By Deborah Peikes, Greg Peterson, Randall S. Brown, Sandy Graff, and John P. Lynch

Transcription

1 doi: /hlthaff HEALTH AFFAIRS 31, NO. 6 (2012): Project HOPE The People-to-People Health Foundation, Inc. By Deborah Peikes, Greg Peterson, Randall S. Brown, Sandy Graff, and John P. Lynch How Changes In Washington University s Medicare Coordinated Care Demonstration Pilot Ultimately Achieved Savings Deborah Peikes (DPeikes@ mathematica-mpr.com) is a senior researcher at Mathematica Policy Research, in Princeton, New Jersey. Greg Peterson is a health researcher at Mathematica Policy Research. Randall S. Brown is a vice president and director of health research at Mathematica Policy Research. Sandy Graff is director of care management in the Department of Internal Medicine at the Washington University School of Medicine in St. Louis, in Missouri. John P. Lynch is vice president and chief medical officer at Barnes-Jewish Hospital and a professor of medicine at the Washington University School of Medicine in St. Louis. ABSTRACT As one of the initial fifteen participants in the Medicare Coordinated Care Demonstration, the Washington University School of Medicine in St. Louis was not able to demonstrate any reduction in hospitalizations or Medicare spending for the patients it served. In fact, the Washington University program increased total Medicare spending by 12 percent. But after a redesign, the results changed. The program stopped conducting care management of most of its patients via telephone from a remote site in California and, instead, served all patients through frequent phone and occasional in-person contact from local care managers in St. Louis. Care management efforts were focused especially on patients deemed at greatest risk of hospitalization, and stronger hospital transition planning and medication reconciliation were provided, among other changes. After that point, the program reduced hospitalizations by 12 percent and monthly Medicare spending by $217 per enrollee more than offsetting the program s monthly $151 care management fee. The results underscore findings from the overall Medicare Coordinated Care Demonstration that suggest that programs with more in-person contacts were more likely than others to build trusting relationships with patients and providers, improve patient adherence to care plans, and address additional needs and barriers that entirely telephonic contacts had been unable to identify. The results also indicate that programs can be more effective by focusing on the highestrisk patients, for whom the largest savings resulted. Improving care for people with chronic illness is a cornerstone of efforts to generate savings and improve quality because this care accounts for a disproportionate share of costs and is a key driver of increased Medicare spending. 1 Care management is one promising approach that is not currently covered by fee-for-service Medicare. The goals of care management include preventing expensive and unnecessary hospitalizations, as well as rehospitalizations after an initial discharge. This goal is to be accomplished by increasing patient and to a lesser degree provider adherence to evidence-based treatment guidelines, while coordinating care across providers. Reflecting the promise of care management, more than half of large employers offer it in their health plans. In addition, the accountable care organization and patient-centered medical home models require it. 2 4 The evidence to date on care management is mixed. Many studies have shown only modest effects on quality and overall cost increases because savings, if any, in regularly covered health 1216 Health Affairs JUNE :6

2 care services are smaller than the additional costs of the care management interventions. 5,6 However, recent evidence, including the results of the study on which we report here, suggests that strong care management models, delivered to the right people, can reduce hospitalizations and possibly costs A Randomized Trial Of Two Interventions This article examines how Washington University School of Medicine s care management program in the Centers for Medicare and Medicaid Services (CMS) Medicare Coordinated Care Demonstration reduced hospitalizations and Medicare spending, following a major program redesign. The redesigned program provided all participants with the same type and intensity of care management services that had previously been provided to just 20 percent of enrollees. This shift provided universal access to stronger transitional care after hospitalization, more comprehensive medication management, and occasional in-person contacts by the care managers with patients and providers. Including Washington University, nine of the eleven extended demonstration programs responded to a February request, with telephone follow-up, for more detailed information on specific approaches that we hypothesized were playing an important role in programs successes. The features adopted by Washington University after the redesign were similar to the features that distinguished the other three programs that succeeded in reducing hospitalizations for highrisk Medicare beneficiaries from the five programs that did not reduce hospitalizations. 17 The redesigned Washington University program also introduced streamlined and standardized care plans that care managers said made it easier for them to focus on the most important goals. A key strength of the study on which we report is that it included random assignment of subjects to care management and usual care groups. This ensured unbiased estimates of the impact of care management. All study enrollees, whether in the care management or usual care group, continued to obtain their traditional Medicare-covered services from fee-for-service providers in the usual manner. The Medicare Coordinated Care Demonstration In early 2002 CMS selected Washington University School of Medicine in St. Louis as one of fifteen demonstration programs (out of fiftyeight applicants) in its Medicare Coordinated Care Demonstration. Mandated by the Balanced Budget Act of 1997, the demonstration tested whether paying the fifteen programs to provide care management to Medicare beneficiaries in fee-for-service, in addition to whatever services they received from their usual health care providers, either reduced total Medicare spending or increased the quality of health care services and satisfaction of beneficiaries and providers without increasing total expenditures. Total expenditures included Medicare payments for the regular Medicare-covered Part A and B services that enrollees used as well as the additional care management fees that CMS paid the demonstration programs. The per beneficiary per month care management fee varied across the fifteen programs and over time. CMS and Washington University negotiated a monthly fee of $ for the first year of the demonstration, $ for the second year, and $ for the remaining years. Programs began enrolling patients in 2002 and were authorized to operate for four years. Eleven of the fifteen programs later received twoyear extensions so they would not have to cease operations while waiting for the Medicare claims data needed to assess program effectiveness through the original 2006 end dates. CMS selected Mathematica Policy Research to conduct the independent evaluation of the demonstration. Detailed cost and quality results of the demonstration through June 2006 have been reported elsewhere. 6 From August 2002 through February 2006, Washington University provided care management services to eligible Medicare beneficiaries in the St. Louis, Missouri, area who agreed to participate in the demonstration and who were randomly assigned by Mathematica to the care management group. The intervention was implemented in partnership with StatusOne, a population-based health management vendor. Washington University s care managers, based in St. Louis, cared for the 20 percent of patients who were deemed the most complex based on conditions, unmet needs, caregiver resources, and recent history of acute service use. StatusOne managed care for the remaining 80 percent of patients from a remote call center in San Diego. Then, following a program redesign in March 2006 through July 2008, Washington University s care managers provided all care management services. Serving High-Risk Medicare Beneficiaries Throughout the demonstration program s six years, Washington University enrolled 2,780 JUNE :6 Health Affairs 1217

3 Exhibit 1 adult fee-for-service Medicare beneficiaries living in the St. Louis metropolitan area. Prospective participants initially were identified by a proprietary algorithm of StatusOne that was applied to the Washington University physicians professional claims data, and by direct referral from health care providers. StatusOne s algorithm used two years of claims data to identify patients likely to become clinically unstable and require a hospitalization in the following twelve months. The algorithm was rerun monthly. Dedicated recruitment staff explained the program to eligible beneficiaries in person and invited them to participate. The 20 percent who accepted were randomly assigned to the care management or usual care group. After the program s redesign, Washington University used hospital data and provider referrals to identify high-risk patients for enrollment. The program enrolled Medicare beneficiaries who were much sicker than average beneficiaries (Exhibit 1). Enrollees were two to three times more likely to have congestive heart failure, coronary artery disease, or diabetes than Medicare beneficiaries nationwide. Their average monthly Medicare expenditure in the year before enrollment was $2, times the national average in Furthermore, 26 percent of enrollees were eligible for Medicare because of Characteristics Of Enrollees In The Washington University Care Management Demonstration At Program Enrollment, And Of Medicare Beneficiaries Nationwide, 2005 Characteristic Enrollees Medicare beneficiaries nationwide Younger than age sixty-five 27% 14% Age eighty-five or older Black, non-hispanic 38 9 State Part B buy-in a Entitled to Medicare based on disability b Congestive heart failure c Coronary artery disease c Chronic obstructive pulmonary disease c Diabetes c Number of annualized hospitalizations per year d Medicare Part A and B spending per month d $2,498 $552 SOURCES For Medicare entitlement reason for beneficiaries nationwide, the Medicare Current Beneficiary Survey. For all other characteristics, authors analysis of the Medicare Enrollment Database, National Claims History File, Standard Analytic File, and the 2005 Medicare 5% File. NOTES For the care management group: n ¼ 1; 397. For the usual-care group: n ¼ 1; 383. For Medicare beneficiaries nationwide: n ¼ 30:6 million. We assessed whether enrollees had chronic obstructive pulmonary disease using Medicare claims from the year prior to enrollment in the program, and whether they had other diagnoses using claims from the two years prior to enrollment. a Proxy for Medicaid coverage. b We assessed entitlement due to disability based on the original reason for entitlement for demonstration enrollees and the current reason for entitlement (in 2005) among Medicare beneficiaries nationwide. c Diagnoses are not mutually exclusive. d Average individual medical care use during previous year. disability much higher than the national rate. Reflecting St. Louis demographics, 38 percent of enrollees were non-hispanic black (versus 9 percent nationally). The Original Care Management Intervention During the first four years, care managers in St. Louis and San Diego assigned patients to one of five levels of acuity, or risk, based on disease; physical, social, psychological, and financial status; and ability to self-advocate. Some criteria were objective, such as whether the patient had been discharged from a hospital within the previous six months. The care manager had to assess subjective criteria, such as a patient s level of health literacy. The risk levels reflected the patient s likelihood of being hospitalized or visiting an emergency department over the next six months. Because depression can both exacerbate and be worsened by chronic conditions, care managers were instructed to screen patients for depression at intake and quarterly thereafter, using the twoquestion Patient Health Questionnaire Guided by the assessment process and depression screening, care managers developed individualized care plans for their own use and routinely contacted patients. Care managers from StatusOne contacted patients by telephone only, but care managers from Washington University made contacts by telephone and in person. The most acutely ill patients, many of whom also had serious psychosocial needs, were contacted every one to two weeks; the least acutely ill were contacted every four to six weeks. St. Louis care managers served the most complex 20 percent of patients and concentrated their in-person contacts on the highest-risk patients in this group. Care plans guided care managers interventions to support patients clinical treatment plans. Care managers educated patients on their diseases and how to recognize and manage worsening symptoms; explained and encouraged the use of special diets, exercise, and preventive clinical care tailored to patients conditions; and encouraged patients to maintain up-to-date medication lists and share them with providers. Care managers also encouraged patients to see their primary care doctors and other providers as appropriate and checked to ensure that these visits took place. To prevent readmissions, care managers contacted patients when they were hospitalized and revised their care plans as necessary. The program was connected with the patients usual physicians and hospitals, which facilitated 1218 Health Affairs JUNE :6

4 Washington University s program redesign increased the frequency of in-person contacts. the intervention. The principal investigator, a Washington University physician, met regularly with physician-advisers; care managers facilitated communication across all involved providers; and the program sent regular updates to physicians. Care managers had direct access to current and past inpatient data from their academic hospital partner, Barnes-Jewish Hospital, and to home care and clinic notes from other providers in the BJC HealthCare system, a nonprofit health care organization that is based in St. Louis. The hospital sent a daily fax of all Medicare admissions to the care management program. The program provided care managers with a list of their patients who had been hospitalized so they could prepare them for the transition out of the hospital. The St. Louis and San Diego care managers differed in their levels of interaction with patients regular providers. San Diego care managers attempted to contact providers but reported difficulty reaching them. This may have reflected the lack of a relationship between the San Diego care managers and patients St. Louis based usual providers. St. Louis care managers contacted the office staff of their patients providers during and after hospitalizations and at other times of crisis for their higher-acuity patients to exchange information, resolve medication discrepancies, and clarify treatment plans and goals. St. Louis care managers also talked to patients physicians when needed and often attended physician appointments with patients. Whenever possible, the local care managers worked with other clinicians on the treatment team instead of physicians, to avoid burdening them. Study Data And Methods We used a randomized experiment with an intent-to-treat design to estimate the program s impacts. The sample included all enrollees, regardless of whether they received care management services.we calculated program impacts on spending per month and hospitalizations per year, using Medicare Part A and B claims, which also contained all of the care management fees that CMS paid to Washington University for the demonstration; and demographic and Medicare eligibility information from the Medicare Enrollment Database. We used multivariate regressions to assess program impacts on hospitalizations and Medicare spending. The covariates, prespecified in a detailed design document 19 and listed in the online Technical Appendix, 20 increase the precision of the impact estimates and adjust for any chance baseline differences in the observable characteristics of the care management and usual care groups. The regressions include a binary variable to indicate whether the beneficiary was randomly assigned to the care management group or to the usual-care group.we estimated the program s effects on the following three outcomes: the number of hospitalizations per year and average monthly Medicare Parts A and B spending during the follow-up period, with and without CMS s care management fees. Outcomes were measured from the beneficiary s enrollment in the study through February 2006 for the period before the program s redesign, and from March 2006 through July 2008, the end of the observation period, for the redesign phase. The Technical Appendix provides details of randomization and estimation of program impacts. 20 Study Results Before the redesign, Washington University s program did not reduce hospitalizations or Medicare Part A and B spending (Exhibits 2 and 3). In fact, the intervention increased total per beneficiary Medicare spending by $236 per month (90% confidence interval: $76, $395), or 12.3 percent, after including the care management fees that CMS paid to Washington University (Exhibit 3). Because CMS paid for only active participants, the average monthly fee paid, $167, was lower than the negotiated rates described above. The Program s Radical Redesign When CMS extended the Washington University program in March 2006, Healthways, which had acquired StatusOne, stopped participating so it could focus its demonstration infrastructure instead on CMS s larger Medicare Health Support pilot program. As a result, Washington University care managers assumed responsibility for the 80 percent of enrollees that StatusOne had managed remotely. The university s care managers provided all care management locally until JUNE :6 Health Affairs 1219

5 Exhibit 2 Impact Of The Washington University Care Management Demonstration On Rates Of Hospitalization, Before And After Program Redesign Number of enrollees Care management group Usualcare group Average follow-up (months) Average annualized hospitalizations, usual-care group a Program impact on annualized hospitalizations Impact as percent of the mean b All enrollees before redesign 1,078 1, All enrollees after redesign 1,087 1, ** Higher-risk enrollees c before redesign Higher-risk enrollees c after redesign ** SOURCE Authors analysis of data from the Medicare Enrollment Database, National Claims History File, and Standard Analytic File. NOTES Program impact is the regression-adjusted difference in the annualized rate of hospitalizations among enrollees in the care management group versus those in the usual-care group. The impact estimates for the period before redesign include people who enrolled from August 1, 2002, through February 28, 2005, and outcomes from August 1, 2002, through February 28, The impact estimates for the period after redesign include people who enrolled from August 1, 2002, through July 31, 2007, including those who enrolled on or after March 1, 2006, and outcomes from March 1, 2006, through July 31, The estimates for the period after redesign exclude enrollees who were not eligible for the program for at least one month from March 1, 2006, through July 31, 2008; in most cases, ineligibility was because the beneficiary died. The Technical Appendix contains details on the regression method used for estimating impacts (see Note 20 in text). a Usual-care group mean per person per year. b Program impact divided by average annualized hospitalizations in the usual-care group. c Higher-risk enrollees met program eligibility criteria and had two or more hospitalizations in the two years before enrollment. **p < 0:05 the program ended on July 31, In retrospect, the transfer of responsibility marked an unexpected, but nevertheless critical, turning point for the better. Although the program was free to refine the intervention at any time during the demonstration, the need to absorb StatusOne s patients using Washington University s local approach, and to stop using StatusOne s care planning software, spurred the program s overall redesign. Exhibit 3 Impact Of The Washington University Care Management Demonstration On Medicare Spending, Before And After Program Redesign Average monthly care management fee paid a ($) Program impact, without fees Program impact, with fees Expenditures b without care management fees ($) Dollars Percent Dollars Percent All enrollees before redesign 167 1, ** All enrollees after redesign 151 2, * Higher-risk enrollees c before redesign 165 2, Higher-risk enrollees c after redesign 149 2, ** * SOURCE Authors analysis of data from the Medicare Enrollment Database, National Claims History File, and Standard Analytic File. NOTES The average monthly care management fee paid is less than the negotiated fee because the Centers for Medicare and Medicaid Services paid fees only for actively engaged enrollees in the care management group. Although a small number of beneficiaries in the care management group withdrew from the program, we kept them in the analysis to maintain the intent-to-treat randomized design. The impact is the regression-adjusted difference in Medicare Part A and B expenditures among enrollees in the care management group versus those in the usual-care group. The impact estimates for the period before redesign include people who enrolled from August 1, 2002, through February 28, 2005, and outcomes from August 1, 2002, through February 28, The impact estimates for the period after redesign include people who enrolled from August 1, 2002, through July 31, 2007, including those who enrolled on or after March 1, 2006, and outcomes from March 1, 2006, through July 31, The estimates for the period after redesign exclude enrollees who were not eligible for the program for at least one month from March 1, 2006, through July 31, 2008; in most cases, ineligibility was because the beneficiary died. The Technical Appendix contains details on the regression method used for estimating impacts (see Note 20 in text). a Mean fees that CMS paid the program for care management services. b Monthly Medicare Part A and B expenditures per enrollee. c Higher-risk enrollees met program eligibility criteria and had two or more hospitalizations in the two years before enrollment. *p < 0:10 **p < 0: Health Affairs JUNE :6

6 Washington University extended to all enrollees the approach it had been using with the 20 percent of enrollees it had managed from the beginning. It also introduced streamlined and standardized care plans, as described below. CHANGES TO STAFFING: Before the redesign, the program had four care managers in St. Louis and eight in San Diego. To replace the San Diego care managers after the redesign, the program hired six new nurse care managers, one licensed clinical social worker (to address patients psychosocial needs), and three part-time care manager assistants, all based in St. Louis. The care managers caseloads increased slightly, to ninety patients per nurse from about seventyfive. The nurses received increased support from the assistants and social worker. In addition, the patients who had worked with StatusOne s care managers were less complex on average than patients already in Washington University s direct care. STRONGER TRANSITIONAL CARE: The redesign expanded strong transitional care services to all patients. Many studies have shown that transitional care from hospital to home can reduce readmissions. 7,11,14 San Diego care managers had talked to patients in the hospital and tried to follow up by telephone within two weeks of discharge. In contrast, care managers from Washington University generally met in the hospital with hospitalized patients, their caregivers, and their providers; helped with discharge planning; and contacted patients within two days of discharge. The care managers focused on whether the patient understood and followed the discharge plan, including taking prescribed medications and making appropriate follow-up appointments. MORE COMPREHENSIVE MEDICATION MAN- AGEMENT: After the redesign, the program used the Washington University care managers more active approach to managing medications for all enrollees. Increasing patients adherence to appropriate medications and decreasing duplicative or conflicting medications can improve health outcomes and reduce costs among chronically ill patients. 15,16 The San Diego care managers had encouraged patients to develop their own medication lists and to share them with providers to eliminate duplicative or conflicting prescriptions a process known as medication reconciliation. The St. Louis care managers continued to educate patients about the side effects of newly prescribed medications, use software to maintain and update the list of each patient s medications during each contact, resolve any apparent discrepancies with pharmacists or the treating physicians, and share revised lists with patients but after the redesign, they did so for all patients. The managers also annually counseled patients on selecting a Medicare Part D plan that best covered their needs, after Part D became effective in January As a result, beneficiaries in the care management group were four percentage points more likely than beneficiaries in the usual care group to have Part D coverage at any point from 2006 through MORE IN-PERSON CONTACT: In-person (versus phone) contact between care managers and patients, caregivers, and providers can be important to the success of care management. Among the eleven Medicare Coordinated Care Demonstration programs that CMS extended, those with more in-person contacts were more likely than others to reduce hospitalizations and emergency department use. 17 Washington University s program redesign increased the frequency of in-person contacts by using the local care managers approach for all patients. Care managers met with patients, caregivers, and providers in the hospital, providers offices, or patients homes. This helped them build trusting relationships with patients and providers, which improved patient adherence to care plans and coordination with providers; helped patients establish a relationship with a usual source of care; and discovered and addressed additional needs and barriers that entirely telephonic contacts had been unable to identify. Before the redesign, the 80 percent of patients managed by San Diego care managers were contacted in person only if the care manager specifically requested that a St. Louis care manager follow up with a patient, which occurred only occasionally. Nonetheless, even after the switch, 90 percent of contacts were by phone, to ensure efficiency. Each patient received an average of 1.47 contacts per month. Only 5 percent of these contacts, or an average of 0.76 per patient per year, were in person. Eighty percent of them occurred in providers offices, the hospital, or skilled nursing facilities; the rest were in patients homes. The remaining 5 percent of the total contacts were mailings that reinforced the support offered during a previous contact, such as a diet plan or a list of community agencies that provided social services. In-person contacts focused on patients with significant clinical and psychosocial needs. MORE-THOROUGH ASSESSMENTS OF UN- MET NEEDS: The redesign made the care managers assessments and monitoring of patient acuity, or risk, and unmet needs more thorough and systematic in two ways. First, the care managers supervisor ensured that all care managers used the assessment process to assess patients acuity levels at each contact and used the depres- JUNE :6 Health Affairs 1221

7 sion tool described above quarterly. Although the acuity scoring guidelines and the depression tool were used before the redesign, their use was systematic only after the redesign. Second, the care managers supervisor reviewed care managers contacts daily to ensure that they contacted patients at the recommended frequency and provided guidance on care planning. The new approach to assessing patients needs strengthened the intervention in two ways. First, care managers identified medical and psychosocial needs that had previously been undetected because of the limitations inherent in contacting patients only by telephone. As a result, the care managers increased the acuity ratings for many of the patients previously managed by San Diego care managers, revised the patients care plans accordingly, and contacted them more frequently. In some cases, the managers helped patients enroll in Medicaid or more fully use benefits that were already in place, such as blind pension funds and survivor benefits. Second, the standardization of the acuity assessments, which drove the contact frequency, and the supervisor s monitoring of contacts focused care managers efforts on the most acute cases. STREAMLINED AND STANDARDIZED CARE PLANS: When StatusOne left the program, Washington University had to replace its care planning software, which both sets of care managers had used. Care plans covered the following six domains: coordination of care, self-reliance, physical activity and fitness, community involvement, social supports, and mental fitness. According to program leaders, the software produced broad and undefined plans, making it hard for care managers to track and support patients progress on specific activities. Many of the plans were also subjective, leading to unwarranted variation in plans for similar patients. In response,washington University developed its own standard care plans to better focus on specific conditions and eliminate unwanted variability in interventions for them. These shorter, disease-specific care plans, which were reviewed and approved by Washington University Physician Network physicians, provided clear guidance on appropriate clinical interventions. The program developed general guidelines to modify the single-condition plans to apply to multiple comorbidities. The care managers followed the relevant guidelines based on each patient s specific conditions. Reduced Hospitalizations After Redesign In the 2.5 years after the redesign, the program reduced hospitalizations by 11.7 percent, or 0.16 hospital visits per patient per year (Exhibit 2), and total Medicare Part A and B spending per beneficiary by 9.6 percent, or $217 per month The program could be delivered even more effectively by focusing on the highest-risk patients. (Exhibit 3). These savings offset the care management fees that CMS paid to Washington University under the demonstration after the redesign, which averaged $151 per beneficiary per month. This average fee was less than the $167 before the redesign, because of the somewhat lower rate of participation in care management after the redesign. (Our intent-to-treat design included patients months after disenrollment in all analyses.) Because the 90 percent confidence interval for the estimated impact on expenditures including care management fees spans zero ( $250, $119), the program did not necessarily generate net savings for all enrollees. Nonetheless, it can be said that the program reduced hospitalizations without increasing total Medicare spending. Our results indicate that the program could be delivered even more effectively by focusing on the highest-risk patients. We expected the program which aimed to reduce the rate of acute exacerbations of patients conditions, which in turn leads to hospitalizations to have the strongest impacts on enrollees who were at highest risk of future hospitalizations. To assess this, we examined separately the 55 percent of enrollees who had had two or more hospitalizations during the two years before enrollment. Indicating their risk for future hospitalizations, usualcare group members in this subgroup had a hospitalization rate in the follow-up period that was 2.5 times higher than the rate for other members of the usual-care group. We found that the program effects for this higher-risk group not only were dramatic, but also drove the overall results. In the higher-risk group, the program reduced average hospitalizations per beneficiary per year by 17.0 percent, or 0.33 hospitalizations (Exhibit 2), and monthly Medicare Part A and B spending by 14.8 percent, or $435 (Exhibit 3). These savings were more than twice the size of the care management fees. As a result, the program reduced total spending by 9.7 percent, or $286 per patient per month. The 90 percent confidence interval for the aver Health Affairs JUNE :6

8 A care management provider will have a farbetterchanceof succeeding if it has a local presence. age monthly savings estimate is wide ($4, $567). The results were not sensitive to outliers, a fact that we assessed by also estimating impacts after trimming cases in the top 2 percent of expenditures to the ninety-eighth percentile. There were no effects on hospitalizations or spending for beneficiaries who did not meet this higher-risk definition. Discussion Alternative Explanations For Success In assessing the observed outcomes and the factors most likely to be responsible for them, we considered whether the stronger effects achieved after the redesign resulted from the redesign itself or from one or more of three following alternative explanations: longer average enrollment among participating beneficiaries; general learning by program staff members about effectively delivering the intervention; and the advent of Part D prescription drug coverage in 2006, which preceded the redesign by two months. We conclude that the first two alternatives did not explain the results. The program did not reduce hospitalizations or spending in beneficiaries first, second, or third years of enrollment if those years preceded redesign, but it did reduce both when those years followed redesign. This difference suggests that enrollees tenure was not the decisive factor. Regarding the second alternative, the clear lack of impacts on hospitalizations over the first four full years of the program suggests that program learning did not result in major improvements during that period. Furthermore, a prior program that Washington University had operated for two years provided administrators and several program nurses with a good deal of experience with care management. This history suggests that substantial program learning was unlikely to have occurred at the time of the redesign. 21 Additional studies would be needed to ascertain whether the introduction of Part D made the program more effective. However, the modest difference in the proportion of treatment- and control-group members who had Part D coverage at any point from 2006 through 2008 (four percentage points) is not large enough to account for the large effect on hospitalizations after redesign. Factors That Kept The Program Financially Sustainable Two factors enabled the program to remain financially viable after redesign, with no real change in the care management fee that CMS paid aside from a small increase for inflation despite the provision of a more intensive intervention. First, the program after redesign employed care manager assistants, overseen by the assigned care manager, to make three of four contacts for patients at the lower acuity levels of 3 5. These patients accounted for approximately 65 percent of the enrollees. The use of assistants enabled the care managers to provide greater attention to the patients at the higher acuity levels of 1 2 and to make all contacts with patients in the hospital. Second, according to the program director, the care management fee that CMS paid exceeded the cost to the program of delivering the intervention before the redesign. This was true both because the cost of delivering the intervention was somewhat lower than anticipated and because the fee at that time supported two organizations. Because the fee was not riskadjusted to account for patient severity, a higher average fee might be required if the program were to serve only higher-risk patients, who need more-intensive services. Even so, Medicare would still see a net savings. Implications For Health Reform Our results show that delivering the right care management program to the right people can work. Serving high-risk patients with an intervention that incorporates strong transitional care, medication management, systematic assessments, focused care plans, and the opportunity for inperson contacts by care managers with patients and providers reduced hospitalizations and Medicare spending. The program did not reduce hospitalizations or spending for the subgroup of higher-risk beneficiaries before the redesign, but it did so afterward. This fact confirms that both the program s features and the patients it served were responsible for its success. The current study also indicates that success is possible without explicitly embedding the chronic care managers into primary care an arrangement found in approaches such as the medical home model as long as the care managers can interact at least minimally with the patient s providers, including primary and spe- JUNE :6 Health Affairs 1223

9 cialty care providers, hospitals, and home health agencies. Although embedding care managers may improve outcomes, 22 most primary care physicians are in one- or two-physician practices, 23 which makes it difficult to hire, train, and supervise a care manager dedicated to assisting high-risk patients. In these cases, housing a program specializing in chronic care management within the same health care organization as the physician s practice or in the community can be a viable alternative. Accountable care organizations might wish to consider having care management provided by an entity within the accountable care organization but not necessarily embedded in the primary care practice. This study suggests, however, that the care management provider will have a far better chance of succeeding if it has a local presence, permitting not just telephone contacts but in-person visits to the hospital, the patient s providers offices, or the patient s home as needed. 24,25 The study also suggests the importance of having access to timely information when patients are admitted to the hospital, so that care managers can provide transitional care, and using a team-based approach that includes care assistants who maintain contact with lower-risk patients, allowing nurses to focus their efforts on the highest-risk patients. Finally, streamlining and standardizing care plans for patients with similar needs is important for focusing care managers attention on the most critical needs and reducing unwarranted variations across care managers. The results of the current study also have implications for future CMS demonstrations. They confirm that evaluations that move beyond solely focusing on whether programs work and that ask why and how they work can generate useful information. 26 Conclusion The lessons learned here about the importance of targeting and about successful intervention approaches emerged for two reasons. First, the implementers had time to refine their model. Second, the evaluation examined qualitative data about program implementation as well as quantitative effects over time and on different groups of patients. Still more powerful would be demonstrations that have the resources to provide rapid-cycle feedback on hospitalizations the primary driver of Medicare costs to implementers so that they could continuously refine their interventions. 26 The Center for Medicare and Medicaid Innovation is building these capacities, but the process will take time and money. The program s results after redesign, especially for the highest-risk enrollees, are promising. There is nothing unique about this care management intervention, such as a sophisticated electronic health record system or specially trained primary care providers, to suggest that it would not be similarly successful in other health care markets. To confirm this expectation, we recommend that the model be incorporated into protocols, replicated, and tested in multiple settings. The study was presented at the AcademyHealth Annual Research Meeting in Boston, Massachusetts, June 27, The authors gratefully acknowledge the support of the Robert Wood Johnson Foundation s Health Care Financing and Organization initiative and themedicarechroniccarepractice Research Network, which was funded by a grant from the Centers for Medicare and Medicaid Services (CMS). The authors also thank CMS for allowing them to reuse the data for this work, and Washington University School of Medicine in St. Louis for support to prepare the manuscript. NOTES 1 Thorpe KE, Ogden LL, Galactionova K. Chronic conditions account for rise in Medicare spending from 1987 to Health Aff (Millwood). 2010;29(4): Patient-Centered Primary Care Collaborative. Joint principles of a patient-centered medical home [Internet]. Washington (DC): The Collaborative; 2007 Feb [cited 2012 Apr 25]. Available from: McClellan M, McKethan AN, Lewis JL, Roski J, Fisher ES. A national strategy to put accountable care into practice. Health Aff (Millwood). 2010;29(5): Meyers D, Peikes D, Peterson G, Taylor EF, Lake T, Smith K, et al. The roles of patient-centered medical homes and accountable care organizations in coordinating patient care. Rockville (MD): Agency for Healthcare Research and Quality; 2010 Dec. (Report No. 11-M005-EF). 5 Bodenheimer T, Berry-Millett R. Care management of complex patients with complex health care needs. Princeton (NJ): Robert Wood Johnson Foundation; (Research Synthesis Report No. 19). 6 Peikes D, Chen A, Schore J, Brown R. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries: 15 randomized trials. JAMA. 2009;301(6): Coleman EA, Parry C, Chalmers S, Min SJ. The care transitions intervention: results of a randomized controlled trial. Arch Intern Med. 2006;166(17): Counsell SR, Callahan CM, Clark DO, Tu W, Buttar AB, Stump TE, et al. Geriatric care management for low-income seniors: a randomized controlled trial. JAMA. 2007; 298(22): Counsell SR, Callahan CM, Tu W, Stump TE, Arling GW. Cost analysis of the Geriatric Resources for Assessment and Care of Elders care management intervention. J Am Geriatr Soc. 2009;57(8): Dorr DA, Wilcox AB, Brunker CP, Burdon RE, Donnelly SM. The effect of technology-supported, multi Health Affairs JUNE :6

10 disease care management on the mortality and hospitalization of seniors. J Am Geriatr Soc. 2008; 56(12): Naylor MD, Brooten DA, Campbell RL, Maislin G, McCauley KM, Schwartz JS. Transitional care of older adults hospitalized with heart failure: a randomized, controlled trial. J Am Geriatr Soc. 2004;52(5): Brown R, Peikes D, Peterson G, Schore J, Razafindrakoto C. Six features of Medicare Coordinated Care Demonstration programs that cut hospital admissions of high-risk patients. Health Aff (Millwood). 2012;31(6): McCall N, Cromwell J, Urato C. (RTI International, Research Triangle Park, NC). Evaluation of Medicare Care Management for High-Cost Beneficiaries (CMHCB) demonstration: Massachusetts General Hospital and Massachusetts General Physicians Organization (MGH). Final report. Baltimore (MD): Centers for Medicare and Medicaid Services; 2010 Feb. (RTI Project No ). 14 Jack BW, Chetty VK, Anthony D, Greenwald JL, Sanchez GM, Johnson AE, et al. A reengineered hospital discharge program to decrease rehospitalization: a randomized trial. Ann Intern Med. 2009; 150(3): Gibson TB, Wang S, Kelly E, Brown C, Turner C, Frech-Tamas F, et al. A value-based insurance design program at a large company boosted medication adherence for employees with chronic illnesses. Health Aff (Millwood). 2011;30(1): Roebuck MC, Liberman JN, Gemmill-Toyama M, Brennan TA. Medication adherence leads to lower health care use and costs despite increased drug spending. Health Aff (Millwood). 2011;30(1): Schore J, Peikes D, Peterson G, Gerolamo A, Brown R. (Mathematica Policy Research, Princeton, NJ). Fourth report to Congress on the evaluation of the Medicare Coordinated Care Demonstration. Baltimore (MD): Centers for Medicare and Medicaid Services; 2011 Mar. (Contract No I [0012]). 18 Kroenke K, Spitzer RL, Williams JB. The Patient Health Questionnaire 2: validity of a two-item depression screener. Med Care. 2003;41(11): Brown R, Aliotta S, Archibald N, Chen A, Peikes D, Schore J. Research design for the evaluation of the Medicare Coordinated Care Demonstration. Princeton (NJ): Mathematica Policy Research; 2001 Feb To access the Appendix, click on the Appendix link in the box to the right of the article online. 21 Lynch JP, Forman SA, Graff S, Gunby MC. High-risk population health management achieving improved patient outcomes and near-term financial results. Am J Manag Care. 2000;6(7): Peikes D, Zutshi A, Genevro JL, Parchman ML, Meyers DS. Early evaluations of the medical home: building on a promising start. Am J Manag Care. 2012;18(2): Liebhaber A, Grossman J. Physicians moving to mid-sized, singlespecialty practice: results from the community-tracking survey. Washington (DC): Center for Studying Health System Change; (Issue Brief No. 18). 24 Cromwell J, McCall N, Burton J. Evaluation of Medicare Health Support chronic disease pilot program. Health Care Financ Rev. 2008; 30(1): Chen A, Brown R, Esposito D, Schore J, Shapiro R. Report to Congress on the evaluation of Medicare disease management programs. Princeton (NJ): Mathematica Policy Research; Foote SM. Next steps: how can Medicare accelerate the pace of improving chronic care? Health Aff (Millwood). 2009;28(1): ABOUT THE AUTHORS: DEBORAH PEIKES, GREG PETERSON, RANDALL S. BROWN, SANDY GRAFF & JOHN P. LYNCH Deborah Peikes is a senior researcher at Mathematica Policy Research. In this month s Health Affairs, Deborah Peikes and coauthors discuss how the Washington University School of Medicine in St. Louis, one of the fifteen participants in the Medicare Coordinated Care Demonstration, fine-tuned its initial unsuccessful efforts to reduce avoidable hospitalizations of Medicare beneficiaries. A complete redesign of its processes and management did eventually reduce hospitalizations and netted substantial cost savings. As in pie baking, a successful outcome for a care management program depends on which ingredients are included, how much of each is added, and how they are combined, says Peikes, a senior researcher at Mathematica Policy Research. Key to the university s program,shesays, were more comprehensive medication management; more inperson contacts to build rapport with patients, while still mainly using phone contacts; more thorough assessments of unmet needs; and streamlined and standardized care plans. Peikes has expertise in evaluating care coordination interventions for patients with chronic illnesses and using the patient-centered medical home model to improve primary care delivery. She led the analysis of the Medicare Coordinated Care Demonstration and is currently the project director and a principal investigator of the Centers for Medicare and Medicaid Services (CMS) evaluation of one of the demonstration programs. Peikes received a master s degreeinpublic affairs, with a concentration in economic policy, and a doctorate in public policy, both from Princeton University. JUNE :6 Health Affairs 1225

11 Greg Peterson is a health researcher at Mathematica Policy Research. Greg Peterson is a health researcher at Mathematica Policy Research, specializing in payment reform, chronic care management, long-term care, and quality of care for Medicare and Medicaid beneficiaries. He is a co principal investigator for the ongoing evaluation of the Medicare Coordinated Care Demonstration, and he is heading a study for the Center of Excellence for Research on Disability on the impacts of waiting periods for home and community-based services through Medicaid. Peterson has a master s degree in public affairs from Princeton and is pursuing a doctorate at the George Washington University. Randall S. Brown is avicepresident and director of health research at Mathematica Policy Research. Randall Brown is a vice president and director of health research at Mathematica Policy Research. He specializes in studies of care coordination and long-term care andinevaluationdesign.brownis principal investigator for the Money Follows the Person longterm care support program evaluation and directs the federally funded Center of Excellence for Disability Research. He recently led two evaluations of care coordination demonstrations funded by CMS. Brown currently serves on the AcademyHealth Methods Council. He holds a doctorate in economics from the University of Wisconsin. Sandy Graff is director of care management in the Department of Internal Medicine at the Washington University School of Medicine in St. Louis. Sandy Graff is the director of care management in the Department of Internal Medicine at the Washington University School of Medicine in St. Louis. She is a registered nurse and certified case manager with twenty years of experience in community-based complex case management for both payers and providers. She developed and implemented several care coordination programs for the School of Medicine, including the Medicare Coordinated Care Demonstration in St. Louis. Graff received her nursing degree from the Barnes Hospital School of Nursing. John P. Lynch is vice president and chief medical officer at Barnes- Jewish Hospital. John Lynch is vice president and chief medical officer at Barnes- Jewish Hospital and a professor of medicine at the Washington University School of Medicine in St. Louis. Lynch was the principal investigator for the Medicare Coordinated Care Demonstration in St. Louis. He has a medical degree from Georgetown University Health Affairs JUNE :6