Results of the Medicare Health Support Disease-Management Pilot Program

T h e n e w e ngl a nd j o u r na l o f m e dic i n e Special article Results of the Medicare Health Support Disease-Management Pilot Program Nancy McCall, Sc.D., and Jerry Cromwell, Ph.D. A bs tr ac t From RTI International, Washington, DC (N.M.), and Waltham, MA (J.C.); and the University of Massachusetts College of Nursing and Health Sciences, Boston (J.C.). Address reprint requests to Dr. McCall at RTI International, 701 13th St. NW, Suite 750, Washington, DC 20005, or at nmccall@rti.org. N Engl J Med 2011;365:1704-12. Copyright 2011 Massachusetts Medical Society. Background In the Medicare Modernization Act of 2003, Congress required the Centers for Medicare and Medicaid Services to test the commercial disease-management model in the Medicare fee-for-service program. Methods The Medicare Health Support Pilot Program was a large, randomized study of eight commercial programs for disease management that used nurse-based call centers. We randomly assigned patients with heart failure, diabetes, or both to the intervention or to usual care (control) and compared them with the use of a difference-indifferences method to evaluate the effects of the commercial programs on the quality of clinical care, acute care utilization, and Medicare expenditures for Medicare fee-for-service beneficiaries. Results The study included 242,417 patients (163,107 in the intervention group and 79,310 in the control group). The eight commercial disease-management programs did not reduce hospital admissions or emergency room visits, as compared with usual care. We observed only 14 significant improvements in process-of-care measures out of 40 comparisons. These modest improvements came at substantial cost to the Medicare program in fees paid to the disease-management companies ($400 million), with no demonstrable savings in Medicare expenditures. Conclusions In this large study, commercial disease-management programs using nurse-based call centers achieved only modest improvements in quality-of-care measures, with no demonstrable reduction in the utilization of acute care or the costs of care. 1704

Management of care for chronic illness, with its focus on high-risk, highcost patients, is touted as a valuable initiative to improve health outcomes while potentially saving Medicare billions of dollars. Medicare fee-for-service beneficiaries must navigate a health care system structured and financed to manage their acute, rather than long-term, health problems. 1,2 When Medicare beneficiaries seek medical care, their health problems are typically treated in costly and uncoordinated settings rather than managed in a holistic fashion. 3,4 As a result, nearly all growth in Medicare spending between 1987 and 2002 can be accounted for by beneficiaries with five or more medical conditions. 5 The concept of managing care for chronic illness has been evolving for years with the emergence of competing models. As described by Boden heimer, 6 the practice-based model emerged from the initiative Improving Chronic Illness Care. This model focuses on improving the care of chronic illness by embedding disease-management programs in physician practices and modifying clinical practice to inform patients about the nature of the care they need and to involve them in processes of their own care. 7-13 In contrast, the commercially based disease-management model relies much less on physicians to provide support for the management of chronic care. Historically, commercial disease-management programs were marketed to commercial health plans and employers as tools to achieve medical cost savings and to reduce absenteeism from work, with improvement in the care of chronic illness being a secondary goal. The commercial disease-management model uses teams of health coaches, usually in remote call centers, to enhance beneficiaries knowledge of and skills in provision of self-care and to coordinate care across providers. To date, the published literature on the effectiveness of commercial disease-management programs has shown both positive and negative returns on investment. 14 Although a number of studies have shown gains in the quality of care processes, 13,15,16 serious concerns with study design and the methods used to assess outcomes raise questions about the ability of disease-management programs to effect meaningful population-based improvements in health status and to lower medical expenditures. 14-22 In the early years of this century, the Centers for Medicare and Medicaid Services (CMS) funded the first major Medicare study of the effectiveness of interventions intended to enhance the coordination of care. 23 None of the 15 interventions evaluated significantly reduced hospitalization rates or saved the Medicare program money, but small samples limited detectable savings to rates of 20% or more. 23 In 2005, the CMS launched the Medicare Health Support Pilot Program to test the effectiveness of the commercial disease-management model, which relies primarily on remote nursebased call centers for beneficiaries with heart failure or diabetes. This article examines the question of whether commercial disease-management companies were able to achieve meaningful savings for the Medicare program while improving the quality of care for beneficiaries and reducing acute care utilization. Me thods Study Design The eight companies launched their programs between August 1, 2005, and January 16, 2006. The programs used a model involving remote nursebased call centers to assess the needs of individual beneficiaries and health coaches to target beneficiaries at immediate high risk for adverse events. In addition to achieving cost savings, the goals of the intervention were to improve beneficiaries understanding of their disease or diseases, their ability to manage self-care, and their ability to communicate with providers. Various educational resources including literature, videos, and Internet resources were provided. A small portion of the intervention population received intensive casemanagement services. All companies communicated with patients medical providers, but datasharing arrangements between the companies and physicians were limited. Over time, all companies modified the outreach and intervention programs that had been developed for private employers and Medicare managed-care plans in response to the challenges of working with the chronically ill population served by Medicare on a fee-for-service basis. To cover the cost of intervention activities, the CMS paid each company a monthly administrative fee per participant that was negotiated between 1705

T h e n e w e ngl a nd j o u r na l o f m e dic i n e the company and the CMS. To retain all its fees, a company had to meet preestablished targets for clinical quality and beneficiary satisfaction and to reduce Medicare expenditures in the intervention group by at least the amount of the fee (thereby achieving budget neutrality). Faced with substantial financial liability, five companies requested early termination of their programs. RTI International won a competitive CMS solicitation to conduct the independent evaluation and was solely responsible for the design and conduct of the analyses, creation of the analytic data sets, interpretation of the data, and preparation of the manuscript. The two authors conceived of and wrote the article, made the decision to submit it for publication, and vouch for the data and analyses. Study Sample The CMS used a population-based strategy to identify roughly 30,000 Medicare beneficiaries in each of eight geographic areas (240,000 altogether) who met the selection criteria for heart failure or diabetes and had a Hierarchical Condition Category (HCC) risk score of 1.35, indicating that their feefor-service cost was at least 35% greater than the average. An HCC score is a cost-weighted score for level of risk that is based on a regression of annual total Medicare beneficiary costs on more than 100 patient characteristics, including age, sex, and 70 categories of coexisting conditions. Beneficiaries were then randomly assigned to the intervention or control group in a ratio of 2:1, within chronic disease, Medicaid eligibility, and HCC risk score strata. After randomization, the CMS notified the beneficiaries assigned to the intervention group that they were eligible to participate in the Medicare Health Support Pilot Program. The companies then conducted recruitment calls to obtain beneficiaries oral consent. Beneficiaries were eligible to participate in the pilot program and were included in our analyses until they died, no longer had Medicare as their primary payer or lost eligibility for Part A and Part B, or enrolled in a Medicare program for end-stage renal disease or hospice care or in a Medicare Advantage plan. The beneficiaries who orally agreed to participate constituted the intervention group. Beneficiary contact rates with Medicare Health Support staff were calculated only for those beneficiaries who were fully eligible and participating after an initial 6-month recruitment period. The number and months of contact, without adjustment for risk, were standardized according to the number of months the company participated in the pilot program. Data Sources Claims filed under Medicare Parts A and B were assembled for 12 months before the start date for each company s intervention (to establish a baseline period) and for the next 36 months or less, if the company terminated participation early. Information on enrollment in Medicare during the baseline period, demographic characteristics, and date of death was obtained from the Medicare enrollment database. The number of days of eligibility during the intervention period was obtained from Medicare daily eligibility files. Claims, enrollment, and eligibility information was assembled for all beneficiaries who underwent random assignment to the intervention and control groups. The frequency of contact (by telephone and in person) was recorded by each of the eight participant companies. Statistical Analysis The Medicare Health Support Pilot Program followed an intention-to-treat, difference-in-differences study design, with changes from baseline compared between the intervention and control groups with regard to the quality of clinical care provided, the utilization of acute care, and Medicare expenditures. Beneficiaries had to be eligible for Medicare Part A or Part B for at least 1 day during both the baseline and intervention periods. The individual beneficiary was the unit of analysis. Observation weights were used to account for the stratified sampling design and each beneficiary s fraction of eligible days during the intervention period. Medicare claims were included only if the starting date of the service occurred when the beneficiary was eligible in the baseline and intervention periods. We analyzed four annual evidence-based, process-of-care measures for diabetes: glycated hemoglobin testing, urinary protein screening, retinal eye examination, and low-density lipoprotein (LDL) cholesterol testing. Only the LDL cholesterol test was used to evaluate beneficiaries with heart failure. Rates per 100 beneficiaries were compared between the intervention and control groups for the 12-month baseline period and for months 7 through 18 and the last 12 months of each company s intervention period. We omitted the first 1706

6 months of company participation from the analysis, since the primary focus during that period was outreach and recruitment. The effects of the intervention were estimated by means of a twoperiod, beneficiary-level, logistic-regression model with robust variance estimation (Stata software, version 8). Dichotomous variables, with interactions, were included for group assignment (intervention vs. control group) and study period (baseline vs. intervention period). Rates of hospitalization and emergency room visits per 1000 beneficiaries were also compared for the intervention and control groups during the baseline period and the same two intervention periods used in the comparison of processes of care (i.e., months 7 through 18 and the final 12 months). We compared rates of utilization for any condition and for a set of 10 ambulatory care sensitive conditions (conditions for which utilization of acute care facilities may be avoided if good primary care is provided through ambulatory services): heart failure, diabetes, asthma, cellulitis, chronic obstructive pulmonary disease or chronic bronchitis, dehydration, bacterial pneumonia, septicemia, ischemic stroke, and urinary tract infection. effects were estimated with the use of a two-period, beneficiary-level, weighted, negative binomial generalized linear model with robust variance estimation and similar parameters for the intervention effect, as described above. For both sets of analyses, we report comparisons between baseline and the last 12 months of the intervention to assess the performance of mature programs; the results did not differ materially from analyses that compared findings from the baseline period with those from months 7 through 18. Analysis of Medicare costs included all payments for Part A and Part B (excluding patient and third-party obligations). Part D drug costs were excluded. No maximum spending threshold was applied to capture possible intervention savings on high-cost beneficiaries. Average costs per beneficiary per month were constructed at the beneficiary level in the baseline and intervention periods by dividing total Medicare payments by the number of months that the beneficiary was eligible. Calculations of gross savings were based on mean differences in the changes from baseline in the cost for individual beneficiaries. Z statistics for the differences in growth rates of Medicare costs between the intervention and control groups were used to test for intervention effects at the 95% confidence level. Weighted analysis of covariance regressions of intervention-period costs per beneficiary per month on the beneficiary s baselineperiod costs and characteristics (e.g., HCC risk score, age, and sex) produced a gain in statistical accuracy of 10 to 15%, allowing us to detect savings as small as 3.0 to 4.5% of control-group costs, depending on the company. Net program savings were defined as average monthly gross savings minus fees paid to the company. R esult s Patients This study included 242,417 patients who were randomly assigned to receive disease-management services (163,107 patients) or usual care (79,310 patients). Because of space limitations, we have excluded the results of analyses conducted for the 47,000 beneficiaries assigned to receive usual care or to receive disease-management services from seven of the eight participating companies 1 year after the pilot program began. The results of these analyses were materially the same as the results of the analyses reported here. The characteristics of the beneficiaries were well balanced between the intervention and control groups at baseline (Table 1). Patients in both groups were quite sick, were receiving costly care, and had a substantial level of comorbidity, with an average of more than one hospitalization annually and, in 2004, an average of $15,000 in Medicare expenditures. Beneficiary participation rates ranged between 75 and 96%, with an average rate of 85% across the eight company programs (Table 2). Because the study had an intention-to-treat design, lower participation rates reduced the likelihood of success across all beneficiaries in the overall intervention group. For beneficiaries who participated continuously after the recruitment period until death, the average number of contacts with a Medicare Health Support company was less than 1 per month (0.7) and ranged across companies from 0.4 to 1.0 per month. On average, the companies contacted participants every 2.7 months, with 80 days between contacts. Gaps in months between contacts were greater than indicated by average monthly contacts because some beneficiaries were contacted two or more times in a single month. On average, during the course of a 1707

T h e n e w e ngl a nd j o u r na l o f m e dic i n e Table 1. Baseline Characteristics of the Study Population.* Characteristic Aetna Healthways CIGNA Health Support (N = 20,259) (N = 10,118) (N = 20,031) (N = 10,016) (N = 20,361) (N = 10,146) Primary condition (%) Heart failure only 26 25 20 20 19 19 Diabetes only 51 51 60 60 60 60 Heart failure and diabetes 23 24 20 20 21 21 Mean HCC score 2.5 2.5 2.5 2.5 2.4 2.4 Enrolled in Medicaid (%) 16 16 16 15 25 25 Other chronic condition (%) Coronary artery disease 47 46 42 42 40 40 Respiratory disease 23 24 22 22 24 24 Renal disease 17 15 15 16 17 16 Cardiac dysrhythmia or conduction disorder 33 32 28 28 26 28 Prior hospitalization for any reason (rate/100 104 101 86 85 79 75 beneficiaries) Prior emergency room visit for any reason (rate/100 51 50 59 59 87 84 beneficiaries) Prior total Medicare payments per beneficiary per month ($) 1,534 1,503 1,397 1,413 1,198 1,127 * There were no significant differences between groups except as noted below. A Hierarchical Condition Category (HCC) risk score of 1.0 represents the predicted expenditures for average Medicare fee-for-service beneficiaries. A mean HCC score of 2.5 indicates that expenditures for this group of beneficiaries are expected to be 150% higher. P<0.01 for the comparison with the control group. P<0.05 for the comparison with the control group. P<0.001 for the comparison with the control group. 30-month period, 59% of participating beneficiaries were contacted at least 10 times, and 23% were contacted fewer than 5 times. Processes of Care Baseline rates were generally quite high for all processes of care except retinal eye examination. Of the 40 evidence-based, process-of-care measures, 14 differed significantly between the intervention and control groups, with trends favoring the disease-management intervention over usual care, although only 7 of these differences represented absolute improvements in quality (Table 3; and Table 1 in the Supplementary Appendix, available with the full text of this article at NEJM.org). However, the absolute percentage-point changes in the rates were small, ranging from 1.3 to 3.1 per 100 beneficiaries. During the pilot period, the control groups generally had larger rate declines than the intervention groups, suggesting a lower quality of care. One company had significant improvement in performance on four of five process-of-care measures. Four companies had significant improvement with regard to one process of care. Utilization of Care Medicare Health Support companies had little success in slowing the rates of growth in hospital admissions and emergency room visits for any condition and for ambulatory care sensitive conditions (Table 4, and Table 2 in the Supplementary Appendix). Only one company was successful in slowing the rate of growth in hospitalizations for any condition ( 44 per 1000 beneficiaries, P = 0.04). Another company slowed the rate of growth in hospitalizations for ambulatory care sensitive conditions ( 5 per 1000 beneficiaries, P = 0.05). Cost Savings All eight companies sustained substantial increases in average monthly costs per beneficiary in both groups (Table 5). Three of the eight companies had 1708

Health Dialog Green Ribbon Health LifeMasters McKesson XLHealth (N = 20,039) (N = 8,018) (N = 22,605) (N = 11,316) (N = 20,120) (N = 10,078) (N = 20,174) (N = 10,107) (N = 19,518) (N = 9,511) 22 22 24 25 26 26 21 21 20 20 55 55 55 55 49 49 54 55 58 57 23 23 21 20 25 25 25 24 22 23 2.6 2.6 2.5 2.5 2.4 2.4 2.3 2.3 2.4 2.4 17 16 14 14 21 21 43 42 34 33 53 54 49 48 47 46 38 38 40 41 26 26 25 25 27 27 23 23 26 27 16 17 15 15 13 14 14 14 16 16 33 33 34 34 27 27 23 23 25 26 98 95 73 72 91 90 90 88 76 75 69 65 55 55 78 76 103 101 84 87 1,330 1,290 1,231 1,214 1,292 1,296 1,241 1,216 1,153 1,138 smaller increases in costs in the intervention group than in the control group, but none of the differences were significant. Given the large samples, the study had the ability to detect savings as small as $50 to $60, or 3 to 4% of average costs per beneficiary per month in the control groups (1.96 times the standard error for average gross savings). At best, the observed gross savings were 1.24% of the control group s cost per beneficiary per month for one company, which was not significantly different from zero. Analysis-of-covariance regressions, with adjustment for beneficiary characteristics, produced similar null findings for gross savings (data not shown). Negotiated monthly fees per beneficiary ranged from $74 to $159, or 4.9 to 11.0% of the cost per beneficiary per month for the control groups. The range of monthly fees was equivalent to what Medicare would pay a physician for two or three office visits involving a moderate level of complexity in clinical decision making. Medicare management fees, net of any gross savings, ranged from 3.8 to 10.9% per beneficiary per month. To achieve budget neutrality and cover all fees received, monthly net fees had to be negative, not positive. Discussion We found that the eight commercial disease-management companies in this large, randomized study had limited success in improving the processes of care studied and had no consistent success in reducing hospital admissions or emergency room visits in a population of fee-for-service Medicare beneficiaries. There were no significant savings in Medicare expenditures on health services. Our results are consistent with previous findings on the use of disease-management interventions in the Medicare fee-for-service population 24 but are not entirely consistent with reported findings for commercially insured or managed-care populations. 14-22,25 We suggest five possible explanations for the lack of success with the Medicare fee-for-service population. First, success in the private sector, as discussed above, usually has not been based on rigorous, population-based, experimental designs and evaluation methods. It is likely that success has been due in part to the act of simply following a targeted group of initially sick workers over time as their health eventually improved. 1709

T h e n e w e ngl a nd j o u r na l o f m e dic i n e Table 2. Rates of Program Participation and Rates of Contact with Health Coaches among Fully Eligible and Participating Beneficiaries in the Group. Company Overall Participation Rate (%)* Mean No. of Contacts per Active Mo Mean No. of Mo between Contacts Mean No. of Days between Contacts 10 Contacts in 30-Mo Period (%) <5 Contacts in 30-Mo Period (%) Aetna 84 0.5 3.3 100 50 22 Healthways 90 0.8 2.1 64 71 12 CIGNA Health Support 89 1.0 1.8 55 75 12 Health Dialog 96 0.8 2.1 64 82 8 Green Ribbon Health 86 0.7 2.5 74 75 12 LifeMasters 76 0.9 2.2 66 39 53 McKesson 82 0.4 3.8 113 20 47 XLHealth 75 0.5 3.4 103 62 14 Average across companies 85 0.7 2.7 80 59 23 * The overall participation rate applies to eligible beneficiaries in the intervention group. The mean number of contacts per active month is the ratio of the mean number of completed telephone and in-person contacts per month to active intervention months, excluding the first 6-month recruiting period. The mean time between contacts exceeded mean contact rates for beneficiaries with multiple contacts in a single month. The mean number of days between contacts was calculated by multiplying the number of months between contacts by 30. Not all companies participated in the Medicare Health Support program for 30 months. Table 3. Differences in Rates of Growth for Processes of Care between the Last 12 Months of the Medicare Health Support Pilot Program and a 1-Year Baseline Period.* Company Difference in Growth Rates per 100 Beneficiaries Diabetes Heart Failure Glycated Hemoglobin Urinary Protein Retinal Eye Examination LDL Cholesterol LDL Cholesterol Aetna 1.6 0.1 1.3 1.0 1.1 Healthways 2.4 0.3 0.9 2.1 1.4 CIGNA Health Support 1.3 3.1 0.1 1.2 1.2 Health Dialog 0.1 0.1 0.3 0.5 3.1 Green Ribbon Health 1.5 1.3 0.3 0.3 1.4 LifeMasters 0.9 1.7 2.1 2.4 2.7 McKesson 1.3 0 0.8 2.9 0.1 XLHealth 0.6 1.7 2.7 0.5 1.7 * There were no significant differences between groups except as noted below. LDL denotes low-density lipoprotein. The difference in rates of growth is equal to the rate of change in the intervention group minus the rate of change in the control group for each company. P<0.05 for the comparison between the intervention and control groups. P<0.01 for the comparison between the intervention and control groups. Second, the care of elderly, chronically ill patients is difficult to manage. They are much more likely than younger persons to have new acute conditions such as stroke, pneumonia, and hip fracture. 26 Although each of the companies in our study intended to manage the care of the whole person, the health coaches were surprised by the number of health and psychosocial problems that were prevalent among Medicare fee-for-service beneficiaries. 5 Third, because of the sheer size of the pilot intervention populations with multiple coexisting 1710

conditions combined with the requirement for budget neutrality from the CMS, the companies adopted a strategy of targeting high-cost beneficiaries at greatest risk for an adverse event. Any model predicting future high utilization and costs will not perform well because of the substantial regression to the mean as the healthier, lower-cost beneficiaries have disease flare-ups and the sicker, higher-cost beneficiaries recover and subsequently require less care at least for a while. 26 Fourth, the unpredictable nature and immediacy of chronic disease flare-ups call for real-time information on health status. Often, beneficiaries were identified for intensive case management weeks or months after they had been discharged from the hospital, when the health coaches made a prescheduled, routine call to beneficiaries, or after analyses of Medicare claims. At that point, major inpatient costs had already been incurred along with the services provided after the need for acute care had ended. Fifth, the health coaches were not integrated into the beneficiary s primary health care team. This hindered their ability to interact directly with the beneficiary s primary care provider and facili- Table 4. Differences in Rates of Growth for Acute Care Utilization per 1000 Beneficiaries between the Last 12 Months of the Medicare Health Support Pilot Program and a 1-Year Baseline Period.* Company Hospitalizations Emergency Room Visits For Ambulatory For Any Care Sensitive Condition Conditions For Any Condition For Ambulatory Care Sensitive Conditions Aetna 44 10 13 2 Healthways 29 15 13 2 CIGNA Health Support 27 3 12 6 Health Dialog 6 22 4 7 Green Ribbon Health 12 5 15 5 LifeMasters 21 5 62 5 McKesson 18 13 43 9 XLHealth 20 27 22 16 * The difference in rates of growth is equal to the rate of change in the intervention group minus the rate of change in the control group for each company. There were no significant differences between groups except as noted below. The conditions defined as ambulatory care sensitive (in which utilization of acute care services may be prevented if good primary care is provided through ambulatory services) were heart failure, diabetes, asthma, cellulitis, chronic obstructive pulmonary disease or chronic bronchitis, dehydration, bacterial pneumonia, septicemia, ischemic stroke, and urinary tract infection. P<0.05 for the comparison between the intervention and control groups. Table 5. Overall Change in Monthly Gross Savings, Net Management Fees, and Net Fees.* Company Change in PBPM Cost Gross Savings Monthly Fee Mean % of PBPM Cost Initial Fee % of PBPM Cost** Net Fee $ $ $ % Aetna 444 466 22±31 1.24 137.36 7.8 6.6 Healthways 448 441 7±33 0.40 105.38 6.3 6.7 CIGNA Health Support 281 278 3±25 0.23 115.60 8.9 9.1 Health Dialog 364 360 4±29 0.30 120.15 8.1 8.4 Green Ribbon Health 398 414 16±23 1.07 73.96 4.9 3.8 LifeMasters 260 222 38±28 2.67 92.77 6.5 9.2 McKesson 264 255 9±26 0.65 113.82 8.3 9.0 XLHealth 399 401 2±25 0.14 158.97 11.0 10.9 * Plus minus values are means ±SE. The period of time covered is the 36 months of the Medicare Health Support Pilot Program (or the length of time each company participated) as compared with the 12-month baseline period. PBPM denotes per beneficiary per month. The change in costs is defined as the difference in mean costs between the baseline and intervention periods PBPM. The mean gross savings is defined as the mean change in PBPM costs for the intervention group minus the mean change in PBPM costs for the control group. The percentage of PBPM cost is equal to the mean gross savings as a percentage of PBPM costs for the control group during the intervention period. The monthly fee is the monthly management fee that the Centers for Medicare and Medicaid Services paid each company for each beneficiary who agreed to participate. The initial fee was the fee negotiated at the beginning of the intervention period. ** The percentage of PBPM cost is equal to the negotiated monthly fee per beneficiary as a percentage of PBPM costs for the control group during the intervention period. The net fee is equal to the monthly fee plus the change in gross savings as a percentage of PBPM costs for the control group. 1711

tate changes in medical care plans made by the primary care provider to mitigate deterioration in health status and avoid the need for acute care services. Although this was a randomized study, the companies involved had difficulties recruiting sicker beneficiaries into their intervention groups. The problem of recruiting candidates for intervention is inherent in any population-based strategy and may be less pronounced with a provider-based approach for clinicians who already have established relationships with patients. In light of the results of other CMS disease-management programs 24 and the unsuccessful interventions in this large, randomly designed study, it is unlikely that simply managing the care of elderly patients through telephone contact or an occasional visit will achieve the level of savings that Congress had hoped for when it mandated the Medicare Health Support Pilot Program. These findings also suggest that for such programs to be effective, they need to include intensive, costly, personal clinical attention. Disclosure forms provided by the authors are available with the full text of this article at NEJM.org. References 1. McGlynn EA, Asch SM, Adams J, et al. The quality of health care delivered to adults in the United States. N Engl J Med 2003;348:2635-45. 2. Jencks SF, Huff E, Cuerdon T. Change in the quality of care delivered to Medicare beneficiaries, 1998-1999 to 2000-2001. JAMA 2003;289:305-12. [Erratum, JAMA 2002;289:2649.] 3. Todd W, Nash T, eds. Disease management: a systems approach to improving patient outcomes. New York: Jossey- Bass, 2001. 4. Anderson G. Chronic conditions: making the case for ongoing care. Baltimore: Johns Hopkins University, December 2002. 5. Thorpe KE, Howard DH. The rise in spending among Medicare beneficiaries: the role of chronic disease prevalence and changes in treatment intensity. Health Aff (Millwood) 2006;25:w378-w388. 6. Bodenheimer T. s to improve chronic illness care: evaluating their effectiveness. Dis Manag 2003;6:63-71. 7. Wagner EH. Chronic disease management: what will it take to improve care for chronic illness? Eff Clin Pract 1998;1:2-4. 8. Institute of Medicine. Crossing the quality chasm. Washington, DC: National Academies Press, 2001. 9. Vladek BC. You can t get there from here: obstacles to improving care of the chronically ill. Health Aff (Millwood) 2001;20(6):175-9. 10. Foote SM. Population-based disease management under fee-for-service Medicare. Health Aff (Millwood) 2003;Suppl Web Exclusives:W-3-342 W-3-356. 11. Glasgow RE, Orleans CT, Wagner EH. Does the chronic care model serve also as a template for improving prevention? Milbank Q 2001;79:579-612. 12. Wagner E. The changing face of chronic disease care. In: Accelerating Change Today (A.C.T.) for America s Health. Curing the system: store of change in chronic illness care. Washington, DC: National Coalition on Health Care, Institute for Healthcare Improvement, 2002:2-5. 13. Wagner EH, Austin BT, Davis C, Hindmarsh M, Schaefer J, Bonomi A. Improving chronic illness care: translating evidence into action. Health Aff (Millwood) 2001;20(6):64-78. 14. Goetzel RZ, Ozminkowski RJ, Villagra VG, Duffy J. Return on investment in disease management: a review. Health Care Financ Rev 2005;26:1-19. 15. Rich MW, Beckham V, Wittenberg C, Leven CL, Freedland KE, Carney RM. A multidisciplinary intervention to prevent the readmission of elderly patients with congestive heart failure. N Engl J Med 1995;333:1190-5. 16. Knight K, Badamgarav E, Henning JM, et al. A systematic review of diabetes disease management programs. Am J Manag Care 2005;11:242-50. 17. Buntin MB, Jain AK, Mattke S, Lurie N. Who gets disease management? J Gen Intern Med 2009;24:649-55. 18. Duncan I, Lodh M, Berg GD, Mattingly D. Understanding patient risk and its impact on chronic and non-chronic member trends. Popul Health Manag 2008; 11:261-7. 19. Serxner S, Mattke S, Zakowski S, Gold D. Testing the DMAA s recommendations for disease management program evaluation. Popul Health Manag 2008;11:241-5. 20. Mays GP, Au M, Claxton G. Convergence and dissonance: evolution in privatesector approaches to disease management and care coordination. Health Aff (Millwood) 2007;26:1683-91. 21. Mattke S, Seid M, Ma S. Evidence for the effect of disease management: is $1 billion a year a good investment? Am J Manag Care 2007;13:670-6. 22. Motheral BR. Telephone-based disease management: why it does not save money. Am J Manag Care 2011;17(1):e10-e16. 23. An analysis of the literature on disease management. Washington, DC: Congressional Budget Office, October 13, 2004. 24. Peikes D, Brown R, Chen A, Schore J. Effects of care coordination on hospitalization, quality of care, and health care expenditures among Medicare beneficiaries: 15 randomized trials. JAMA 2009; 301:603-18. 25. RAND Corporation. Expert Panel meeting on disease management outcomes measurement. Washington, DC: The Office of Assistant Secretary for Planning and Evaluation, August 2008. 26. Wennberg DE, Marr A, Lang L, O Malley S, Bennett G. A randomized trial of a telephone care-management strategy. N Engl J Med 2010;363:1245-55. Copyright 2011 Massachusetts Medical Society. apply for jobs at the nejm careercenter Physicians registered at the NEJM CareerCenter can apply for jobs electronically. A personal account created when you register allows you to apply for positions, using your own cover letter and CV, and keep track of your job-application history. Visit NEJMjobs.org for more information. 1712