Currently, there are about 400,000 patients with end-stage renal disease (ESRD) with Medicare as primary payor. When including late-stage chronic kidney disease (CKD), kidney patients make up about 1% of the Medicare primary payor population, and the programs account for a disproportionate 10% of the entire Medicare budget. These numbers are expected to increase as the population with hypertension and diabetes ages.
This year, the Centers for Medicare & Medicaid Services (CMS) issued a report prepared by Arbor Research Collaborative for Health on the first 3-years results of a 5-year Medicare demonstration project that allowed patients with ESRD to join 1 of 3 Medicare Advantage Plans (identified in their report as DMO A, B, and C). Each plan was designed to apply disease management principles, and each started out with certain hypotheses on healthcare interventions that were used in the analyses of the programs. At the end of the 5-year project, the first 3 years were evaluated for outcomes that included:
- Improvement in hospitalization, mortality, and transplantation-related measures;
- Improvement in patient outcomes and processes of care;
- Improved quality of life and patient satisfaction;
- Acceptance by providers; and
- Favorable or budget-neutral cost profile to the Medicare program.
To make this evaluation, CMS compared the demonstration population with ESRD patients with traditional Medicare fee-for-service. For more information, see the full report End-Stage Renal Disease (ESRD) Disease Management Demonstration Evaluation Report: Findings from 2006-2008, the First Three Years of a Five-Year Demonstration . It should be stressed, however, that the evaluations had limitations. CMS assessed outcomes only according to the specific initial hypotheses, and each of the 3 projects employed a number of innovative interventions for improving patient outcomes that were not part of the hypotheses. In addition, although the demonstration ran for 5 years, CMS evaluated only the first 3 years, and then stopped the evaluation. The results then did not take into account the interventions not specified in the initial hypotheses or the last 2 years of the programs, which were showing positive trends over time, major deficiencies of the report.
One project, managed by DaVita, Inc. was evaluated on the basis of the following hypotheses:
- Impact of pharmacist involvement on medication-related problems (MRPs);
- Management of cardiovascular disease and risk factors; and
- Improvement in preventive care processes.
To learn more about this specific plan, Medscape interviewed Allen R. Nissenson, MD, Emeritus Professor of Medicine, David Geffen School of Medicine at UCLA and Chief Medical Officer, DaVita Inc., who described their approach and the results of their project.
Medscape: Could you give some of the background leading up to your involvement in the CME project?
Allen R. Nissenson, MD: Back in 1996, a group of nephrologists, including myself, got together because of the dismal outcomes of CKD and dialysis patients, which we thought were mainly related to poor care coordination. Around that same time chronic disease state management was being tried for conditions like diabetes, heart failure, and AIDS, so we went to Baxter Healthcare and convinced them that there was an unmet need for care coordination in kidney disease. Baxter agreed and funded a company called RMS Disease Management. The following year, we discovered that a large part of the total cost of care was related to vascular access complications, so we convinced Baxter to start a second company, RMS Lifeline, which was to develop outpatient vascular access centers in order to improve access outcomes. We then worked with the National Committee for Quality Assurance (NCQA) on the disease management accreditation program, which they were putting together. In fact, we were the first renal disease program to be accredited by NCQA. In 2003, both RMS Disease Management and RMS Lifeline were acquired by DaVita Inc. Subsequently, RMS Disease Management became VillageHealth and RMS Lifeline became Lifeline Vascular Access. VillageHealth is the group that became involved in the CMS demonstration project.
Medscape: How did the CMS demonstration project get started?
Dr. Nissenson: Initially, when DaVita and Fresenius met with the people at CMS to talk about this project, the CMS plan was to collect as much data as possible from the participants, analyze it, and come up with conclusions. We were very uncomfortable with that approach and said, "Why doesn't each of the groups involved come up with some testable hypotheses based on the way each program is structured? Then you judge the programs according to the results of these hypotheses rather than collecting millions of bits of data. And who knows how you'll deal with that?" They finally agreed to that approach. One problem, however, is that CMS ended up taking that idea literally, so they basically said, "So if the innovations you tested that are not part of the 3 hypotheses, we're not evaluating those." They looked at very general things: basically survival, hospitalization, and cost in addition to the specific hypothesis, but not such critical parts of our efforts such as our catheter removal program.
Medscape: Could you describe how your demo project was structured?
The specific demo project that we participated in was a Medicare Advantage Special Needs Plan, and the whole idea was to coordinate care, improve clinical outcomes, and constrain the overall costs of care. Our program was structured a bit differently from Fresenius' demo project [one of the other 2 organizations involved in this program]. We chose to work in 1 geographical region in Southern California -- Riverside and San Bernardino counties – initially with 1 very large nephrology group that had a large number of patients, many dialysis facilities, and outpatient vascular access centers. So we operated in a fairly controlled clinical environment, as opposed to the structure developed by Fresenius, which included working at several different sites around the country .
In terms of the care coordination, we used field-based nephrology nurses as our care managers. We had nurse practitioners, clinical pharmacists, and additional support staff. The clinicians, particularly the care managers, would interact directly with the patients and physicians, either in the dialysis facility or, at times, in patients' homes.
Medscape: How were the payments structured?
Dr. Nissenson: There was a global payment, which was calculated based on current Medicare Advantage payment rates; 5% of the payment was withheld and could be earned back based on hitting quality targets. This was a global capitated program, so we were responsible for all healthcare costs not just those related to kidney disease.
Medscape: As part of your medication management program, you used clinical pharmacists. Could describe this program and how it evolved over the course of the project?
Dr. Nissenson: Just to place this program in context, dialysis patients take on average 10-12 medications – 20 or more different pills -- every day. This is a daunting task. It's interesting that there are data in the literature to show that MRPs account for or contribute to up to 20% of hospitalizations, so we know this is a big problem in kidney disease patients. We thought that a carefully constructed medication management program could significantly improve the consequences of MRPs.
Initially we tried to identify MRPs using a manual process. A VillageHealth pharmacist, who was part of the care team, would review all the medications and look at [4 potential problem areas]:
- Are multiple medications of the same class being given that are redundant?
- Are medications being given that interact with one another adversely?
- Are medications being given at an inappropriate dose for people with kidney problems?
- Are any medications not being given that should be based on the patient's clinical condition?
As an example [of the last issue], if someone has had a heart attack, it's currently recommended that they be put on a beta-blocker [at discharge] unless there's a contraindication. If this was not done we could alert the doctor to consider doing so. So we would look at these 4 categories and then intervene if we felt we'd uncovered an issue with a patient's medication that might lead to adverse events.
After the first year, we acquired a software program that greatly facilitated this process; we fed the medication [information] directly into a computer program and could immediately see if there was a MRP in 1 of the 4 domains described. When we implemented this approach, we also re-engineered the way we intervened when a problem or potential problem was identified. We had previously had the pharmacist contact the patient directly, but this was getting difficult because we had over 400 patients scattered in many different dialysis facilities. [We switched] to having the nurses work with the patients and their physicians on the MRPs that were identified in the computer program and used pharmacists for backup and support of the care manager. In addition, we initially had started out with just scheduled medication reviews. The review might start when the patients enrolled in the program and then [occur] every quarter or 6 months -- it varied. We switched from that to a more focused set of reviews based on sentinel events. For example, if a patient was in the hospital and discharged, that would trigger a new medication review.
We found with this intense involvement and the use of the drug interaction computer software that we had a significant decrease in some of the MRPs, such as incorrect doses, drug interactions, and patients on drugs where there was no clinical indication. When MRPs were identified, close collaboration between the VillageHealth nurses, doctors, and pharmacists resulted in resolution in a majority of cases.
Medscape. Now, the CMS report mentioned that the number of MRPs increased over the course of the first 3 years. How does that fit with your saying that they declined?
Dr. Nissenson: Part of this is a reporting issue. [Over time] we had all the pharmacy claims available, which wasn't the case at the beginning. Second, when you do quality improvement, before you start, you don't find any problems. Once you begin doing it, you find a lot of problems. [Don't forget] this [part of the program] isn't preventative. Its purpose is to look at and reconcile medications. [Of course] you try to prevent the distal part, which is an adverse reaction or a hospitalization, but you don't prevent the problem when you are defining it.
In addition, part of the problem was the definition of MRPs. I distinguish these from adverse events, which are the consequences of an MRP. There's no evidence that the adverse events went up; we tried very hard to demonstrate that they went down. However it's difficult to tease out what's directly attributable to the medication problem. The bottom-line was that, 95% of the time, any recommendation to change the medicine, the dose, or to stop a drug was accepted by the doctors, and the medication was changed. So, we're confident that this program had significant impact on clinical outcomes, although it is not easy to be completely definitive.
I want to mention one thing we accomplished that I think does suggest the program was effective. As part of it, all patients' medications are in the computer system. We have really good data on the medicines they're taking. We found that the medication possession ratio (MPR), which is used in the pharmacy industry [to indicate] adherence, was significantly improved in patients where there was an interaction with a nurse or pharmacist about medications. MPR is reported in the United States Renal Data System (USRDS), so we had something we could compare ourselves with, and we had significantly better adherence, or MPR, for statins, beta-blockers, angiotensin-converting enzyme (ACE) inhibitors, and H2 blockers. This also suggests that the medication management program was effective in helping the patients.
Medscape: Is this medication management program going to continue?
Dr. Nissenson: We have a program called DaVita Rx, an in-facility medication pharmacy program, which is unique to DaVita. The patients order their medication, and the oral medicines are delivered to them at the dialysis unit. So we know the patients get the medicine, and we have a really good tracking system of what the medications are and how frequently they're renewed, which is a surrogate for compliance. That is an ongoing program for all of DaVita. We have about 35,000 patients who are in that program right now, and we have results that are very similar to those we saw in the demo for the patients who are in that part of the program -- better medication adherence, better intermediate outcomes related to medications such as phosphorus levels, fewer hospitalizations, and lower costs.
Medscape: Your third hypothesis was that cardiovascular disease would improve under your management program. Was this successful, and if not what should be the key message of this experience?
Dr. Nissenson: We looked at different aspects of cardiovascular health with respect to control of lipids, use of ACE inhibitors, and control of blood pressure and the bottom-line was that we had a minor impact. We did better on implementing processes, such as improving [the number of] people getting cardiovascular drugs -- at least at the beginning -- and measuring lipids. However, at the end of the day, we may have been overly ambitious with this [part of the project]. There are so many factors that go into cardiovascular health that this was just much more difficult than we had anticipated. I think the message is not that this is unimportant, but that there would have to be a much more structured and holistic approach to really have an impact. Also, unfortunately, by the time people get on dialysis there may be less we can do than if we got to them earlier, so this approach might be more effective in the CKD population than in the dialysis patients.
Medscape: Your demo project was quite successful with improving preventive care, including immunizations and care of diabetes. What was the process for making those improvements?
Dr. Nissenson: Each of the patients in the program had a care plan, which was developed with the patient, the patient's nephrologist and the VillageHealth nurse. The care plan included attention to key preventative care behaviors that are currently recommended for diabetic patients such as getting foot checks, retinal exams, and controlling blood sugar. The care plan also included a focus on ensuring patients received appropriate immunizations, which we administered at the dialysis facilities. The nurses provided education -- why this preventative care was, and how to get preventative care recommended done. They also helped facilitate getting appointments, such as for retinal exams. This was a very intense focused collaborative effort with our care managers and the patients working together on this. With this population, we know that success in preventive care has a significant and tangible impact on hospitalizations and mortality.
By the way, this experience has been great for us throughout DaVita. We are able to use this demo as a kind of laboratory and then scale these approaches to the whole company. Our current success with preventative care in the whole company is now unbelievable.
Medscape: I found the success of that program vs the cardiovascular program interesting.
Dr. Nissenson: Well, you have to realize that preventative care is way more concrete and there are no side effects [compared with taking cardiovascular medication].
Medscape: I'd like to hear more about the catheter reduction program, which wasn't included as part of the hypotheses in the CME project.
Dr. Nissenson: We call the program CathAway, which we developed as part of the VillageHealth demonstration. It takes a holistic view of dialysis vascular access, including avoiding/removing catheters and ensuring that an AV fistula is placed and used whenever possible. The high percentage of catheters [in dialysis patients] ends up being 1 cause, if not the leading cause, of hospitalization, sepsis, and mortality. So, we [asked ourselves] what needs to happen to take a patient from having a catheter to not having one and having permanent access? We broke it down into 7 steps and developed software to track it.
Each of our VillageHealth nurses uses this software to track every patient with a catheter through the process. They make sure by the end of the 7-step timeframe, which ended up being 4.5 months, that each patient has permanent access and no catheter. We actually managed to achieve this. According to USRDS data, catheter rates are still hovering in the low 20% and in this program the catheter rates went down to about 8%. We've applied that same process to the whole company, and DaVita is now down below 17.4%, which is better than what anyone else has done.
Medscape: After 3 years of evaluation, your project showed a decrease in hospitalization and mortality. However, when compared with a CMS-determined control group, these rates were no longer significantly different from Medicare. Do you believe that your program truly improved these primary outcomes, and if so, why did the comparisons of the control group not show this?
Dr. Nissenson: When we look at our data over time, clearly, we had significantly better outcomes, both hospitalization and mortality than reported for FFS patients. For example, the 2-year survival rate for our patients was 72%, and for the same time period, fee-for-service was 66%, almost a 10% improvement. Our hospitalization rates were 69% vs fee-for-service rates of 74%. That's about an 8% better performance. Just looking at the raw data, we were very confident that we were improving hospitalization and mortality. The problem was with the control group. As you know, the best way to do any kind of a study like this is to do a randomized control trial (RCT).
Because it was impossible to assign people either to our disease management group or not, to do a true RCT, [CMS] instead used a statistical method known as propensity score matching. They tried to find a similar population and then pick as many characteristics in the treatment group and in the control group that they thought would be important in driving outcomes, and then they matched them up and adjusted the results as if they were randomly assigned groups. This is a legitimate statistical technique, but just like randomization, it sometimes doesn't work. What I mean by that is you can organize controlled trials so that people are randomly assigned to group A vs group B, but then when you look at the 2 groups, it sometimes turns out they are different in important ways.
The propensity matching done as part of this project evaluation we believe suffered from the same kind of flaw. I'll just give you 1 example. In our demo project, a quarter of all the patients in the project were incident patients. They were new patients. In the control group, only 11% were incident patients. We know that incident patients have higher hospitalization rates, higher mortality, and higher costs than prevalent patients -- stacking the deck against our demo group. A separate control group was created for each of the 3 demonstration projects. When you look at the other major demo project, the fraction of incident patients between their demo and the control groups was fairly similar. We had over twice as many incident patients, which we believe significantly voided the validity of the propensity match. So, this is just 1 example of how the groups weren't comparable.
Medscape: Is there any way of relooking at that control data?
Dr. Nissenson: We've actually hired an independent third party to reanalyze the CMS data and to look at other ways to do a matched control analysis. We hope to have this done within the next 6-9 months. If CMS got the right answer, we want to know that, and if we confirm that we in fact significantly lowered hospitalization rates and cost and improved survival, as I suspect we will, we want to know that as well.
Medscape: Will CMS evaluate the remaining 2 years?
Dr. Nissenson: CMS agreed that there are a lot of issues with the analyses, but that their part of the project is over. They're done. Although this is a 5-year demo, the report only covers the first 3 years. We thought that's fine, but then at least make sure the reader knows this is an interim analysis, and analyze the last 2 years subsequently, because these projects are called demos for a reason -- they're for learning, and we continually improved our outcomes over time. We think omitting the final 2 years in the analysis really understates the accomplishments of the demos, but unfortunately [3 years] was all CMS was budgeted for.
Medscape: Once you finish the whole 5-year analysis, can you submit the results back to CMS?
Dr. Nissenson: Our plan is to [send our results] to peer review journals and CMS clearly has access to those publications.
Medscape: One of the primary premises of integrated care is that outcomes will improve while costs will be constrained. Because the evaluation did not look at actual costs, but Medicare Advantage payments instead, which are acknowledged by the report to exceed fee-for-service payments, it is unclear if savings were actually seen. What is your interpretation of the impact of your program on overall health costs for your enrollees?
Dr. Nissenson: The report really doesn't assess the cost of care in the demonstration; it just describes payments. The payments were structured based on current Medicare Advantage rates, and as was reported and is well-known, Medicare Advantage plans are paid more -- about 7%-12% -- than Medicare fee-for-service for the same services. That analysis is entirely unrelated to the key question raised in the demo- did care coordination lead to lower overall costs of care for these patients?
What our data show, again this is with our own analysis, is that we were able to save about $1500 per patient during 2008, the third year of the demo. In 2009 and 2010, we continued to improve in terms of inpatient cost, which is really what drives most of the total cost.
Medscape: CMS withheld funds to be paid back to the DMOs if they achieved important quality targets. How did you succeed in hitting these targets, and how did you do relative to fee-for-service and the other projects?
Dr. Nissenson: This question is important because quality measures are a major focus of CMS, and some of these process measures correlate with primary outcomes. This part of the demonstration also had very important financial consequences for us, because CMS withheld 5% of the payments due to us and we had to earn those back based on hitting the quality targets. Over the course of the demo, we had to hit 69 quality targets in various domains. The targets were based on a comparison with national benchmarks and also improvement within our program. So they not only looked at our absolute performance compared with the benchmark but also how much we improved from our own previous performance. We hit 64 out of the 69 targets and earned almost all of our quality with hold back.
We did a little bit better than the other major demo project, but they did very well also. The point is that the fee-for-service world would not hit anywhere near this number of targets. This was just another affirmation that, at least for these more typical dialysis related process metrics, care coordination can really improve outcomes. I should conclude that the 3 demo sites were never in a competition. We all did different things. [The other major project] did home weight monitoring, blood sugar monitoring, and used nutritional supplements. We had CathAway, immunization programs, and medication management. The message to me from the whole project is care coordination is needed and can improve clinical outcomes while constraining costs. If you mix it all together and pull out the approaches that worked and don't do the things or improve those that didn't work, you can significantly decrease mortality, hospitalizations, and lower the overall costs of care. The CMS demonstration project really helped all of us who were engaged in it to understand where we need to focus and get the most bang for the buck. It positions the renal community well for the new world of integrated care management and Accountable Care Organizations.
- U.S. Renal Data System, USRDS 2009 Annual Data Report: Atlas of Chronic Kidney Disease and End-Stage Renal Disease in the United States. National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases. Vol 2; Bethesda, MD; 2009:191, Figure 1.