The approach toward care coordination in an intervention can be captured by looking at a set of key components or features, which we will call “intervention domains.” These domains have been identified through an evaluation of the Medicare Coordinated Care Demonstration (MCCD; Peikes et al. 2009) and in developing a design for a new care coordination demonstration, the Medicare Chronic Care Practice Research Network (MCCPRN). These domains are based on experiences and the recent literature, in addition to management and through review of project directors of each program identified in Table 2, Chapter 5 (see Schraeder & Shelton 2009). Each of these 14 domains, as identified in Table 3.1, can play a crucial role in affecting the outcomes of an intervention. Also, specific domains are likely to be more important for some patients than others, depending on their healthcare needs, and this may vary over time as their health and social situations change.

Table 3.1 Domains of Care Coordination Interventions

As mentioned above, in order to have any hope of generating sufficient medical savings to cover its cost, a care coordination intervention needs to accomplish the difficult task of reducing beneficiaries’ needs for hospitalizations. Achieving a sizeable reduction in the number of hospitalizations for a chronically ill population of Medicare beneficiaries is inherently fraught with difficulties, due to the so called “funnel effect,” which can perhaps be best explained with the following illustration. For a voluntary (opt-in) care coordination intervention targeted at Medicare beneficiaries with one or more serious chronic illnesses such as congestive heart failure (CHF), chronic obstructive pulmonary disease (COPD), or coronary artery disease (CAD), let us assume that patients who enroll in the program have, on average, one hospitalization per member per year. Even if we assume that half of these hospitalizations are theoretically preventable, and the care coordination intervention succeeds in actually preventing 30% of the theoretically preventable hospitalizations, the overall reduction in the number of hospitalizations would only be 15% (which would be expected to produce an associated cost reduction of something around 10%). In other words, even under quite optimistic assumptions about the potential to reduce hospitalizations and the effectiveness of the intervention, it is difficult to achieve a sizeable reduction in the number of hospitalizations for beneficiaries suffering from multiple, serious chronic illnesses.

In spite of such difficulties, some interventions do succeed in reducing hospitalizations and controlling costs, but other similar sounding interventions are not successful on this dimension. Hence, it is important to identify the reasons behind the success of specific interventions. In a recent commentary, Mahoney (2010) identifies three constructs that determine the success of multifactorial fall prevention interventions: content, process, and the choice of target group. These factors are equally applicable to the success of care coordination interventions. Content, according to Mahoney (2010), refers to the components of an intervention that are integral to its success (for example, refer to the domains of care coordination in Table 3.1). Process refers to the way in which the intervention is delivered to induce uptake and behavior change – a particularly difficult task in a fee-for-service environment. The choice of an appropriate target group is equally crucial, if enrollees are at limited risk of hospitalization, it will be difficult to deflect many hospital admissions; conversely, some patients could be so ill that it is too late in the progression of their disease to reduce the need for admissions. It is useful to note here that among the handful of successful care coordination models reviewed in detail in Section VI, including the MCCD results noted above, one common feature was that each of these models succeeded in reducing expenditures only for a high-risk subgroup of patients, although the exact definition of high-risk varied from one intervention to another.

Earlier reviews of the literature

Boult et al. (2009b) performed a systematic review of models of comprehensive healthcare that have been found to be promising with respect to improving the quality, efficiency, and health-related outcomes of care for elderly patients with chronic diseases. They identified 15 models belonging to the following six model categories that have improved at least one outcome: interdisciplinary primary care (1), models that supplement primary care (8), transitional care (1), models of acute care in patients’ homes (2), nurse-physician teams for residents of nursing homes (1), and models of comprehensive care in hospitals (2). We reexamined the evidence presented by Boult et al. (2009b) to further limit the set of promising models to those that have evidence of reducing hospitalizations. For instance, evidence from a meta-analysis they reviewed suggests that interdisciplinary primary care teams focused on heart failure can reduce hospitalizations and total costs. However, among models that supplement primary care, the evidence is mixed. Transitional care interventions that facilitate safe transitions from hospital to another healthcare setting or to home have shown clear evidence of reducing hospital readmissions and costs; however, these interventions are short-term in nature. Finally, hospital-at-home programs that provide acute care for certain conditions in patients’ homes and nurse-physician teams for nursing home residents have also demonstrated their capacity to reduce inpatient admissions.

Bodenheimer and Berry-Millett (2009) synthesize evidence from a large number of studies on care management for patients with complex healthcare needs, including Medicare care management demonstrations. They offer a useful summary of findings from such studies, including the characteristics of successful care management programs, where success is broadly defined in terms of quality improvement or reductions in hospitalizations and costs. Again narrowing our focus to models for which Bodenheimer and Berry-Millett (2009) find evidence for reductions in hospitalizations, their review shows (like Boult’s) that transitional care interventions that target complex patients with multiple diagnoses who are being discharged from hospitals are highly effective in reducing hospital readmissions and total costs. However, the evidence on reductions in hospital use and costs is somewhat mixed for care management interventions in a primary care setting or for care management within integrated delivery systems, and any favorable evidence on cost savings from disease management is based on weak and questionable research methodologies. Bodenheimer and Berry-Millett (2009) also note that Medicare demonstrations of care coordination have usually failed to find evidence for reductions in service utilization and costs. They go on to summarize the characteristics of successful care coordination programs, using a broad definition of success, as previously mentioned: selection of the right patients or the right target group; person-to-person encounters between patients and care managers, including home visits; appropriate training of care managers; a multidisciplinary care management team including physicians; presence of informal family caregivers at home; and the use of coaching techniques as part of care management.

Brown (2009) adopts a more stringent definition of an “effective” care coordination intervention in that to be deemed effective, an intervention should have reduced hospitalizations and costs. His review identifies three types of interventions that have demonstrated effectiveness in reducing hospitalizations for Medicare beneficiaries with multiple chronic diseases: (1) transitional care interventions, in which patients are first engaged within the hospital and then followed intensively over four to six weeks after discharge, to ensure adherence to medication and self-care instructions, recognition of symptoms requiring immediate attention, and in keeping appointments with primary care physicians; (2) self-management education interventions that engage patients for four to seven weeks in community-based programs in order to activate them in the management of their own conditions; and (3) coordinated care interventions that identify patients with chronic conditions at high risk of hospitalizations, conduct initial assessments and care planning, and provide ongoing monitoring of patients’ symptoms and self-care, working with the patient, primary care physician, and caregivers to improve the exchange of information. Brown (2009) discusses the implications of the findings from his review for the design of the PCMH model, and for approaches to financing of care coordination models, and ends by identifying issues for ongoing research.

Bott et al. (2009) provide a qualitative summary of the findings from seven disease management (DM) demonstrations conducted by the Centers for Medicare & Medicaid Services (CMS) since 1999 involving around 300,000 fee-for-service Medicare beneficiaries. In general, the findings from these demonstrations have not been encouraging in that the results have not consistently shown evidence for improvement in evidence-based care, satisfaction with care, or broad behavior change. Also, very few programs have led to cost savings for Medicare, net of program fees. For instance, the Case Management Demonstration involving beneficiaries with heart failure and diabetes achieved some reduction in hospitalizations and Medicare costs, but cost savings were not sufficient to offset program fees, and in the Care Management for High-Cost Beneficiaries Demonstration, three of the six participating programs achieved financial savings of which only two had sufficient savings to cover program fees and meet the required 5% savings target, net of fees. Bott et al. (2009) conclude by noting that several factors will determine the success of future demonstrations involving beneficiaries with chronic diseases, including developing a better understanding of the targeted population and their barriers to getting high-quality care, redesigning future demonstrations based on current evidence, and reassessing ongoing demonstrations to define parameters of success.

Below we update the findings in these existing reviews with findings from recent demonstration evaluations, and then go into more detail about the most successful of the original studies that these reviews have highlighted.

Findings from medicare demonstrations

Over the past decade, CMS has conducted several demonstrations involving thousands of Medicare beneficiaries with chronic illnesses. Barring a few exceptions, the findings from these demonstrations have not been encouraging in that, cost reduction, if any, achieved by any of these demonstrations, was not sufficient to cover intervention costs (Bott et al. 2009). These demonstrations can be grouped into the following two categories based on their mode of beneficiary targeting: (1) population-based disease management programs, and (2) voluntary care management programs. Key features of and findings from several demonstrations falling in either of these categories are summarized below, followed by a brief summary of the lessons learned from the evaluations of these demonstrations.1

Population-based disease management

The congressionally mandated Medicare BIPA Disease Management Demonstration was intended to provide disease management services and a comprehensive prescription drug benefit to certain chronically ill beneficiaries to test whether disease management in the traditional fee-for-service program led to improved outcomes and lower total costs to Medicare. Medicare FFS beneficiaries with advanced stage CHF, diabetes, or CAD were recruited into three programs operating in Louisiana, Texas, California, and Arizona. The demonstration, which began in 2004, however, did not continue to conclusion, as none of the programs had impacts on the key outcomes of Medicare Part A and Part B expenditures and service use (Chen et al. 2008); the programs were therefore terminated before the end of their originally scheduled three-year duration.

Although all three disease management programs in the demonstration used Medicare claims data to identify potentially eligible beneficiaries and recruited beneficiaries through letters, telephone calls, or referrals by physicians and hospitals, they encountered unanticipated difficulties with recruitment, and two programs did not meet their enrollment targets. An independent evaluation found that the programs differed in their timeliness of conducting initial patient assessments and also in their frequency of patient contacts that additionally varied with program maturity. Although, overall, none of the programs had any impacts on hospitalizations, emergency room visits, or Medicare expenditures, one program (XLHealth) reduced hospital admissions and Medicare expenditures among a later cohort of enrollees, specifically those who enrolled after the first six months of program operations. However, even for this later cohort of enrollees in the program, the cost savings were not sufficient to cover program fees (Chen et al. 2008). Finally, the programs had limited or no impacts on prescription drug access, patients’ satisfaction, functioning, mortality, and potentially preventable hospitalizations.

The Lifemasters Supported SelfCare Demonstration, a randomized controlled trial covering the three-year period from January 2005 to December 2007, was a CMS sponsored population-based disease management demonstration program implemented by LifeMasters Supported SelfCare (LifeMasters). The program targeted Medicare fee-for-service beneficiaries who were dually enrolled in Medicaid (dual-eligibles); resided in Florida; and had CHF, CAD, diabetes, or a combination of the three. A sample of Florida beneficiaries meeting the target criteria were randomly assigned to the treatment group, which was eligible to receive the LifeMasters intervention, or the control group, which was not eligible for it; both groups retained their regular Medicare and Medicaid fee-for-service benefits. The LifeMasters intervention included patient assessment and care planning, routine nurse monitoring, patient self-monitoring, patient education, care coordination, and (limited) service arrangement. The intervention was conducted primarily over the telephone by registered nurses, although some nurses also conducted in-person visits to patients classified by the program as frail. LifeMasters classified enrolled treatment group members as either active or inactive. Active beneficiaries were those who agreed to participate, at some level, in the demonstration and for whom LifeMasters received a monthly management fee. LifeMasters further classified active beneficiaries as mediated or instructional. Mediated patients participated fully in the disease management program; instructional patients received only a quarterly health magazine and an occasional telephone call from program staff. LifeMasters was at financial risk for this demonstration, meaning that while it would receive a substantial share of any net savings, it was required to reimburse CMS for any net losses (that is, gross savings in Medicare savings minus fees paid). Gross savings were calculated by the Actuarial Research Corporation (ARC) as the difference in average Medicare expenditures per month for the full treatment group (that is, all beneficiaries assigned to the treatment group, regardless of whether they actually engaged with the program) minus average monthly expenditures for the control group, multiplied by the number of treatment group member months.

The formal evaluation of the LifeMasters demonstration, using an intent-to-treat design, found that it did not reduce total Medicare expenditures, hospitalizations or emergency room visits (Esposito et al. 2008a; Esposito et al. 2008b) for the sample of 36,959 enrollees and 14,797 control group members. However, monitoring reports produced by ARC indicated that the intervention might have had impacts on the subset of beneficiaries with CHF and those who had both diabetes and CAD, and on beneficiaries residing in a subset of the demonstration counties. Therefore, at LifeMasters request, CMS approved a demonstration redesign, to begin in March 2007, for which eligibility criteria were narrowed to include only beneficiaries who resided in select counties and who had CHF only, or at least two of the three targeted conditions (CHF, CAD, and diabetes). Additionally, LifeMasters attempted to enhance its efforts to mediate patients and continued with several program enhancements that were started in the fall of 2006, including wound care, end-of-life planning programs, and complex case management. The evaluation study population for the redesign period of March 2007 through August 2009 included two groups of randomly assigned beneficiaries: those enrolled in the demonstration before March 1, 2007, and eligible for the redesign (Cohort 1) and those enrolled after the start of the redesign (Cohort 2), and thus provided a unique opportunity to study the impact of the intervention on two distinct populations with varying degrees of exposure to the program. An evaluation of the redesign phase of the intervention, however, found that the program had no effects on quality of care, health care utilization, or Medicare expenditures (Stewart et al. in press).

The program’s failure to generate savings, despite the apparent comprehensiveness of the interventions and the targeting on high-risk beneficiaries who are dually enrolled in Medicaid and have serious chronic illnesses, suggests that it may be difficult for telephonic disease management programs to achieve savings in a Medicare fee-for-service environment. However, even after the redesign, LifeMasters succeeded in contacting only 67.6% to 77.3% of patients in the two cohorts within 13 weeks of enrollment, and no more than 30% of the active treatment group members were mediated (fully engaged) at any point of time during the redesign phase. Furthermore, the program had limited contact with even the mediated patients, and LifeMasters made no use of the Medicaid prescription drug data that it had purchased from the state for its enrollees. Thus the intervention may have been too weak to be effective for this population.

Medicare Health Support (MHS): The Phase I of the MHS Demonstration, was a three-year (2005–2008) pilot designed to test a variety of care management interventions for invited fee-for-service Medicare beneficiaries with CHF or diabetes, tested a range of program models serving diverse populations in urban and rural areas. The programs offered self-care guidance and support to chronically ill beneficiaries to help them manage their health, adhere to their physicians’ plan of care, and assure that they obtain medical care that they need to reduce their health risks. The Phase I programs together served approximately 100,000 chronically ill Medicare beneficiaries.

The program was cancelled in August 2008 because experience from Phase I of the MHS program suggested that the program did not meet the statutory requirements of improved clinical quality outcomes, improved beneficiary satisfaction, and financial savings (CMS 2008). An initial evaluation conducted by RTI based on the first six months of program operations for the eight Medicare Health Support Organizations (MHSOs) found few or no significant differences in acute care utilization between the treatment and control groups. The report also notes that negotiated program fees were not covered by reductions in Medicare expenditures over the first six months, and without a reduction in these fees programs may not be budget neutral in the long run (McCall et al. 2007). While the early report noted an unexpected and growing divergence between the treatment and control groups from the time of random assignment to the time of program start-up in monthly Medicare expenditures that could affect the ability of the MHSOs to meet their savings targets, none of the analyses suggested that the programs reduced hospitalizations or costs.

The Care Management for High Cost Beneficiaries (CMHCB) demonstration, originally approved for three years, was launched by CMS in 2005 with sites in different areas of the country. It tests provider-based intensive care management services as a way to improve quality of care and reduce costs for fee-for-service beneficiaries with one or more chronic illnesses (with the primary focus on CHF, diabetes, and chronic kidney disease [CKD] and high Medicare costs). Six organizations were selected to provide disease management services to a large number (1,800 to 15,000) of beneficiaries. CMS pre-selected beneficiaries for the demonstration according to eligibility criteria, but participation in the demonstration is voluntary. Program services were intended to increase adherence to physician prescribed care, reduce unnecessary hospital stays and emergency room visits, and help participants avoid expensive and debilitating complications. Each of the six organizations chose their own care management interventions that ranged from a distributed network of personal visiting physicians (PVPs) who see patients urgently and routinely in their homes and nursing facilities, or intensive disease management directed by nephrologists in supplementary clinics to use a technology platform in patients’ homes to coach them about their health, collect vital signs, and transmit risk-stratified results to multispecialty medical groups. While detailed results from the evaluation of the CMHCB demonstration is awaited, CMS granted three-year extensions to three of the six original programs in the demonstration, based on the initial success of these programs in meeting or exceeding the savings required by the demonstration agreement (CMS 2009a). These three organizations, serving beneficiaries in New York, Massachusetts, Oregon, and Washington, would be subject to monthly operational monitoring and quarterly financial evaluations of performance with the demonstration being extended one year at a time based on their financial status or yearly projected savings. Savings were not measured using a randomized design, but rather were based on projections.

Voluntary care management for high-risk patients

The Medicare Case Management Demonstrations studied the appropriateness of providing case management services to beneficiaries with catastrophic illnesses and high medical costs. It tested case management as a way of controlling costs in the fee-for-service sector. These demonstrations were implemented in three Midwestern sites (Indiana, Iowa/eastern Nebraska, and suburban Detroit) in October 1993 and continued through November 1995. The three programs chose different target populations, (for example, those with CHF only, CHF or COPD, or any of eight conditions), and the style and focus of case management differed greatly across the three programs, specifically in their levels of in-person contact, use of nurses and social workers, degree to which case management was structured or allowed to evolve, and their emphasis on education and service coordination. However, the programs also shared a number of key activities, such as client assessment and periodic reassessment, service coordination and monitoring, condition-specific self-care education, and emotional support to clients and their informal caregivers.

An evaluation of the Case Management Demonstrations found that although the programs succeeded in identifying and enrolling Medicare beneficiaries at risk of high Medicare spending, participation rates were much lower than expected for all three programs, with beneficiaries older than 85 years and those who died within six months of the participation decision being less likely to participate (Schore et al. 1999). Also, in spite of high levels of satisfaction among eventually participating beneficiaries, none of the programs improved patient self-care or reduced hospital admission rates and Medicare spending. In fact, there was a significant increase of 10% points in the proportion of patients admitted to the hospital in one site, which was a hospital; the number of hospital admissions increased by 34%. While the evaluation was unable to conclude whether the observed increase in hospitalizations in this site was a true program effect, it did note that since the program was hosted by a hospital, it might have been more receptive to admitting program clients for observation or treatment. The paper noted four possible reasons for the lack of intended impacts on health behavior or Medicare service use and expenditures: no involvement of clients’ physicians in the interventions, lack of a clear focus in the interventions and goals, lack of staff with sufficient case management experience and clinical knowledge, and inadequate financial incentives to reduce Medicare spending.

The Community Nursing Organization (CNO) Demonstration tested a capitated, nurse-managed system of care that provided a specified package of community-based services, in conjunction with case management, under a capitated payment methodology. In 1993, CMS selected four organizations in Illinois, Arizona, Minnesota, and New York to provide community nursing and ambulatory care services to Medicare beneficiaries. Services covered as part of the CNO service package included home health services, medical supplies, appliances, and devices, durable medical equipment, ambulance services, outpatient physical therapy, services provided by a clinical psychologist or a clinical social worker, and case management services–defined as services which assist enrollees in gaining access to and coordinating/approving utilization of needed medical, social, educational, and other services. Each of the CNO sites was free to define and configure the process of case management in the way it judged to be most beneficial to the members and efficient for the organization. Methods of assessment, resources devoted to planning and monitoring, as well as the number of members whose care was actively managed, therefore, differed from site to site.

Applicants to the CNOs were randomized to treatment (CNO) or control (traditional Medicare) groups, and based on this randomized research design, the preliminary evaluation report covering the period from January 1994 through 1999, found that Medicare spending per person per month was higher for members of the treatment group than for members of the control group in each of the programs, with these differences being statistically significant at three of the four sites. Furthermore, average monthly Medicare spending in the treatment group kept increasing relative to the population over the course of the demonstration (Frakt et al. 2003). However, based on data collected from two sites on beneficiary satisfaction, the final report also found that an overwhelming majority of enrollees at both sites were satisfied with the care received and felt that their nurse consultant was available when needed.

The Informatics for Diabetes Education and Telemedicine (IDEATel) Demonstration tested the effects of providing home-based telemedicine services to 1,093 eligible Medicare beneficiaries in two cohorts who had diabetes mellitus and lived in medically underserved areas in New York City and upstate New York. The demonstration began in February 2000, and was originally scheduled to end in February 2004. However, Congress extended the demonstration and the evaluation for a second four-year period and the demonstration ended in February 2008. In both phases eligible Medicare beneficiaries from New York City and upstate New York who volunteered to participate in the demonstration were randomly assigned to either the treatment or a control group. During the demonstration, control group members in both sites received usual diabetes care from their primary care physicians. Treatment group participants also continued to see their primary care physicians and additionally received a home telemedicine unit (HTU), which they could use to (1) measure and monitor blood pressure and blood sugar and transmit their measurements to a nurse case manager, (2) communicate with a nurse case manager via audio/videoconferences, known as televisits, and (3) access web-based chat rooms and educational materials available only to participants.

The evaluation of IDEATel (Moreno et al. 2008; Moreno et al. 2009) found that the intervention as delivered was neither as intensive nor as technologically sophisticated as originally designed, since the Consortium delivering the intervention encountered unexpected challenges and deliberately departed from its plans in some areas. Most importantly, a relatively low proportion of treatment group members made consistent use of the HTUs, and very few used a wide range of its functions. While IDEATel improved clinical outcomes in one site, it had no statistically discernable impact on Medicare Part A and Part B expenditures or the use of expensive services, such as hospital admissions in either phase of the demonstration, in either site. Furthermore, the high intervention costs of the demonstration – more than $8,000 per participant per year – exceeded the control group’s combined Part A and Part B expenditures in upstate New York, and were far higher than the costs of comparable home telemedicine programs. Given the absence of effects on costs or services, even a less expensive version of this demonstration would not have produced sufficient Medicare savings to offset demonstration costs (Moreno et al. 2008).

The Medicare Coordinated Care Demonstration (MCCD): In early 2001, CMS selected 15 demonstration programs for the MCCD out of 58 applicants in a competitive awards process under which each program was allowed to define, within broad boundaries, its own intervention and target population. Each program began enrolling patients between April and September of 2002, and was authorized to operate for four years. Eleven of the 15 programs later requested, and were granted, two-year extensions, and continued to operate into 2008. The program hosts included disease management companies, community hospitals, academic medical centers, integrated delivery systems, hospice, and an upscale retirement community, which were located in geographically diverse areas. Although the care coordination interventions of the 15 programs varied widely, all programs assigned patients to a nurse care coordinator–usually a registered nurse–and in all programs the care coordinator assessed patients and developed patient care plans.

The evaluation of the 15 original MCCD programs showed that none generated net savings to Medicare (Peikes et al. 2008) and 14 programs showed no statistically significant differences in hospitalizations. However, one program (Mercy Medical Center) significantly reduced hospitalizations and had a sizable yet statistically insignificant reduction in Part A and B expenditures, and another program (Health Quality Partners [HQP]) showed promise, with both hospitalizations and expenditures about 11% lower in the treatment group, though these decreases were not statistically significant (Peikes et al. 2009). CMS subsequently granted these two programs extensions to operate for two more years. A Report to Congress on these two promising MCCD programs (Mercy and HQP) found that they significantly reduced hospitalizations among a high-risk subgroup (enrollees with COPD, CHF, or CAD and at least one hospitalization in the year before randomization, or with any of 12 chronic conditions and at least two hospitalizations in the prior two years). HQP also achieved significant cost reductions for this subgroup (Schore et al. 2010). Further, ongoing analysis of data from 11 of the original 15 programs over 2002–2008 shows that four of these programs significantly lowered hospitalizations by 8% to 33% among a high-risk subgroup of beneficiaries (Peikes et al. 2010). The implications of these recent findings from the ongoing analysis of data from the MCCD are discussed in greater detail in Section V below.

Lessons learned from the evaluations of medicare demonstrations

The findings from the Medicare demonstrations, whether the population based disease management programs or voluntary care management, are, in general, disappointing in that none of these meet our criteria for a successful intervention that reduces the number of hospitalizations and Medicare expenditures. Also, some of these demonstrations, such as the Disease Management and the Case Management Demonstrations encountered difficulties in beneficiary recruitment and participation, while for some others such as the IDEATel demonstration, intervention costs were too high to lead to any savings in Medicare expenditures. Although full evaluation results for the CMHCB demonstration are awaited, none of the other demonstrations met their budget neutrality requirements. Valuable lessons, however, have emerged from the evaluation of the MCCD, in particular from recent findings that four programs reduced hospitalizations and costs among a high-risk subgroup of beneficiaries. Also, as briefly discussed above in Section III, these four successful programs share certain intervention characteristics that are lacking in some of the other MCCD programs. Overall, the findings from the ongoing analysis of data from the MCCD suggest that care coordination can generate savings to Medicare if the right interventions are targeted to the right people. In other words, following Mahoney’s (2010) framework, identifying an appropriate target group, carefully developing the content of an intervention tailored to the needs of the target population, and delivering the intervention in a patient-centered way by triaging and activating patients are the key determinants of success in a care coordination program. Also, as noted by Bott et al. (2009) in their review of disease management demonstrations in Medicare, reassessing ongoing demonstrations to define parameters of success in future demonstrations, and striking a balance between program costs and expected savings or between program costs and value in terms of improvements in quality and outcomes would be crucial to future success in Medicare demonstrations.

Successful programs/models

Very few care coordination or care management programs that used a randomized design to compare an intervention group with an equivalent control group have reported significant impacts on hospitalizations or costs of health services. Effective care coordination program models that have reported impacts in hospitalizations or cost fall into one of three categories: (1) transitional care interventions (Naylor et al. 2004; Coleman et al. 2006), (2) self-management interventions (Lorig et al. 1999; Lorig et al. 2001; Wheeler et al. 2003) and (3) care coordination interventions (Care Management Plus, the Geriatric Resources for Assessment and Care of Elders, Guided Care, and four sites that participated in the CMS Medicare Coordinated Care Demonstration [MCCD]). In the next section we briefly describe these programs and their outcomes.

Transitional care

Some of the strongest evidence of care transition interventions that reduce hospitalizations and costs are two well-tested models designed to reduce readmissions to hospitals. The transitional care intervention developed by Naylor et al. (2004) targeted patients who were hospitalized for CHF and used highly trained advanced practice nurses (APNs) to administer the intervention. The intervention was highly structured and effective. The APNs met with patients in the hospital and in their homes shortly after discharge to provide intense coaching and education on medications, self-care, and symptom identification. The intervention lasted a total of 12 weeks, and patients were followed for one year. The intervention was evaluated with a randomized design and intent-to-treat approach. During the year following the hospital discharge, the number of rehospitalizations per patient year was 34% lower in the treatment group than the control group. In addition, rehospitalization rates in the treatment group were 44.9% compared to 55.4% in the control group, a difference of 10.5%. At one year, treatment group patients also had mean total costs 39% lower than control group patients ($7,636 versus $12,481). The total intervention cost was $115,856 ($982 per patient).

The other successful transitional care model, developed by Coleman et al. (2006), also used advanced practice nurses as the care coordinators (referred to as “transition coaches”), but targeted hospitalized patients with a range of chronic conditions. Under Coleman’s model, the one-month intervention provided patients with (1) tools to promote cross-site communication, (2) encouragement to take a more active role in their care, and (3) continuity of care and guidance from their transition coach. The intervention was evaluated with a random design. Intervention patients had lower rehospitalization rates than control subjects at 30 days (8.3% versus 11.9%) and 90 days (16.7% versus 22.5%), as well as lower rehospitalization rates for the same condition that precipitated the initial hospitalization at 90 days (5.3% versus 9.8%) and 180 days (8.6% versus 13.9%). In addition, mean hospital costs were $488 lower for intervention patients than controls at 180 days ($2,058 versus $2,546). The annual cost of the intervention was $74,310 ($196 per patient), resulting in a net cost savings of approximately $147,797 over the six-month follow-up.

Self-management

Another model that has been shown to generate reductions in hospitalizations is one that focuses on educating patients in how to self-manage their conditions. Kate Lorig and John Wheeler both developed self-management models and produced studies with favorable results. The programs focus on four factors: (1) identifying patients’ goals; (2) improving their self-management skills; (3) building their sense of self-efficacy; and(4) assessing their mastery of these skills.

Lorig et al. (1999; Lorig et al. 2001) offered a community-based self-management program to patients who were 40 years of age or older and had a physician-confirmed diagnosis of heart disease, lung disease, stroke or arthritis. In seven weekly group sessions, course leaders provided program participants with instruction on exercise, cognitive symptom management techniques, nutrition, fatigue and sleep management, use of medications, dealing with emotions, communication, problem-solving, and other topics.

The program was evaluated with a six-month randomized, controlled trial with an intent-to-treat approach. Compared to control subjects, treatment subjects demonstrated improvements at six months in weekly minutes of exercise, cognitive symptom management, communication with physicians, and other healthy practices. Treatment subjects also had one-third fewer hospital stays (0.17 versus 0.25) and spent, on average, half as many nights in the hospital as control subjects (0.8 versus 1.6). Treatment subjects also generated $820 less in average six-month health care costs than control subjects. The cost of the intervention was only $70 per participant, which produced health expenditure savings of approximately $750 per participant over the six-month follow up.

Wheeler’s (2003) model is similar to Lorig’s work. Drawing from six hospital sites, Wheeler administered the program to women who were 60 years or older and had a diagnosis of cardiac disease. The four-week program featured weekly group meetings in which health educators taught program participants to manage cardiac problems such as diet, exercise and taking medicine. The program was assessed with a randomized design, and an intent-to-treat approach. The results show that over a 21-month period following the intervention’s conclusion, the treatment group experienced 39% fewer in-patient days and 43% lower in-patient costs than women in the control group. The program cost about $374 per patient, resulting in a ratio of medical expenditure savings to program costs of approximately 5 to 1.

Care coordination

Care Management Plus (CMP) is a primary care based interdisciplinary team model from Intermountain Healthcare and Oregon Health and Science University. CMP includes care managers, either registered nurses (RNs) or social workers, located on-site, working in health care teams and using information technology (IT) to help care for patients with complex chronic illness. CMP prioritizes health care needs through structured protocols and tools to assist patients and caregivers to self-manage their chronic diseases. Specialized IT tools allow clinicians to access individualized care plans and receive reminders about best practices, and they facilitate communication among the health care team and with other providers such as subspecialists. RN care managers typically have a caseload of approximately 350 to 500 patients and provide individualized assessment and care planning, using disease management guidelines; teaching/coaching self-management skills; assisting in care transitions and coordinating services with different providers; and proactive monitoring and providing ongoing guidance and support (Dorr et al. 2006; Dorr et al. 2007a). CMP has demonstrated several benefits to chronically ill patients. It reduced annual mortality rates by more than 20% and hospital admission rates by 24% to 40% for patients with diabetes and depression (Dorr et al. 2008), and improved quality of care through higher adherence to clinical testing guidelines and increased physician productivity (Dorr et al. 2007b).

In the Geriatric Resources for Assessment and Care of Elders (GRACE) program, on-site support teams, comprised of an advanced practice nurse and social worker, provide comprehensive, home-based primary care for low-income seniors receiving care through community health centers. The support team meets with off-site geriatrics interdisciplinary teams to review each patient at least quarterly (Counsell et al. 2006). Each support team has a caseload of approximately 100 to 125 patients and are seen face-to-face by the support team at least every two months. The focus of the intervention is on an initial in-home assessment and individualized care plan; the use of specific clinical protocols; an electronic medical record, and a web-based care management tracking tool; the integration of affiliated pharmacy, mental health, home health and community-based and inpatient geriatric services; and the coordination and continuity of care among all health care providers and sites of care. GRACE has demonstrated improvements in quality of care, reductions in hospital admissions, and cost savings for a pre-defined subgroup of patients (Counsell et al. 2007; 2009) at high risk of hospitalization when they entered the program based on a PRA score greater than or equal to 0.4 (Boult et al. 1993). Overall, GRACE reduced emergency department visits by 17% for all intervention patients and 35% for high risk patients, reduced hospital admissions by 44% for high risk patients, and reduced expenditures (not including program costs) by 23% for high risk patients one-year post intervention.

In Guided Care, two to five primary care physicians partner with an on-site RN to provide comprehensive primary care to 55 to 60 of their elderly, chronically ill patients at high risk for using costly health services in the upcoming year. The RN conducts a comprehensive in-home assessment and then monitors and contacts patients every month, engages them and their caregivers in self-management activities, provides transitional care when needed, and coordinates important health-related services across different provider settings (Boyd et al. 2007; Boult et al. 2009a). The RN also helps patients to create a personal care guide and action plan and provides emotional support on an on-going basis. Preliminary data indicate that Guided Care improves quality of care (Boult et al. 2008; Boyd et al. 2010), reduces family caregiver strain (Wolff et al. 2010), improves physician satisfaction with chronic care (Marsteller et al. 2010), and reduces the use and cost of expensive health services (Leff et al. 2009; Boult et al. in press).

As noted above, although the 15 original MCCD national sites did not show net savings to Medicare (Peikes et al. 2009), ongoing analysis identified four programs that significantly reduced hospitalizations by 11% and Medicare expenditures (including program costs) by $178 per patient per month among their high risk patients (Peikes et al. 2010). High risk patients were defined as having chronic obstructive pulmonary disease (COPD), congestive heart failure (CHF) or coronary artery disease (CAD), and having at least one hospitalization in the prior year or any of 12 chronic illnesses and two or more hospitalizations in the previous two years. The four programs were implemented in four different geographical settings: (1) a hospital that is part of an integrated delivery system in rural northern Iowa (Mercy Medical Center, Mason City), (2) an urban academic medical center (Washington University, St. Louis), (3) a hospice and home health care provider (Hospice of the Valley, greater Phoenix area), and (4) a quality improvement service provider in suburban and rural Pennsylvania (Health Quality Partners). While the programs were different, they had core similarities. All four programs used RNs trained in comprehensive care coordination and focused on improved self-care, chronic symptom recognition and management, improved medication management, and improved transitional care and physician communication.

Two common features highlight the successful care management/care coordination models. The first was that transitional care and self-management education were integral components of the model. The second is that they all succeeded in reducing expenditures for a high-risk subgroup of patients, although the exact definition of high-risk varied from one intervention to another. The lessons learned from these successful models form the basis of essential care coordination components discussed in Chapter 5. We return to the identification of high-risk patients in the concluding section of this chapter.

Selected studies on medicaid and commercial populations

In this section, we critically evaluate findings from two studies that evaluate care coordination programs in Medicaid and a commercially insured population respectively. Although the demographic and health profiles of Medicaid beneficiaries differ from that of Medicare beneficiaries in important ways, including the higher prevalence of younger beneficiaries, homelessness, substance abuse, and mental and behavioral health problems among the former, a sizeable number of patients with chronic illnesses are enrolled in Medicaid, either with or without Medicare coverage. Hence, we are interested in interventions funded by Medicaid that served a similar population of beneficiaries with chronic illnesses. Similarly, even though a commercially insured population is likely to be younger and healthier on average, we review a recent evaluation of a telephone-based care management strategy for a commercially insured population that included beneficiaries with chronic illnesses.

Indiana chronic disease management program (ICDMP)

The ICDMP, assembled by Indiana Medicaid and implemented in 2003, had several components and was primarily designed to improve the quality and cost effectiveness of care for Medicaid beneficiaries with CHF, diabetes, asthma, and other conditions (Rosenman et al. 2006). The ICDMP was implemented in three stages: it was launched on July 1, 2003, in central Indiana for eligible participants with diabetes or CHF, and northern and southern parts of Indiana as well as a statewide asthma disease management program were added to ICDMP during 2004. ICDMP for adults with diabetes or CHF had several components: identification of eligible participants through their ongoing participation in the state’s primary care case management programs (PCCMs) for Medicaid beneficiaries and through queries of Medicaid claims to identify people with specific conditions; risk stratification based on predicted costs whereby the highest-risk (20%) of eligible participants were assigned to nurse care managers and the remaining (80%) to telephonic care management; use of information systems for decision support; and the creation of quality improvement collaboratives for primary care practices. It is interesting to note that more than 40% of the participants in the ICDMP who had diabetes or CHF were dually enrolled in both Medicare and Medicaid.

Holmes et al. (2008) evaluate the net fiscal impact of the ICDMP, based on a randomized research design, in which eligible Medicaid beneficiaries who had CHF, or diabetes, or both received chronic disease management services or standard care, based on the random assignment status of their Indianapolis-based primary medical provider’s practice. The study uses multivariate methods controlling for baseline differences in age, sex, risk status, prescription drug use, Medicare coverage, and levels of claims paid during the pre-ICDMP period to find that ICDMP significantly reduced claims paid by Medicaid for all beneficiaries with CHF, especially the low-risk beneficiaries with CHF, but it did not have any cost-saving effect for beneficiaries with diabetes. However, the study’s findings and conclusions are questionable due to several methodological issues.

The first problem with the study is that in spite of random assignment, the treatment and control groups were not balanced in their baseline characteristics – in particular, they differed widely in average Medicaid expenditures during the year prior to enrollment. While this difference can be controlled for statistically, the advantage of randomization is not realized in this study—the sizeable difference in pre-enrollment expenditures raises concerns that the two groups differed along unobserved or unmeasured characteristics, which could bias the impact estimates obtained by the authors. Moreover, the authors do not examine treatment-control differences in pre-intervention trends in Medicaid claims that could highlight potential differences in the claims pattern over time between the two groups, although they report having 21 months of pre-ICDMP data for each patient in the analysis sample. This problem arises due to the small sample size of individuals with CHF, only 186 beneficiaries, resulting in imprecise estimates that are heavily influenced by outliers. The problem is exacerbated by the finding that the effects are limited to the subset of 117 sample members who are in the low-risk subgroup of beneficiaries with CHF.2 Finally, the authors do not examine any other outcomes, such as service utilization outcomes, which would provide a reasonable robustness test and guide as to the potential source (for example, reduction in hospitalizations) of the estimated cost savings. The surprising findings of no cost savings for high-risk beneficiaries and favorable program impacts on only the low-risk beneficiaries with CHF are at odds with virtually all of the other studies discussed above, in which impacts are observed only for the high-risk subgroup of sample members.

Telephone-based care management for a commercially insured population

A recent population-based randomized study by Wennberg et al. (2010) evaluated a telephone-based care management strategy for a commercially insured population that included patients with selected chronic conditions, such as heart failure, COPD, CAD, diabetes, and asthma. In this study, 174,210 subjects were randomly assigned to either a usual support group (control) or an enhanced support group (treatment). The study team used predictive models to predict health service utilization and the likelihood of surgical intervention for a beneficiary with a preference-sensitive condition. The key difference between the two groups was in the extent of outreach – a greater proportion of patients received outreach through health coaches or interactive voice-response calls in the enhanced support group through the lowering of cutoff points for predicted health care costs or utilization. The intervention content delivered to the two groups – that reached 10.4% of the patients in the enhanced support group and 3.7% of the patients in the usual support group – was otherwise similar, and consisted of telephonic contacts by health coaches providing behavioral change and motivational counseling, and promoting shared decision making. The study finds that telephonic care management led to a statistically significant 4% reduction in total healthcare expenditures that was mainly driven by a significant 10% reduction in hospital admissions in the treatment group. Additionally, there were significant reductions in hospital admissions for at least two beneficiary subgroups, those with selected chronic conditions and those with high-risk conditions other than the selected chronic or preference-sensitive conditions. With low program costs of $2 per person per month, the results suggest that the net savings from the program was about $6 per person per month.

In spite of these impressive findings and an apparently strong research design, there are several problematic aspects of this study that make these results suspect. First, given that the treatment in reality is the differential outreach to the two groups, we believe that the analysis should have ideally been limited to those who met the new (lowered) threshold for the intervention but did not meet the original higher threshold, in both the treatment and comparison groups. Such a strategy would have ensured that (1) the analysis included only beneficiaries for whom the program could conceivably have had any effect, by excluding the 82% of sample members for whom it was irrelevant; (2) the study had the correct counterfactual – those eligible for the intervention who did not receive it; (3) the analysis was still an intent-to-treat approach in that all beneficiaries in both the treatment and control groups who met the target eligibility criteria were included in the analysis regardless of whether they actually received the intervention or not; and (4) despite the much smaller sample size, the analysis would have had more statistical power, given that the impacts have to be concentrated in the individuals selected.3 However, the authors conduct their analysis on all beneficiaries in both groups regardless of whether they were eligible for the intervention (that is, met the lower cutoff for predicted costs or utilization). This in turn leads to the second issue of whether these effects are reasonable or not. Given the nature of eligibility for the intervention, the effects detected for the full sample of beneficiaries can arise only from the greater proportion of beneficiaries in the treatment group who received the telephonic care management services, in other words, for the additional 18% or so of beneficiaries who became eligible to be contacted in the treatment group due to the lowering of the threshold or cutoff,4 or more specifically for the additional 6.7% of beneficiaries who were actually contacted in the enhanced support group (10.4% versus 3.7% in the usual support group). Since the effects observed for the full sample can only arise from this small percentage of patients in the treatment group who actually received services, the actual effects for this subgroup of treatment group patients must be extremely large.5 Such effects are not only highly unlikely for such a weak intervention, but are totally at odds with findings from previous randomized controlled trials that report little or no improvements in outcomes from telephone-based care management. The weak methodology raises major concerns regarding the validity of the findings.

Conclusions

CMS-funded demonstrations and empirical work over the last few years has yielded a wealth of new information on the effectiveness of various forms of care coordination/care management for patients with chronic illnesses. The importance of these findings for getting U.S. health care costs under control cannot be overstated. Some of the lessons learned are drawn from studies that have shown certain interventions to be ineffective, while others are drawn from singularly successful interventions. This non-systematic review of the evidence from recent studies published in the literature or in federally funded reports leads us to a number of conclusions:

Stay updated, free articles. Join our Telegram channel

Join