Disease management: promising, but not yet proven

Challenge of improving chronic care

One of the biggest long-term fiscal challenges for industrialized countries is controlling the cost of care for people with chronic illnesses, such as heart disease and diabetes. In 1998, for example, 14% of US Medicare beneficiaries, most of whom had multiple chronic illnesses that negatively affected their quality of life, accounted for more than three quarters of all Medicare payments. Coordinating the care of these patients is difficult, as US Medicare beneficiaries with chronic illnesses see many doctors – an average of 11 different physicians per year [1]. Despite the costs and complexity of providing effective chronic care, studies have shown that many acute health problems, and resulting monetary and social costs, can be prevented if: patients receive medical care that is consistent with recommended standards; patients adhere to recommended medication, diet, exercise and self-care regimens; patients have access to transportation and social support services; and providers communicate better with each other and with patients. This has led many health maintenance organizations and insurers to develop, or contract with, disease management (DM) programs to help inculcate these changes in patient and provider behavior. DM greatly expands the intensity and focus of managed care programs, which were widely tested in the early 1990s.

Disease management holds promise for improving chronic care

DM has been growing rapidly as a potential way to improve chronic care. DM is a “system of coordinated healthcare interventions and communications for populations with conditions in which patient self-care efforts are significant.” DM [101]:

•	Supports the physician or practitioner/patient relationship and plan of care
•	Emphasizes prevention of exacerbations and complications utilizing evidence-based practice guidelines and patient empowerment strategies
•	Evaluates clinical, humanistic and economic outcomes on an ongoing basis with the goal of improving overall health

Components of DM may include ways to identify eligible populations, evidence-based practice guidelines, patient self-management education, evaluation and reporting/feedback. DM is typically provided by nurses, sometimes with the assistance of social workers, nutritionists and pharmacists. However, some programs require more highly trained nurses (registered nurses [RNs] or advanced practice), while others use call center staff without a nursing degree for some patients and tasks.

The use of DM is burgeoning as insurers and employers seek ways to control costs and improve the quality of chronic care. The Disease Management Purchasing Consortium estimates that the revenue of DM organizations that provide outsourced services grew from US$78 million in 1997 to nearly 1.2 billion in 2000 and is projected to grow to 1.8 billion by 2008 [102]. Those figures do not count the DM programs run internally by payers and hospitals. Matheson and colleagues interviewed 120 of the 150 largest health insurance payers in the USA (measured by the number of commercially insured people). They found that 49% offered outsourced DM and 4% ran their own DM programs. While it is harder to obtain numbers on how many people actually use DM programs, a recent survey indicated that 50% of covered workers are enrolled in managed care plans that offer such programs [103]. In the public sector, many states now require their Medicaid programs to offer DM initiatives. Some states offer enrollees incentives to sign up for DM initiatives. South Dakota (USA) has gone so far as to penalize beneficiaries of the state’s high-risk health insurance pool if they do not participate in DM programs.

Implementation of disease management has outpaced evidence proving its worth

Implementation of DM has outpaced evidence that it is effective. Most studies have suffered from one of three flaws. First, many use a pre–post analysis to compare spending for an equal amount of time before and after patients are enrolled (e.g., the year before patients are enrolled with the year after). Most patients are identified as eligible after an acute event, such as a hospitalization, and the cost of that event is sometimes counted in the ‘preintervention’ costs. Not surprisingly, unusually high costs from the hospitalization return to normal in the post-period, in what statisticians refer to as ‘regression to the mean’. This phenomenon occurs even if the identifying stay is excluded from the calculations, because average costs are higher during the 6 months after a hospital stay than in month 7, or later after discharge. As a result, studies that do not use a comparison group to account for natural trends tend to show dramatic reductions in hospitalization and costs, even though costs would have likely declined in a group of similar patients that was identified at the same point and did not receive DM. The second common error is to compare only those who successfully complete or remain in the program with nonparticipants. People who volunteer for a program and complete it are often more likely to take an active role in managing their health and adopting the program’s suggested behavior changes. Furthermore, including only those who completed a program is not a fair way to measure cost–effectiveness if payers have paid for people who later drop out. To address these methodological weaknesses, the DM industry is currently trying to standardize ways to calculate the return on investment in DM. Finally, other studies use reliable methods, but enroll small samples that may not be indicative of results when offered to the larger populations that insurers typically serve. Some small pilot programs designed to improve patients’ adherence to treatment regimens and physicians’ adherence to professional guidelines have been found to be effective in improving patient outcomes and reducing costs, for exampe, see [2–7]. As a result, the Congressional Budget Office’s review of the literature finds evidence that some DM programs might improve the quality of care, but finds inconclusive evidence on whether the improvements translate into reductions in costs of care that are large enough to offset the costs of DM [104].

Disease management can mean many things

Insurers considering purchasing DM services tend to focus on whether DM works. The question they should be asking instead is what DM features are likely to lead to improved quality and lower costs, for what target populations, and over what time frame, as the exact features of DM vary widely.

Whether DM works depends on the features of the program and target population, and the time period being examined. The term ‘disease management’ is broad enough to encompass many approaches, none of which are exactly the same. In other words, if you’ve seen one DM program, you’ve seen one DM program. The target populations, philosophies, models, intensity of contacts, and even the providers of DM are incredibly diverse. For example, the 15 programs being evaluated using a randomized design in the Medicare Coordinated Care Demonstration (MCCD) are operated by a variety of organizations, including five commercial DM providers, three hospitals, three academic medical centers, an integrated healthcare delivery system, a hospice, a long-term care facility and a retirement community. The programs vary widely in the target populations they treat, the length, nature and intensity of the interventions they deliver (ranging from an average of less than one contact per patient per month to more than eight contacts) and the cost of the interventions. Finally, different programs might be more and less effective over different periods of time, depending on how long it might take to create behavioral changes and how long any changes are maintained.

To help link specific features to outcomes, in its evaluations of two Centers for Medicare & Medicaid Services (CMS) DM demonstrations, Mathematica Policy Research, Inc. developed a scoring algorithm that assessed each program’s performance on ten domains. These domains were staffing, initial assessment, care planning, physician practice, educating and encouraging better self-care by patients, efforts to improve communications and coordination across providers, service arrangement, quality of patients’ electronic medical records, ongoing monitoring, and quality management and outcome measurement. We also designed and collected information on the types of patients enrolled, how often contacts occurred and the purpose and content of those contacts, who initiated each contact, whether the contact was by telephone or in person, the frequency of contacts, reasons for patients dropping out, use of home monitoring devices, and other indicators reflecting the nature and intensity of the intervention. All these factors can affect the cost and effectiveness of a particular DM intervention.

Evidence on the horizon: Centers for Medicare & Medicaid Services is rigorously testing disease management

CMS is currently running a series of demonstrations that use randomized designs to expand the evidence base on whether and how DM interventions can improve care for Medicare beneficiaries with chronic illness in its fee-for-service program. While Medicare beneficiaries are, on average, sicker than people covered by employer/commercial plans, findings regarding DM effectiveness for the Medicare population are likely to provide important lessons for a younger population with chronic illness too. The DM providers who participated in two of the earliest demonstrations were not required to carry any financial risk for reducing traditional Medicare costs sufficient to offset program fees. Informatics for Diabetes Education and Telemedicine (IDEATel) began in 2000 and had enrolled 2164 people as of April 2006. The MCCD is evaluating DM provided by 15 different providers. The providers began enrollment in 2002 and have since enrolled 21,691 beneficiaries (as of April 15, 2006). Programs charge fees per beneficiary per month ranging from US$50 to 444.

CMS decided to require providers participating in later demonstrations to guarantee reductions in traditional Medicare costs large enough to offset program fees. The Medicare Disease Management Demonstration is testing DM provided on a larger scale by three DM providers. In return for monthly program fees per beneficiary, the programs must provide DM services and prescription drug coverage. Enrollment, which began in 2004, totaled 26,538 (18,993 of whom were in the treatment group) as of January 7, 2006.

These three ongoing demonstration programs – IDEATel, MCCD and the Medicare Disease Management Demonstration – will provide important information on the ability of different types of DM programs to produce favorable effects on cost and quality. However, the programs only serve people who willingly enroll for a random-assignment study. In subsequent demonstrations, CMS decided to answer questions about the effectiveness of practical, large-scale models where the DM provider takes responsibility for entire populations of patients with particular conditions. This population-based approach eliminates concerns about incentives for favorably selecting (‘cherry picking’) beneficiaries. It also provides incentives for DM providers to encourage all patients with the target condition who might benefit to participate, and to encourage usual care providers to make systemic improvements in the way they deliver chronic care. The LifeMasters Demonstration program is a population-based program targeting people dually eligible for Medicare and Medicaid with particular diagnoses and is also at financial risk for program fees. Enrollment through January 2006 was 50,654 (36,182 of whom were in the treatment group). LifeMasters’ fees are lower because it is not providing prescription drug coverage. The Medicare Health Support Program (formerly called the Chronic Care Improvement Program) provides DM on a population scale to all eligible beneficiaries in a geographic area – again, bearing risk for financial performance. The nine providers began operating in 2004 and are expected to serve 180,000 beneficiaries.

Early evidence suggests caution

Early evidence from the first three CMS demonstrations – IDEATel, MCCD and the Medicare Disease Management Demonstration – emphasizes the importance of waiting for more evidence before wholesale adoption of DM. The three early CMS demonstrations required patients to agree to participate in DM. All three have had significant problems enrolling patients. IDEATel needed more than an extra year to obtain its target sample size and had a higher-than-expected dropout rate of 19% in the first year [8]. During the first year of the MCCD demonstration, six of the 15 programs enrolled fewer than half the number of patients they had targeted, citing initial overestimates of the number of eligible patients from referral sources, physicians’ failure to encourage their patients to enroll, high patient refusal rates, and difficulty having nurses both recruit and serve patients [9]. These enrollment shortfalls suggest that purchasers of DM programs need to be careful when determining how to pay for services, to ensure they are not paying for coverage of people who are not receiving DM services. They also suggest that different models that rely more on partnerships with usual care providers may be more effective at gaining the trust of beneficiaries – an essential factor in convincing patients to make difficult behavioral changes.

Short-term findings provide some evidence that DM may improve quality of care, however anticipated reductions in costs may not materialize during the first year. One-year results from the IDEATel demonstration indicate that daily testing of blood sugar levels increased, but the demonstration did not increase adherence to healthy diet regimens; and had no effect on exercise regimens in one site and only a small effect in the other site. Not surprisingly, the intervention had no clear-cut effects on body mass index and waist-to-hip ratio. The intervention did have somewhat favorable effects on the use of recommended medications and, notably, had substantial impacts on diabetes control, lipid levels and blood pressure. However, the implementation costs were more than US$8000 per participant per year, which was greater than the enrollees’ total Medicare expenditures for Part A and B services, making it impossible for the intervention to be cost-neutral [8]. Early evidence from the MCCD demonstration also shows a need for caution. Survey data indicate that the DM programs are popular with patients (once they enroll) and patients’ usual care physicians. Early enrollees in the treatment group reported better access to information and appointments, better communication among their providers, and greater understanding of their health conditions, however, the interventions have not led to statistically significant increases in self-reported adherence to medication, diet and exercise regimens, despite their focus on patient education [9].

Results of all three of these evaluations should be available in the coming years, with the results from the first 25 months of operations for the MCCD slated to be released later this year [10]. Until then, without rigorous evidence on the effects of DM, purchasers should proceed with caution.

Disclaimer

This editorial, based in part on research funded by the Centers for Medicare & Medicaid Services, presents the opinions of the author alone. It does not necessarily represent the opinions of the Centers for Medicare & Medicaid Services.