I don't want to give a lecture on what makes for a good indicator. So, I will run through the bullets very quickly because I know we have experts far more expert in the room than I. Measurable and reliable, this simply speaks to the idea of developing an indicator and why it ends up being important.
One of the things that we really tried to do was stick with the MDS as much as possible because we realized that there were lots of organizations that were using the MDS, and that would only grow over time.
I have to say that I was a little disconcerted when I heard John saying that MDS 7.0. I thought oh, my goodness, the last one we looked at was the 2.0. So, we are really behind if we haven't been using the 7.0. So, I am glad that that was just a draft. I will take that back with me and make sure that we are on target.
Universally applicable, this becomes a real challenge when you are looking at facilities that sort of call themselves different things, skilled nursing units, subacute units, transitional care units, all of these different kinds of facilities.
They want an opportunity to compare across the walls, real or imagined that exist. Is it valid for these different kinds of units to be talking to each other and comparing rates? That is something that you can only understand and come to appreciate over time. It requires a lot of exploration.
Nonetheless in developing the indicators we did where possible endeavor to describe indicators that could be used by a variety of different long-term care facilities. Clinically relevant I think is fairly obvious until you actually begin developing indicators because again it is that separating the need-to-know from the nice-to-know and what really is going to be useful.
Risk adjustment provides a real opportunity for discussion, I think. There are lots of organizations that won't even begin to use indicators unless they are risk adjusted, and presenting them with opportunities to risk adjust then becomes a very loud debate over the most appropriate methodologies for risk adjustment, and our feeling is that whether or not an indicator is risk adjusted should never be the sole reason for using one or not.
There are lots of opportunities to compare and discuss and learn with indicators that are not risk adjusted, and one of the biggest issues associated with risk adjustment is whether or not the facilities involved in the data collection in comparison endeavor are technologically there regarding data collection and gathering of patient level data and things of that nature and whether or not they are technologically there.
If they are not technologically there it should not be an excuse not to participate in some sort of comparative analysis opportunity.
Next?
Let us skip this one and move to the next one. I am going to try to bring it in in under 7 minutes here. One of the things that was important to us in developing this indicator set when one looks in the gender for pilot testing, the generate interest; don't force it issue is very important.
We didn't come up with this idea for the pilot test; actually developing this out of long-term care indicators as I indicated earlier was a result of the interest of our participants. So, we knew that we had a group that was ready and willing to use it.
Obtaining pilot site commitment we thought was very, very important because there are lots of people who express interest and enthusiasm at the outset and then once they get a look at what it is they have to do say, "Well, you know, we would if we had time," and nothing really gets done that way.
So, one of the things that we did at the outset, we said that we are going to give you all of the information up front. We are going to give you 2 weeks to review it, and if you want to do it we want a letter from your CEO saying that you are committed to the pilot test process, and that actually worked out very well because essentially what it did was generate discussion regarding the pilot test prior to involvement in the pilot test, and very few CEOs will sign on the dotted line without saying, "What is this that I am signing, and what does it mean, and how is it going to benefit our organization?" in particular because pilot testing takes time which means that the volunteer facility is actually making a commitment, a financial investment of sorts in the endeavor.
So, it was important to us to obtain that. Obtain feedback, solicit feedback, that is that second item in the third bullet there, that is a very important process. To test and implement is something we found to be very disastrous. What makes more sense is to test and to obtain feedback we provide sort of feedback guidelines, different bullet items that we expect our pilot sites to comment on whether or not the data are collectible and who in the facilities was actually collecting the data; where did definitional issues become a problem or become a point of discussion; where do we need to enhance our definitions and finally, just looking, eyeballing this indicator set how do you see it being involved in your quality improvement efforts and what kind of comments do you have there?
Take this feedback from the pilot sites because they are the ones who are using it, the idea again, design the tool for the users, take this feedback, make some changes to the indicator set and then retest again another quarter's worth of pilot testing, and we found that that has been very, very useful.
The idea that scrapping it is a possible outcome, it is just like anything else. When you spend a lot of time developing it, you almost want to shoehorn it into a situation. I cannot say that we have never made that mistake. There have been a couple of indicators along the way that we just are certain we could make work and make useful, but when it came right down to it we couldn't and we had to scrap them. So, that is a hard lesson to learn, but it is definitely a worthwhile lesson to learn. It is the cutting your losses section, okay?
Just to give you an idea of the pilot test process, we are very lucky with the structure that we have with the quality indicator project because of the sort of network, the web that we have established around the country.
When we announced the fact that we were developing this indicator set and that we were ready to launch a pilot test we actually had a tremendous amount of response from our group of participants and other organizations, and we had 150 organizations apply to be in the pilot test, and that sounds very lofty. We never expected to say that we had some apply and only accepted a few because we sort of assumed that we would be talking people into the pilot test. It didn't actually work that way. We had so much interest in the pilot test that we were actually concerned that we may have sort of an unmanageable pilot test, and of course, we didn't want that. We wanted a very sort of smooth and effective pilot test process.
So, what we did was we looked at the 150 that had expressed interest in participating and we said, "All right, let us get a representative sample here. We want a certain percentage of them to be hospital based, a certain percentage to be free-standing. We would like geographic representation."
So, that ended up working out very well for us. We had 18 states represented in the pilot test. We conducted two rounds of data collection, and we officially implemented the set July 1, 1997.
Today we have 263 registered participants, and I took that count on Friday which to us was sort of mind-boggling growth in less than a year to have gone from zero participants in July 1997 to 263 today. It was something else. I think there is a little program called ORIX(?) that may have had an impact on this. I don't know if any of you are familiar with that or have any sort of involvement, but I think there has been an impact there.
This is actually worthy of pausing for a second and making a comment. We are so concerned about organizations getting involved in the quality indicator project for the wrong reasons because we believe this whole concept of our philosophy; it is not the data; it is what you do with it, and we want to see people getting involved for the right reasons. It is not pedantic so much as it is associated with data reliability. If there is a commitment to participating and there is a commitment to collecting the indicators according to our definitions, and there is a commitment to participate in reliability surveys and conformance assessment surveys, then you have that much more of an assurance of your database integrity. So, we take that very seriously. It is not something that we take very lightly.
So, when we launched this indicator set July 1, we had actually just followed a 2-year pilot test long before anyone knew what an ORIX was. We were working on this long-term care indicator set, and there was quite a lot of interest generated as a result of ORIX. So, it has been an interesting process for us walking through this interest level and making sure that we truly have a committed group of participants in the project.
Quickly to take you through the long-term care indicators and let you know what we have done with them and what some of the breakdown areas are, each of these essentially represents a domain. Most of them have breakdown areas, and I bet John could comment very quickly on which ones have data elements that are obtainable from the MDS and which ones do not. Not all of the data elements are obtainable from the MDS.
In looking at the MDS and in looking at the areas that our expert panel recommended we collect data for there are just simply going to be clinical areas that are not collectible from MDS or other sorts of data collection mechanisms of that nature.
Unplanned weight change looks at weight change, and actually I will, also, comment as we move through this indicator set on which ones the Joint Commission liked for ORIX and which ones they did not like for ORIX, and before I even move into this I will just comment again on the idea of developing a tool specifically for its user, and a lot of comments that we and other organizations, performance measurement organizations got back from the Joint Commission regarding their measures were that they were not useful in the accreditation process and word had it on the street that there were lots of organizations that kind of felt that that was somewhat of an affront, but really it is not at all an affront because most of the organizations that have indicator sets that they use and provide have developed them for the health care organizations to use in their quality improvement and performance assessment endeavors. They did not develop measures that were geared to be used as accreditation tools.
So, when you look at this indicator set and when I comment on different breakdown areas that is an important thing to keep in mind that these indicators are developed for users to use to assess their quality, to identify opportunities for improvement.
Unplanned weight change we looked at a 5 percent change in body weight either a gain or a loss, specifically again focusing on the unplanned aspect. This was one of the indicator domains that the Joint Commission did not like because of the data collection methodology. We use a snapshot technique essentially. We take weights on one single day in the month. It is the same. It might be the first Monday of every month or the third Thursday of every month, but it is uniform for each facility.
Of course, this only pertains to the residents who have been in the facility for 30 days or greater. Pressure ulcer problems --
DR. IEZZONI: I am sorry, we should try to wrap up.
MS. WOOD: Okay, I will stop then.
DR. IEZZONI: No, I think maybe you could just give us the bullets on each of the bullets.
MS. WOOD: Okay, that is fine. Pressure ulcer prevalence, the breakdown areas here, the indicators are categorized by pressure ulcer stages one through four, and again this is a snapshot data collection process and it was not approved for use in connection with ORIX.
Resident falls, we look at falls. We look at a repeat fall rate, and we look at falls resulting in injury. Unscheduled transfer to inpatient acute care, we look at transfers within 72 hours of admission, and then we look at an overall transfer rate, and we look at transfers according to reason for transfer. So, transfers for infection, transfers for cardiovascular decompensation, we look at transfers for evaluation and treatment of fracture and transfers for all other reasons.
The last group is nosocomial infection incidence, and we look here at lower respiratory infections, at symptomatic UTIs and the symptomatic UTIs are further broken down according to catheter-associated UTIs and non-catheter-associated UTIs.
The last bullet on this one, this is another one that was not deemed appropriate for use in connection with accreditation. The concern there was that we were only looking at symptomatic UTIs as opposed to asymptomatic, and this, of course, is despite the fact that there is copious information in the literature regarding the benefits of focusing your surveillance on symptomatic UTIs, but again, it was the design of the tool.
DR. IEZZONI: Okay, thank you very, very much.
Why don't we take 5 minutes for questions. Certainly somebody has a question for one of our panelists.
DR. MOR: I have a question for Ye-Fan. The 18 facilities or 18 organizations in Cleveland that have come together as a long-term care integrated delivery system, whatever, is there any common information set for them across the multiple kinds of service settings in which they provide care?
DR. GLAVIN: Yes, actually there are a few of the indicators. MDS definitely we pull out a lot of --
DR. MOR: No, I didn't mean indicators. I meant a common sort of clinical information system across the multiple settings.
DR. GLAVIN: You mean in terms of assessment tool?
DR. MOR: Yes.
DR. GLAVIN: Okay. MDS probably is the most common assessment tool across all 18 organizations. The next one will be the FIM, you know because a lot of rehab programs are going on there, and ORIX, nobody uses ORIX. It probably is too new, and nobody ever has ideas. As a matter of fact, that is one of the concerns from the 18 organizations because it takes a lot of money in terms of the investment in the software and in terms of looking at OASIS. Of course, you have to take that into consideration. Nobody really knows much about OASIS at this time, and one of the reasons the long-term care alliance was put together, the reason is that you are surviving the managed care organizations. So, there is a lot of information. MDS will not, or FIM cannot answer to the major care organizations, and we have to collect like the length of stay, go to emergency rooms, physician use. So, that will be a consistent data through the whole system.
DR. IEZZONI: Simon, you had a question?
DR. COHN: Yes, actually I have perhaps a clarification for Ye-Fan. You had indicated in one of your overheads that you felt that the MDS allowed you to take 95 percent of your expenditures for --
DR. GLAVIN: For resources. We pretty much based it on the Ohio data, if the patient has case mix scores, and then we go back and look at a resource allocation and in terms of operation supplies and we think 95 percent of the case mix and compared to actually spending of the resources are pretty predictive.
DR. IEZZONI: I have never heard of anything that predicts 95 percent. I actually don't believe it.
DR. COHN: If it is we ought to stop, forget to have six.
DR. GLAVIN: I think if you ask the state, the State of Ohio has 6 years' data to support it, and I think they are, also, getting a lot of information. I think they are concerned. You see, I think, one of the reasons you have two things. The first thing is are the scores accurate. That is the first question, and what we look at is if the scores are accurate and there is a very strong prediction in terms of resource allocation.
DR. IEZZONI: I think I need to see the data to feel really comfortable with that, having seen a lot of efforts to predict resource consumption.
DR. MOR: Actually if it is a per diem then --
DR. IEZZONI: Okay, so the variation is a lot less that you are --
DR. MOR: Because you are not predicting a future event.
DR. GLAVIN: It is a 3-month data if, you know, you look at Ohio report.
DR. IEZZONI: There are theoretical maximums on R squareds and I think that takes us a little bit, you know. May I just ask a question of John? You have got your post-acute care draft 6. Is this, maybe I wasn't listening; I probably wasn't, carefully enough; is this supposed to relate to home health care as well?
DR. MORRIS: That is HCFA's decision.
DR. IEZZONI: But if you have been designing, I mean is this a potential --
DR. MORRIS: We have taken the basic MDS moved it into home health care around the world. We published our reliabilities in JAX(?) in August and showed that the basic MDS elements that we have put into version 2.0 are equally reliable when moved into home care given the nature of the data acquisition possibilities in a home care setting, so that there is no technical reason why any of the items that we have designed in 2.0 wouldn't be transportable to home care environment, at least the ones that would be appropriate.
What we are designing here I think eventually does have some applicability as HCFA goes forward with OASIS to think about those elements.
DR. IEZZONI: That is why I am a little confused because we heard earlier a very compelling justification from HCFA for why they wanted to stick to a single approach for the rehab hospitals, for example, but I don't hear a similar compelling argument for the home health because I mean if what your system is could be a substitute or something similar to the purport of OASIS or the goal of OASIS, I am just curious as to kind of the rationale for going with OASIS rather than this when they went with this rather than FIM FRGs for the rehab side. I guess I just --
DR. MOORE: There is some rationale outside of HCFA, I think based on availability and work that was done. I think eventually HCFA is going to have to bite the bullet and ask itself the question what is in fact, minimally necessary information that goes across the whole continuum of care. As Steve said this morning, the UNII contract provides a perfect opportunity, the U-N-I-I. I call it UNII. It is pronounced different ways. That provides a perfect opportunity for HCFA to bring people together to ask themselves the question what do they have in MDS; what do they have in UNII; what do they have in OASIS? That might be a vehicle. HCFA has to ask the question. Clearly what they are going forward with is a series of instruments which will be demonstrated which will be potentially integratable in that never-never land of the perfect system, but what we are going forward with the MDS post-acute is starting with the MDS, asking ourselves the question, where do we have to go beyond it; what can we cut out of what we had before, and if you were to do a cross-walk between the MDS-PAC version 6 that you have there and MDS 2.0, you will see we have dropped material. You will see we have added material, and we can explain the logic of why we have done that because we are working from a patient point of view. The other, I think really big issue here is that the systems are designed for different purposes. The MDS is designed as a patient assessment system. The OASIS is designed as a patient outcome system, monitoring system. They both have utility and they were designed from different perspectives, and you know, there is, also, a different measurement philosophy.
A lot of the functional questions inside of the OASIS system are capacity built questions; what could somebody do? Most of the MDS in 2.0 and still most of it in the PAC is a performance-based thing. So, they are capturing slightly different things in some areas. In some areas they capture things very much the same.
If you look at pressure ulcers the OASIS has lots of great data on pressure ulcers. If you look at mood items in the MDS and cross-walk it as I have with the OASIS mood items they basically measure the same things. Some things, therefore measure different things. Some things measure the same things. I think there is a very interesting intellectual exercise that HCFA at some point will have to bite the bullet and ask how that cross-walk works.
DR. IEZZONI: Interesting. All of this is very complicated and hard, and I think we need a break, and thank you for our last panel. I really apologize for the rush.
Why don't we reconvene in 10 minutes at three-fifteen with our final panel?
(Brief recess.)
DR. IEZZONI: All right, now, why don't we get started? This is our final session. It is the panel on current collection and data standards, and we are going to hear from a group that we have repeatedly heard about during the day and that is OASIS, and Dr. Peter Shaughnessy is going to be starting us out.
DR. SHAUGHNESSY: My only fear is that as I begin there will be less people who want to come back, but I will proceed, nonetheless.
The last panel ended with an interesting discussion on the issue of why is HCFA going with OASIS in this particular area, home care versus therefore not going with MDS per se versus going with MDS more in the rehab area which I thought was a very worthwhile issue to raise.
I would like to follow up on it a bit. John Morris had some very good comments on it. In essence OASIS is really part of outcome based quality improvement, and I will talk about outcome-based quality improvement for home health care, and it has been under development for some time, and it really fulfills a very strong need for HCFA at the present time in terms of reacting to the prospective payment system that will be implemented at least in theory in late 1999.
DR. STARFIELD: Would you please say what you mean by prospective payment? I really don't understand. Are you talking about capitation or are you talking about some episode?
DR. SHAUGHNESSY: In home care the plan, first of all it is not clearly specified in the legislation, but the way it is envisioned for the most part in home health care is that a home health agency will be reimbursed on the basis of an episode of care from start of care to discharge, and the reimbursement rates will vary according to patient severity and case mix and so on.
So, under that system obviously as is the case with various types of prospective payment what we have is an incentive for home health agencies to cut back on services since they will be receiving a fixed reimbursement amount for that episode of care.
Now, whether they respond strongly to that incentive or whether their more altruistic motives to care for patients will override that remains to be seen, but the fundamental issue is that HCFA has a responsibility to determine what happens to patients under prospective payment and therefore to monitor outcomes over time.
So, where HCFA is coming from at the present time is really No. 1, we need to assess outcomes under these payment systems that are now evolving for home care and No. 2, we should use the system that has been developed to monitor outcomes for home health care, and that is the system that we will talk about today of which OASIS is a part, and I would like to say is a small part.
We approached this from the point of view of really how to go about measuring outcomes, how to monitor them in such a way that outcomes could be enhanced over time. OASIS happens to be the data set that was developed for that purpose. We didn't get into this with a charter from HCFA, if you will, or a charge from HCFA to develop a data set. We got into it from the point of view of coming up with a good outcome measure system that could be used for practical purposes of various types.
Now, therefore, OASIS is sort of a by-product, if you will of that set of goals, and, also, I would like to say that over the course of time I certainly don't see any reason why we cannot have a greater convergence of some of these data sets in terms of more common data items.
In the short run I think we have got to go probably with what we have because it satisfies a very important need at hand. With that little bit as background, I would like to say that you really cannot consider OASIS, therefore, without understanding OBQI, outcome based quality improvement, and I will talk about OBQI for a while here.
It is a rather expansive and we hope dynamic approach to a continuous quality improvement that has been under development for over 15 years, mostly with funding from HCFA, to some extent with funding from the Robert Wood Johnson Foundation. All of this focuses on home health care.
There are some misconceptions about OASIS by those who don't understand if you will the fact that it fits into OBQI. In essence the purpose and utility of OBQI should be really comprehended in order to comprehend what OASIS is about. It is, also, critical to note then that OBQI is targeted not just at evaluating outcomes from a regulatory perspective or what have you, but it is targeted at enhancing outcomes from the point of view of changing care behaviors to enhance those outcomes, and we will talk about that in a moment.
With that, I guess I could try the first slide, if we are set. This particular slide should be, everybody should have a handout. This is the first slide in your handout. It is what we call the applications framework for OBQI. It was 2 or 3 years in its development, and a lot of research underpinned it in the ensuing work, and I will talk a little bit about that shortly, but this slide can be looked at from two perspectives. One is the perspective of home health agencies and how they can use OBQI to do what I just mentioned, evaluate and then enhance outcomes, and it can, also, be used b government and regulators and so on for similar purposes.
Kathy is going to summarize this framework if you will from the point of view of providers or agencies, and then I will come back and cover it from other perspectives.
MS. CRISLER: Essentially the place where the provider enters into this whole picture is actually the boxes that fall down the right hand side of the screen because what was anticipated at an early point in the development of the OBQI framework was really the idea that the data would be used, the outcome data would be used at a big system level, in other words at the Medicare system level to evaluate the effects of home health care on the patients. In other words, if I am paying for care from this home health agency what, indeed, is the home health agency doing for these patients; what are the outcomes of the care, and so it had been envisioned that at the large system level there would be reports that would be utilized both by HCFA in its survey and certification process as well as by the individual agencies in terms of their own performance improvement activities and similar to what Nell mentioned earlier from the standpoint of the hospital industry and for example, the Joint Commission's ORIX initiative and moving forward in terms of performance improvement the home health industry has, also, been greatly impacted by the desire and the need to, in order to become and remain accredited to be able to demonstrate that its performance improvement activities are keeping patient outcomes high.
So, in essence the agency embarks on its own outcome enhancement activities in response to outcome reports which it receives, and we will, there are some examples in your packet, but we have given actually agencies outcome reports. They have begun on outcome enhancement investigation activities by investigating the care that was provided in terms of achieving these outcomes, and have made decisions as to whether they needed to improve care or in places where care was exemplary to reinforce care and the idea clearly is then to make this a CQI process to really be able to see the impacts at the point that they receive a second outcome report in the subsequent year's outcome report which allows them, again, to continuously monitor and continuously improve their own performance within their agency.
We actually are seeing this within several demonstration projects, one demonstration project conducted nationally, funded by HCFA. That demonstration currently has 54 agencies. We are, also, conducting the quality assurance component of the current prospective payment demonstration which is using OBQI in a slightly modified form, and we are, also, conducting a similar OBQI project within the State of New York which has one of the very largest home care programs within the country.
DR. SHAUGHNESSY: Okay, this particular as we call it two-stage screen if you will is intended first on the left hand side, and you can probably see the small print a little more easily on your handout to result in after data collection result in outcome analyses to some extent by patient group. We have what we call quality indicator groups, and they include about 15 to 16 different categories or conditions of patients so we can produce condition-specific outcomes for those groups or we can have what we call a global outcome report as well, and the risk adjustment and the risk adjustment models, therefore we spent a lot of time on, and we used them. They vary by group. So, if for example one is talking about the outcome of stabilization and dyspnea or shortness of breath, the model of stabilization for dyspnea for cardiac patients might be different from the model, the risk model for patients overall in the global outcome report.
In any event an agency gets an outcome report of the nature that we will talk about in just a moment. Based on that outcome report, then they select target outcomes, and then they go through with those target outcomes, and they can go through a variety of steps resulting in a plan of action which says, "Okay, in order to enhance this outcome further or in order to reinforce the outcome because I am exemplary rather than perhaps inferior, here is the actual set of procedures I will go through in the plan of action. I will change these care behaviors. I will change them in this way." Specification of whom will change them, how they will change, over what period of time and so on is all documented.
The agency then proceeds with actually implementing that plan and monitoring it. This is all from the agency perspective as Kathy was indicating. In the same way though HCFA or other bodies can use those outcome reports for regulatory purposes in order to monitor, for example, whether or not an agency is markedly inferior compared to all others. That particular agency would obviously require more resources in terms of survey and certification. Agencies that are superior or perhaps just even slightly above average may require very little resources from the perspective of regulation. So, again, this type of approach is useful both to providers and to regulators and even to payers in various ways.
If I could, oh, I wanted to mention one other thing. The focus then is obviously on outcomes and fostering continuous improvement. It is, also, on fostering continuous change in this whole framework so that it can evolve over time particularly in those areas where agencies have useful plans of action to change outcomes that in turn can be disseminated to other agencies. So, with this system the goal is to have a growing body of knowledge produced by providers of care on how to enhance outcomes and how to reinforce outcomes, and that is happening now in some of our demonstration programs.
I will talk about the impacts of this in terms of the actual ability to change outcomes in a moment, but first, if you would turn to Figure 1 in your handout this is not on the slide, you will see an illustrative outcome report, and this report is hypothetical. It is for Nirvana Home Health Agency and as you might expect with a name like that, it is actually a pretty good outcome report, but the first set of outcomes, I won't go through all the details of it, the first set of outcomes is largely functional, ADL and IADL.
I won't get into just in the interests of time, I won't get into the technical definitions of how we define improvement and stabilization and why we have done what we have done and so on except to note that in terms of the outcomes that are here you will note that there aren't any aggregate outcomes that sort of cut across many, many domains of care and give you sort of an overall functional outcome indicator or an overall physiologic outcome indicator, and the reason for that we worked with many different types of measures, the reason for it in essence is that clinicians have to be able to look at these reports and say, "Okay, I am weak in this area, and I want to do something about it." If they are told, for example, that overall your physiologic outcomes don't look too good, the next question is well, which ones, where, why; what should I do about it, and they have to have this level of specificity, the same thing with an aggregate functional indicator.
Now, we use those kind of aggregate indicators for overall evaluation research and that sort of thing, and they can be used for such purposes but to really satisfy the need of individual home care providers we actually resorted to, if you will, after a great deal of consideration these specific kinds of outcome indicators.
This report, and I will just take an illustration, improvement in dressing upper body, one of the functional areas that we consider very important in the developmental aspects of our research that underpins OBQI, if you look at this agency's performance you will see a 69.3 percent in improvement rate for dressing upper body with a white bar. That is the current performance for this year, a 53.2 percent rate for last year, that is the prior year and the reference group is 57.2 percent. Eventually that would be a national group.
So, both of those differences relative to the current improvement rate are statistically significant, .05.
DR. IEZZONI: Could we ask a question of clarification?
DR. STARFIELD: What happens to people who basically didn't have any trouble with these particular things to begin with?
DR. SHAUGHNESSY: First of all if somebody is independent, say, in dressing upper body, the measure is not computed for them. They are a zero on that scale. So, they are excluded from the computation of the measure.
DR. STARFIELD: In fact, you could have someone who is the same on some, better on others, worse on others, and you still don't know anything about that person?
DR. SHAUGHNESSY: I don't understand that.
DR. STARFIELD: Each of these is computed independently?
DR. SHAUGHNESSY: They are computed independently, yes, but --
DR. STARFIELD: You cannot really understand anything about an individual.
MS. CRISLER: This is the agency's outcome report. The data collection is actually computed, is collected using the OASIS items integrated into an agency's assessment. So, just as you would assess anybody on whether they, for example, do or do not have a wound, then that individual assessment is done for that particular patient. So, the person who is independent in dressing upper body, I assume that there is no care plan developed at the agency level that related to dressing upper body, but this is a report that is provided --
DR. STARFIELD: It chops people up into parts of bodies. That is the only point I am getting at.
DR. SHAUGHNESSY: As opposed to having an overall measure that says that this person's overall outcome was good, bad or indifferent, yes, and we did some of those. The only problem is if an agency ends up with as I mentioned, if they end up with the fact that they are basically inferior in terms of their one overall outcome measure they don't know where to go in order to improve it.
You can construct those kinds of measures, and again we have done so, but for an agency these have turned out to be most useful.
Okay, I won't go through every single one of these. I would like to point out that this is the global report, and there are as I mentioned condition-specific reports. So, if you wanted to, and I guess, Dr. Starfield this relates a bit to your question. If you wanted to only look at cardiac patients, and you wanted to look at certain outcomes for cardiac patients, there is a smaller set that pertain to cardiac patients.
DR. STARFIELD: Could I respond? I really don't only want to look at cardiac patients because whenever we do that, we immediately decrease the relevance of what we do to the whole population. Okay, we pick the things that we happen to be interested in, maybe in that population that Congress is interested in, you know, elderly white males, that kind of thing.
DR. SHAUGHNESSY: Sure, and I am not saying that that is all one wants to look at. If one wanted to look at cardiac patients, patients with neurologic problems, patients with certain types of chronic problems you can go the whole gamut and you can do that sort of thing, and you can look at a pretty good spectrum of conditions in terms of assessing the performance of the agency, and you can pool all patients together as is the case here.
In terms of the outcome indicators or measures that are used here, we talked earlier about reliability and so on. There are 25 or 26 global outcome measures that we call end result outcome measures that refer to change over time in health status. Of those 25 or 26, more than half I believe, 14 are based on OASIS items where the reliability coefficient, the weighted kappa is greater than .75, and all of them here are greater than .60. That is reasonably good reliability for these.
Now, is OASIS perfect in terms of reliability? No. We have two more reliability field trials coming along, but it does the trick in terms of these types of reports on the utility.
I should mention here as well, I suppose that we have gone through this with several of the demonstration agencies and at this point the 54 agency demonstration in 26 states that Kathy was talking about, they are about to receive their outcome report for their second year that looks like this. We asked those agencies at the end of the first year to target on two outcomes. One had to be an end-result outcome of this nature, and the other had to be hospitalization, and there was a lot of complaining about targeting on hospitalization because in home care obviously home health agencies don't necessarily have a whole lot of control over hospitalizing, depending on physician behavior and other issues as well, but our belief was, well, if you can do it with hospitalization you can probably do it with anything, and most agencies went along with that fairly well.
So, they developed plans of action for hospitalization, for improving in this case lowering their hospitalization rates and they, also, developed plans of action for the one target outcome they chose.
We are in the process of analyzing those results, and they are preliminary, but overall what we are finding and these are risk-adjusted results is that the rate of decrease in hospitalization rates appears to be on average 7 to 10 percent in 1 year's time, and also, the rate of increase for these end result outcomes that they chose appears to be somewhere between 10 and 12 percent.
As far as we can tell, and we don't have a lot of data to really risk adjust well with nationally, there is no national trend going on in terms of decreased hospitalization rates, and, also, we took a look at other outcomes that they didn't target on using this approach that were uncorrelated with the outcome that they chose for their target outcome, their end result outcome, and those changed in the range of 1 to 2 percent.
Therefore, overall what we are seeing is significant impact in terms of improved outcomes using the approach. We are not done analyzing and as any good researcher and statistician I am not going to say that those numbers I gave you are the final results, but I am comfortable that they are going to be pretty close approximations of the final results, and again, our goal with OBQI is just that is to put something in place that agencies and providers of care can use to enhance their outcomes and that is going on, and it can be used obviously for other purposes.
Okay, we will skim through a few other things. I have been impressed today with what I would call the commonality --
If we could go on to the next slide?
-- with the commonality of what perhaps might be an applications framework. Several people that have been developing data sets talked about it. We just mentioned our applications framework, but, also, another thing has emerged, and that is I think most people either explicitly or implicitly use guiding principles, and perhaps we all ought to be more explicit on what they are as we develop data sets and I am not just talking to those folks that do them as contractors or grantees or what have you but, also, government, specify what the applications framework is for a given data set and, also, specify what the guiding principles are that we are using for developing that data set. I think it would clarify a lot of things. Much of it is implicit and I think very logical. Also, as we move forward with a post-acute care or perhaps some broader term but post-acute care multi-provider, multi-setting data set
I would, also, suggest that we talk about specifically what the applications of that data set would be, very clearly and very concisely and thus have an applications framework for it because right now I think it means a lot of different things to a lot of different people, and all of our objectives, if you will, our intentions are quite noble, but to lay them out explicitly I think is very important.
In addition to that applications framework for this cross-cutting data set I think we, also, need some sort of guiding principles that we say, "Okay, we are going to follow these in developing it," and I think many of those kinds of principles have been enunciated today, but I don't think, again, we have got them together in a tightly integrated, well-documented sort of system, and if I might suggest right in the middle of my presentation as I have gone off onto this tangent, I think that is a critical role that this Committee, this Subcommittee could be involved in and would genuinely facilitate the development of that kind of cross-cutting data set and in some sense I think that is where you are naturally evolving as I see it.
There is quite a research background and in the interests of time I don't think I will go through it all. As a matter of fact, I don't think I will go through any of it, but I would like to point out that on Table 1 of your handout there are approximately 10 or 11 studies of one kind or another that underpin the development of OBQI or are testing OBQI or are taking it to its next stage.
As I mentioned they have been funded largely by HCFA and largely by the Robert Wood Johnson Foundation. I would just like to point out that, let me see, 2, 7, 8 and 9 are demonstration programs. Project 1 was a national quality measure study that was really intended to first address the question of is it feasible to measure outcomes in home care in such a way that it can be useful for providers of care and for government and then secondly, if so, what should they look like. That is where the outcomes system that we developed came from and it used considerable input. Just as John Morris was talking about I think the post-acute care MDS and obviously what took place for MDS for nursing homes and the like, a large number of clinical panels were convened. We had advisory committees and steering committees of various kinds in the development of all that, but what I would like to say that made it somewhat different, I suppose is that we started with specifying outcomes and outcome measures. We didn't start with data items of any kind, and we took quite a bit of time convening people around the issue of what are the most important and critical outcomes that are germane to home care that could be used within this applications framework, and only after that was done did we get into specification of data items that served the purpose of either measuring those outcomes or risk adjusting the outcomes.
We spent a lot of time trying to specify risk factors that were important to each outcome that was specified.
Okay, I think that is probably enough on that, and I would like to get back now to the next slide.
If we have time, incidentally and folks are interested, later and let me know when I should quit; how much time do you want me to have?
DR. IEZZONI: I think you should begin wrapping up in about 5 minutes or so.
DR. SHAUGHNESSY: Okay. I would be willing to comment on the reason for the difference in capacity-based versus performance-based ADLs in home health care versus the institutional long-term care, an important difference, may converge together over time but pretty important.
You will see in your handout that there are several premises that we began with and several principles that I would call guiding principles that we used in Slides 3, 4 and 5, and I would ask you to go through your handout as we are discussing this.
A great deal of time was spent in trying to have a very firm conceptual and clinical foundation for OASIS, and as I mentioned, OASIS was a spin-off of this whole process of specifying outcomes in OBQI. All through this though we were as sensitive as we could possibly be to the needs of home health care providers. If this didn't work for providers, then it was not something that really was going to make it in terms of a parsimonious data set which would be integrated into an assessment.
Another big difference between OASIS and some of the other data sets we have been talking about is OASIS, we don't even consider OASIS to be an instrument. We consider it to be a data set that can be integrated into an assessment instrument and if, for example, an agency has 300 items in its assessment instrument or 400 the way it works is an agency actually inserts the OASIS items, replacing like items, and we have found that to be very effective.
A decision was made after considerable deliberation not to have a comprehensive assessment instrument for home health care mandated, and there are lots of reasons for it which I won't get into, but we could discuss. That decision was made early on and certainly involved in HCFA and as a matter of fact the final decision was made by HCFA on that, but getting back to some of the main points about the industry itself, Kathy, did you want to comment on some of the reactions of providers in the industry?
DR. CRISLER: I think as Pete has indicated while the data set was originally developed for outcome measurement, the fascinating thing to us is how well received it has, indeed, been by the industry. I think that what we have found is that the data collection burden, for example, virtually does not exist primarily because of the fact that these items are integrated into agency records, usually replacing very similar items. In other words, most home health agencies already have, for example, an ADL scale that relates to bathing. So, it is a matter of cutting and pasting and putting the OASIS item in and removing the current item as it relates to bathing, and in fact, we did this ourselves at our center in preparation for the demonstration project to see whether any of the items were really so unusual that it was difficult to actually do this cutting and pasting and found that it indeed was not unusual or not burdensome. In the questions that we have asked of participating agencies we have found that the time commitment in terms of data collection is essentially non-existent, that agencies, one of the gentlemen this morning talked about the Visiting Nurse Service of New York and the fact that they are using OASIS in one borough, not using it in other boroughs, at first, and we interviewed providers from both such boroughs and found that there was no difference in terms of the time that it took to do an assessment or to do the documentation between those that were using OASIS versus those that were not.
So, the data collection burden as it exists in an agency is really not that of the data collection itself. We do require that agencies data enter the information and transmit the data to us either electronically via bulletin board system or on diskette. That does involve approximately 3 minutes of data entry time per record.
Okay, so that is another component which we are actually testing out the receipt of data in that particular manner for the production of the outcome reports, but the National Association for Home Care has reviewed a variety of instruments as it has moved forward in preparation of its uniform data set, and it has endorsed the use of the OASIS for its patient level information and has obviously other additional agency level financial information and so on.
DR. SHAUGHNESSY: In the interests of wrapping up, what is going to happen in the future is Medicare is obviously going to move forward with using something like this, therefore, to assess outcomes and monitor outcomes because of the need to do so under a changing payment system, but refinements of various types are planned, for example, measures that are more relevant to personal care, measures that are more relevant to terminally ill patients, to HIV patients and to short stay or managed care patients.
Chris Murtaugh mentioned this morning that that was an issue, and as I understand it, I wasn't here early on, but the issue of length of stay and how to deal with it in the context of such measurement is a very critical and pivotal issue.
We see OASIS and the whole OBQI process as very adaptable and certainly can adapt core data items I think that will cut across the post-acute setting. There is no reason why that cannot be done, and as we decide upon such a data set to insert it into OASIS and replace OASIS items with it is certainly feasible.
Obviously you are going to have to mess around if you will with new risk models and so on as data items might change or what have you unless they are sort of isomorphic, and you can derive the old ones from the new ones, but nevertheless there is no reason that cannot be done and the flexibility built into the whole OBQI framework and OASIS certainly permits that.
The documentation, very systematic documentation for purposes of training and so on is now in the process of being completed. OASIS-C, we are currently on OASIS-B. We use letters instead of numbers just to be different, but OASIS-C will probably be somewhere 3 years off or something like that, and in conclusion the refinements there will pertain to such things as I just mentioned as well as additional reliability work that is going on, the fact that as foreign language versions of OASIS are implemented, then as well you have got changes that come from that, but overall probably it will be change largely because OBQI will change over time.
I think I should probably stop there and certainly we can have questions later.
DR. IEZZONI: Thank you. Our final speaker, Terry Moore, please?
MS. MOORE: Let me see, so, I have about 15, is that --
DR. IEZZONI: Fifteen minutes.
MS. MOORE: Okay, my colleagues at Abt Associates and I are working on three major case mix related projects, all funded by HCFA, two in the nursing home world and one in the home health care world.
I am going to focus my comments mainly on how we are using the OASIS data set in our home health case mix project so that we can have a discussion about how the purpose of this data set is much different in our purposes and how that might have some implications for data standards, etc.
We were awarded a contract by HCFA to develop a case mix model for the episode reimbursement system, this prospective payment system that will eventually be implemented. I think Hilly and Frieda King talked about 1999, winter 1999.
So, we have recruited and trained 90 agencies in eight states in the country to collect a version of the OASIS which we are calling the OASIS-plus from our nursing home colleagues who needed additional items from the minimum data set to collect for case mix, for resource purposes.
We added elements to the OASIS, and so, we have called it the OASIS-plus. I will describe that actually in more detail just to familiarize you with the kinds of changes that we felt we needed to make in order to use this data set for the purpose of developing a case mix model.
We were required to use the OASIS by the contract that was awarded by HCFA and partly to test the feasibility of whether this particular data set could be used for both purposes. I mean lots of folks today have talked about the multiple purposes that each of their data sets have been used for.
This is another example. It is being used for outcomes based quality improvement. It has been widely tested in the agencies for that purpose specifically. Could it, in fact, be used to measure resource or predict resource utilization in the Medicare home care community?
We collect not only the OASIS-plus and I will describe when we do that. We, also, are capturing information on every visit that is made to a study cohort patient which I will describe. So, we are capturing information on services received and that information will be combined with our information on patient characteristics from the OASIS-plus in our modeling.
Unfortunately, especially for our HCFA audience who would like to hear exactly which elements of the OASIS-plus are required for home care case mix we don't know yet. The project's time frame is basically agency started collecting data this past October.
They will collect for up to 18 months and we do not know at this point nor do we expect to know when our interim report is due which is in this June which of the particular variables that either exist currently in the OASIS data set or in the plus items that we have added which are predictive of resource use, but we have added the items by making some guesses from what we know in the other care settings.
So, let me see, let us go to the first?
Why did we decide that we needed, given that we were mandated to start with this data set, why supplement it, why did we feel we needed additional items? The OASIS is an 89-item tool. In home care that is actually pretty big from the agencies that we are dealing with. That is a daunting number of items to collect though the minimum data set in nursing homes is something like 300 or something like that.
At any rate we took a look at the tool. We convened a clinical panel made up of the industry, home health nurses, therapists, social workers, administrators to talk about what they believe predicts resource use in home care.
We developed some criteria for if we were going to add anything what did it need to be? It needed to be believed to be a measure of resource use clearly. We wanted to use only items that had been previously tested. This is a short turnaround project that needs a case mix model soon.
We didn't want to be recreating the wheel, and of course, it needed to be clinically meaningful. We took a look at the literature in this area. Cognitive factors were found by Mathematica's earlier work in home health case mix to predict therapy treatments and the use of medical social services.
We looked at the OASIS and decided we wanted to add some items on cognition. Nutritional status and dehydration predict poor outcomes at least in the rehab, in the nursing home literature, and we didn't have any reason to believe that isn't true for home care. The OASIS, also, needed to have a little supplementation in those areas.
Next slide?
The end result, what we are calling this OASIS-plus we supplemented several domains and by domains I mean big categories of health status measurement like activity of daily living, sort of demographic information, integument, those kinds of domains.
So, we supplemented by adding items in particular domains. We added an entire domain which is nutrition hydration which was not in the data set, and basically added 41 items total which again is daunting to the home health agencies that we are working with.
These items, it is important to note that they really do replace items that agencies are currently collecting. The biggest challenge is to convince them of that, that you could actually get rid of your other items on things like weight change, nutrition and use these items.
Let us go to the next?
We have required agencies to collect the data in a manner which parallels what will be required under the revised conditions of participation for Medicare-certified home health agencies. The study cohort is Medicare fee-for-service patients admitted for 6 months of an agency's data collection period. Primarily this is October 1, to the end of March.
This tool, subsets of this particular tool which is now I gave you one copy, and if people want copies, they can give me a card or something, different subsets of the 129 items are collected at different measurement points. This has actually been a pretty big deal, and I differ at least at this point in our data collection with Kathy's comment that the data collection burden is minimal only because probably our agencies are new enough at collecting this data, and we have added data to what Colorado has been collecting, but they have struggled with this.
So, they collect the data set on admission to home care. They collect it about 60 days thereafter, when the patient is discharged from home care or if the patient has been transferred to an inpatient facility, and this frequency I would like to comment on a little bit when I am done with these slides here, and actually I might be. Is that the last one? Okay, great.
I want to just talk a little bit about problems to date. I agree that this is an evolving tool. This is probably something which eventually can be consistent with other measures that are used in this post-acute care population.
There is a couple of big problems, some of which have to do with the actual items that we added to this data set, and the other problems are the time frames for data collection.
At this point we don't know how frequently characteristic information on home care needs to be collected in order to measure resource use. So, we have adopted a data collection strategy and a frequency of data collection to make life easier for agencies who are coming into our study because this is what HCFA will require, this frequency of data collection, and perhaps that might be something that we can talk about later, but for OBQI it is important to have the 60-day measure, the discharge measure.
Again, we don't know at this point whether this frequency of data collection for resource purposes is necessary or whether it is too long. So, that is an issue that is outstanding.
There are, also, some administrative requirements that Medicare-certified agencies are up against which make life a little bit difficult when they come to collect the 60-day measure which we could talk about a little bit, and that is just something that HCFA will probably want to consider whether or not having a required physician recertification at day 60 is something that can jibe with a data collection tool that might potentially get in the way of that time period, and there are lots of nitty-gritty details here.
I don't want to go into too much depth, but if people have questions, that is fine.
The other point I mentioned about the issues of whether this data set could eventually be used for case mix and whether it stays in home care or whether it becomes replaced by some other post-acute care tool that Nancy Miller and Bob Cain are talking about, the items that we chose to add, again, I said that they were believed to be items, believed to be measures of resource use.
We went to the minimum data set for home care for these items. So, the OASIS-plus is actually a combined tool that has both measures from the minimum data set for home care, as well as OASIS measures.
The biggest area that the OASIS was supplemented in with MDS measures is in the area of activity of daily living. So, we are asking agencies who are in this case mix project to not only collect capacity measures which were already in the OASIS data set but now performance measures. So, what did you really do in terms of bathing, etc., in the given time period? I think we are using a 48-hour time period on admission.
So, we don't at this point know how well agencies are going to do with the different measures and which set if either perform better in terms of reliability, which set if either perform better in terms of measuring case mix use or resource use.
So, there are lots of -- too bad this panel doesn't convene a year from now. Maybe we would have a little more answer, but that just gives you an overview of the kinds of things that we are doing with the case mix project.
DR. IEZZONI: Thank you.
Barbara?
DR. STARFIELD: Cut me off if I take too long because I could very well take very long with this. Do you know of any, well, let me start off by saying that I think it is reasonably well recognized that this country probably does worse in this kind of care, this non-conventional setting care than almost any other western industrialized nation. Do you know of any other country that takes this approach where a payer and not even a single payer at that is looking at quality in this kind of detail?
DR. SHAUGHNESSY: Are you talking about just home care or nursing home care or all types?
DR. STARFIELD: I am talking about them all. Take what you do. Take the OASIS. Take any one you want.
DR. SHAUGHNESSY: In terms of home care at the present time, no. There are --
DR. STARFIELD: Yes, but there are very well-developed home care systems.
DR. SHAUGHNESSY: Yes, there are data systems in Europe, in Belgium, for example, where information is collected on case mix and used for monitoring quality, but that wasn't its original intent.
DR. STARFIELD: What was its original intent?
DR. SHAUGHNESSY: Its original intent was to collect the information to monitor resource use as a function of intensity of care needs.
DR. STARFIELD: Okay, the second this, and this may sound strange to you, but what you are doing is very comfortable for me because it looks like an academic exercise.
It doesn't look to me like a real world exercise. It is internally extremely elegant, but I am wondering about its external relevance because what you have here is a system that is looking at one aspect of a non-conventional health approach, home care, and yet not taking into account a population focus where different segments of the population may fall in and out of home care. They may substitute other kinds of care for home care. So, it is not clear to me while this is very internally elegant and very au courant in terms of its emphasis on outcome, is there really a population-focused approach to looking at outcomes?
DR. SHAUGHNESSY: Kathy, I know you want to talk about the agency side, but let me comment on sort of the, I think the overall topic which is it is not population based and the like. That is absolutely right, on the one hand, but that is true of every single outcome that has been talked about today in terms of if we want to monitor for example, subacute care providers and we really want to see how they are doing, one has to take a look at the patients that they are caring for despite the fact that they are going through the entire post-acute care system, through the entire health care system. I think our dilemma right now that we are all facing is that we all know the ideal way to do this somehow is to collect information and follow patients longitudinally and collect information on them over time and somehow measure outcomes, the outcomes, if you will of the system that is caring for them, and managed care is pushing us in that direction, but I think our own concepts are pushing us in that direction, as well.
We are a good way away from being able to do that in my opinion. I would hazard the guess that we are at least a decade if not more away because there are so many complexities. When one looks at the different types of providers we are talking here about home care and trying to evaluate home care providers which Medicare must do because they pay these providers a lot of money, and we say, "Okay, let us at least see how home health care agencies are doing in terms of the types of beneficiaries and patients that they are caring for. Let us do the same thing with nursing homes. Let us do the same thing with rehab facilities, recognizing all the while that we are simply on the right path, that all we are able to do is monitor the effectiveness of different types of care providers." That is all we are doing here in home care. That is all we are doing with even the MDS when we are using it for outcome measures and the like in nursing homes, and we have to take that next step of saying, "Okay, now, let us rise above all of these different modalities of care and track patients over time," and once you do that you get into the issue not only of common data items for tracking, but in my opinion you get into much more complicated issues related to how frequently should the data be collected; what is the outcome interval; how do we risk adjust; are there different risk factors actually operating under different modalities of care, and all that has to be thought through before we can take a true population based approach.
So, it is my belief at least that I have no trouble whatsoever with that approach conceptually, but in order to get to it we cannot wait until we design the perfect system. We have to in the meantime make sure that the providers, that in this case HCFA is paying a great deal of money to are at least doing the job that they should be doing, and that is only an approximation of what you are talking about.
DR. STARFIELD: That is why I started off asking about the other countries because they are, also, paying for the care, too.
DR. SHAUGHNESSY: Yes, I personally, don't think that there are a whole lot better systems out there in terms of tracking people over time to evaluate the total health care system and then determine which providers ought to be improved and changed in what ways, but Kathy I think wanted to comment.
MS. CRISLER: Yes, I wanted to comment from the standpoint of the complexity for the providers because that frankly is not congruent with what we hear at all. What we hear is that what agencies are currently doing in terms of their own internal performance improvement activities fits beautifully within OBQI from the standpoint that agencies have been predicting or have been using quality indicators as it were for years, except that where the quality indicators come from is sort of they are plucked off the tree or out of the sky. In other words, if I am at lunch with a group of fellow quality assurance or quality improvement people then what we talk about, what are you monitoring in your agency; what are your quality monitors and so on, what they have and what they come from with from the outcome reports, they have the opportunity to actually see a good foundation, aha, this is my benchmark performance either against a reference database or my benchmark performance against my own performance last year and have the opportunity then to swing into action as it were with the current activities that they do that they use in terms of their own internal agency.
So, they are very appreciative of having a starting point that they feel is much sounder scientifically based than anything that they had in the past and that certainly guides at least the starting point for their activities.
DR. IEZZONI: Kathryn, may I follow up on that by asking two questions, one really quick one? Have the home health agencies in your study been a random sample of home health agencies? Have they been representative of small, big, medium, not well endowed, well endowed?
MS. CRISLER: Actually across the approximately 170 agencies that are participating in all of the various projects when we did the selection process for the national Medicare demonstration which involves the 54 agencies that was purposely designed to indeed take into account all of the four various sized categories that HCFA currently uses in terms of defining agencies.
Therefore there are two agencies in that sample, for example, who admit less than 150 patients a year. They both average right around 100 patients. That is a small home health agency, believe me, but that sample was chosen from the standpoint of being a leadership sample. So, they were not randomly selected. They are national, but they were not randomly selected.
The agencies that we have as part of the prospective payment demonstration component, we inherited. Actually Abt Associates picked those agencies in the five states in which those agencies are participating and in the New York demonstration we, also, have a very wide variety in terms of size, some licensed, not certified agencies, again who are very small, again, admissions of probably less than 100 patients, and so, they represent facility-based hospital-based, nursing home based, you know, free standing VNAs, non-VNAs, etc.
DR. IEZZONI: My second question betrays an embarrassing level of cynicism on my part, okay, so I just lead with that. When you say that the home health agencies love OASIS and are very happy with it, I am not at all surprised by that, and I want you to correct my misperception if there is one. What I hear you saying is that what is happening in OASIS is that a nurse or a home health aid or whoever it is goes out and assesses from her point of view the patient's capacity to do certain things and then 6 weeks later or 60 days later that same person goes out and assesses from her point of view that person's capacity to do things.
Now, if I were a human being and I knew that I were being judged I would say, "Oh, gosh, you know the first time this person wasn't that able to do this stuff, but lo and behold 60 days later this person is able to do this," and so it doesn't surprise me given that what you have set up is a system where a nurse or a home health aid is saying that the patient can do this and then a little bit of time later saying that the patient can do this, and it is always that person's own point of view that they themselves are reporting about they themselves, and so it doesn't surprise me that they would be happy with that.
MS. CRISLER: First of all, Lisa, I think there are lots of people who would regard any data-driven system from that same degree of cynicism. So, I understand that. You have actually addressed several points in there from the standpoint of -- and let me start by saying, one, that the home health aid does not actually complete this. At the present time we are looking at skilled providers. So, we have yes, the discipline mix that is found in home care, in other words the nurse, physical therapist, occupational therapist, speech therapist and the medical social worker.
DR. IEZZONI: It is still the person who is giving the care.
MS. CRISLER: Exactly, but it could be --
DR. IEZZONI: So, it is still a nurse.
MS. CRISLER: But it could be a different one at start of care as opposed to what it would be at discharge.
DR. IEZZONI: But they want their agency to look good.
MS. CRISLER: If they have figured that out, and that is a piece that is of particular interest as we look at the second outcome report and the next outcome report within the agency, the potential for gaming which is exactly what you are talking about, and I think that Pete is dying to get at the microphone, but I think that what we have, I mean that is clearly there, and I would not deny that that possibility were there.
I think from the standpoint of building data audits to be done within the context of an agency itself is an absolutely critical --
DR. IEZZONI: But the audit cannot be simply going back and retrospectively seeing whether the same assessment would be completed based on the documentation in the record. It would have to be a second visit to that patient.
MS. CRISLER: Exactly.
DR. IEZZONI: Which means a burden on the patients, you know, having to have somebody come and review them again, you know simply so their agency can have the quality of their data be checked up on. I think it is intrusive to patients.
MS. CRISLER: Although agencies conduct supervisory visits presently, supervisory visits of several different kinds. For example, I as the nurse supervisor may make either a visit with or immediately following. I as the rehab supervisor may make a visit with or immediately following my rehab staff, and that certainly is a way to conduct, also, a supervisory visit, given the way that an agency presently functions and so on.
DR. IEZZONI: Because the minute this starts getting used for reimbursement I just think the gaming potential is enormous and unlike PPS for hospitals where you went back and retroactively or retrospectively checked ICD-9 CM coding this is a whole different level of need for monitoring, especially when you say that there are some home health agencies that are so small that they only enroll 100 people a year. How are you going to be able to design a prospective payment system for them, but are there other people who want to leap in here to save me from betraying more cynicism?
DR. SHAUGHNESSY: Let me respond a bit to this because this is a very critical issue, there is no doubt. First of all, there are incentives that actually run in the same direction. I am not talking just about OASIS here. I am talking about all data capture systems independent of what the domain is, what the modality is we are talking about. If you are talking about measuring outcomes in essence change in health status over time, and what you are looking for is say improvement or at least stabilization over time, the incentive that any provider has, if you are going to collect data from providers and that unfortunately is the most efficient way to do it; to do it some other way is just immensely expensive. There is no easy way to get around that, but if you are going to collect data using providers their incentive for any type of outcome system, and it doesn't matter whether it is hospital or not is to try to make the patients look sicker, if you will, at start of care and look better at follow-up than they really are.
That is a generic issue that goes well beyond home care and OASIS and then when you come around to payment there is, also, an incentive depending on the time point on which you are going to condition payment, and this is something Carrie was talking about is unknown in home care, and as a matter of fact it is rather unknown in other types of health care. We just made decisions on time points where we are going to collect data and say, "Bango, that is it. We are going to collect data at this time, and then we are going to use it for case mix adjustment," but I won't get off onto that.
The point of the matter there is most payment systems are conditioned with the exception of DRGs where we are talking about discharge diagnosis, but most of them appear to be conditioned on a particular time point early on. So, again, the incentive is to make patients appear sicker than they really are because I am going to get paid more for sicker patients, if you will.
So, those incentives are there. They are there across post-acute care.
DR. IEZZONI: I agree, Peter. It is just your panel had more time. So, I was able to ask the question. You bore the brunt of my tongue on this one.
DR. SHAUGHNESSY: Let me, also, make another quick comment, and that is OASIS in terms of, if you will, the opinion of the provider, that is largely in functional areas, and there is a specific reason why we have gone with capacity-based rather than performance-based measures of functional health status, and it has to do with outcomes, and I will try to summarize it briefly if I can or you may not want me to.
DR. MOR: That was my question. So, proceed.
DR. IEZZONI: Vince, did you want to ask a question?
DR. MOR: No, that was my question.
DR. SHAUGHNESSY: We wrestled with this for years in the early going, and we are still not positive we are right, and we therefore are now conducting or now planning a field trial where we are going to collect the information in precisely the same way on capacity and performance and see how much of an academic question this really is or whether it is a real world question because if the two are correlated about .99 or what have you, then you know maybe it doesn't matter that much. So, that is another piece to the puzzle.
However, the reason we did what we did is in home care you often have, oh, let us just use an example. We will talk about men because they are harder to care for and usually more ornery and that sort of thing, but we will say that a particular fellow is admitted to home care, and he at that time of admission he is able to get around his house. He doesn't have anybody else helping him out. He is able to get around a bit. It is difficult, but he is able to do it without human assistance. He is, also, able to prepare light meals.
After home care, and this is fairly typical after start of care or at least common the daughter moves in. The daughter moves in within a week or so, and what she is able to do is she is able to help him get around better, and she, also, prepares all his light meals so that if the assessment is done at that point using capacity measures and the home care nurse knows that this person still has the capacity to move around without human assistance, and he still has the capacity to prepare light meals, but in fact, he is not. He is not performing that way at that time.
Now, when you translate all this into outcome measures at start of care, or let us just talk about follow-up. At follow-up this person is no longer independent in terms of human assistance.
This person is now dependent using a performance-based measure. Is he actually getting around by himself? Yes or no. Is he actually preparing light meals? Yes or no. What happens is his outcome either remains the same or it drops off, but in fact, that home care nurse knows that he is capable of getting around.
They have actually closed the case and his outcome in fact is better or at least the same as it was.
So, you have an invalid measure of the outcome using performance and a valid measure using capacity. The difficulty there is from the viewpoint of validity it makes sense to use capacity.
From the viewpoint of reliability you have the problem that you are talking about, and there is a trade-off between validity and reliability in this instance.
DR. IEZZONI: But, Peter in responding to Barbara's question earlier you said that if somebody is capable of doing something they are no longer part of your evaluation of whether they can improve or not, and so that man who is capable --
DR. SHAUGHNESSY: No, that is not what I meant to say if that is the way it came across.
DR. IEZZONI: Barbara said, "What about all the people who can already dress themselves?"
MS. CRISLER: That was on the improvement measure, and okay, if that is specifically what you were asking, Barbara, then I misunderstood, also. From the computation of that improvement measure, yes, if they were fully independent and able to do that at start of care, they would not be included in the computation of the improvement measure for dressing upper body.
They would be included in the computation of stabilization measure because they could conceivably decline.
DR. SHAUGHNESSY: And, also, for that matter they could be independent of human assistance but still dependent. So, it is only if they are completely independent are they excluded.
In this case the example I was using, this person might need a cane or something like that to get around. So, he is still dependent and still would enter into that improvement measure.
DR. IEZZONI: Does that answer your question, Vince?
DR. MOR: Yes. Are there other contexts in which the differentiation -- you have conceptualized the disconnect between capacity and performance as exclusively in relation to the inability to accurately or validly display a measure of improvement or deterioration as the case may be.
DR. SHAUGHNESSY: That is right.
DR. MOR: And that is particularly complicated for those general things because men don't do anything. Women do it for them. Are there other contexts in which that makes sense as well, because otherwise it is kind of an empirical thing.
DR. SHAUGHNESSY: It is the functional areas that I just illustrated by and large.
DR. MOR: Because I think Terry will have very, very shortly empirical data from your study to crosswalk precisely those two perspectives. Is that correct, Terry?
MS. MOORE: We will have interrater reliability data.
DR. MOR: No, no, but you will, also, have the relationship between the same measure on performance and the same measure on capacity, didn't you say that was what you had?
MS. MOORE: Yes, in fact, we have that now. I mean we have some of that now. That is not really part of, I mean we will look at that in the context of actually reliability. We will look at that in the context of training needs for nurses, but it is not part of something that we are going to spend a lot of time evaluating.
DR. SHAUGHNESSY: I think what you are bringing up, Vince is even if they aren't, there could be something done with this data set, but my only difficulty with it is assuming that the data are there on capacity and performance they aren't isomorphic.
There are actually different scales used, and you cannot get one scale from the other. Therefore you cannot really assess their correlation. That is why I mentioned what I did. We are undertaking one where it could be done using precisely the same scales.
DR. MOR: I guess my point, Pete is that if there is a real big stumbling block to the crosswalk of OASIS and MDS-type information it is that conceptual split in the interpretation of the relative valence given to capacity versus performance.
DR. SHAUGHNESSY: In functioning.
DR. MOR: Well, in a variety of ADLs and IADLs. To the extent that is knowable how important that code set of instructions is, that is very important for this whole purpose of today which is to talk about how these things can work together.
DR. SHAUGHNESSY: It is.
DR. MOR: So, it would be highly desirable to take a look at that.
DR. SHAUGHNESSY: That is why we are doing what we are doing. It could turn out, as I said that it doesn't really matter that much one way or the other. If it does matter, then I think we all have to come together and make a decision on what I would call the trade-off between consistency across items and validity of outcome measurement.
DR. IEZZONI: Okay, we have about 30 seconds left.
Does anybody have a question to pop into that 30 seconds?
Let me just make an observation then. Pete, I appreciate what you meant when you said that it would be prohibitively expensive to collect data about functioning from anybody other than the care provider.
However, there is a literature that suggests that patients and their providers do not agree on the patient's ability to perform things, and I will tell you it would make me pretty irritated if some physical therapist was saying something about me, and I disagreed with that, and so, you know I think that there might be an issue for some patients around that especially if this tool begins to be used in settings where people feel that they might be vulnerable for having their care cut off at a certain point, no longer getting that benefit, and so, I think that those are all things that people are going to have to look at as you move more towards implementation of a system such as this.
DR. SHAUGHNESSY: I agree, and I think as I mentioned if the results of these empirical trials we are doing are what I suspect they are going to be to move over to a performance-based system is not going to be complicated.
DR. IEZZONI: Right or a patient's perception of their performance. I mean that man you were talking about if you asked him if he could cook a meal, he would say, "Sure." He just doesn't, but he can, and that is her perception.
DR. SHAUGHNESSY: Again, some of this goes beyond even home care, but nevertheless, I agree.
DR. IEZZONI: It does. It is just you bore the brunt of this.
Okay, Barbara and I already caucused among ourselves, and we agree that we don't have any closing remarks except to thank people and especially thank HCFA for hosting us here.
It has been a beautiful facility. We really enjoyed lunch.
DR. STARFIELD: And we are going to come back.
DR. IEZZONI: Although we didn't have enough time for the lunch. So, we really thank you and thank all the people who have really thought very long and hard about what they wanted to tell us in their 15 minutes of allocated time and for bearing with our questions.
So, thank you, and we are adjourned.
(Thereupon, at 4:30 p.m., the meeting was adjourned.)