Long-Term Effects of Care Management on Mortality, Hospitalizations, and Medicare Costs Among Chronically-Ill Medicare Beneficiaries

Abstract

Care management interventions for chronically-ill Medicare beneficiaries are central to reforms in the Affordable Care Act aimed at improving health outcomes while reducing expenditures. This dissertation examines the long-term effects of 15 care management models tested in the Medicare Coordinated Care Demonstration (MCCD). These nurse-led interventions to improve patient self-management and clinical care aimed to reduce medical expenditures by reducing the need for expensive hospitalizations. Approximately 25,000 Medicare beneficiaries were randomly assigned to receive usual care or care management services for up to 4 to 11+ years, depending on the program. This dissertation (1) assesses the programs' long-term effects on survival, and (2) tests whether effects on hospitalizations and Medicare costs changed over time, as would be expected if the interventions put patients on a better health trajectory.

Most programs had no measurable effects on survival or hospitalizations, either over the full follow-up period or during distinct time periods. Including program fees, seven programs increased costs to Medicare by 6 to 55%. Several programs affected survival and/or hospitalizations, but the estimated effects were both favorable and unfavorable. Two programs targeting heart failure patients reduced mortality risk by 18 and 37%. One program may have increased the hazard of dying in later years of enrollment. Two programs targeting high-risk patients with a range of possible diagnoses reduced hospitalizations by about 10% over many years, and may have been cost-neutral to Medicare. As a group, the four programs targeting heart failure patients reduced hospitalizations by 8% soon after patient enrollment but these effects dissipated by the third year. One program increased hospitalization rates by 19%.

These results indicate that many of the care management models tested in the MCCD risk increasing costs to Medicare without measurably improving select outcomes. However, programs with particular design features (e.g., frequent monitoring or in-home visits, depending on the target population) may be able to: (1) improve survival, and/or (2) reduce hospitalizations at least enough to cover program costs. Further testing is needed to verify whether programs with such features can replicate or improve the results seen in the MCCD.