People dually enrolled in Medicare and Medicaid often have complex medical, long-term care, and social needs. However, dual enrollees are at risk of receiving less-than-optimal care, because Medicare and Medicaid medical providers generally address different health care needs and lack incentives to coordinate with each other. To reduce these risks and improve care, the Centers for Medicare & Medicaid Services, managed-care organizations, providers, and other stakeholders have worked together to develop various integrated care models. Understanding the effects of integrated care plans on health outcomes, utilization, and spending remains an ongoing priority to improve care for dual enrollees.
This literature review examines existing research that quantitatively evaluates whether, or the degree to which, integrated plans for dual enrollees meet their objectives. In doing so, we broadly categorize existing studies by methodologies employed and identify potential opportunities for future studies to measure the causal effects of integrated plans.
The main findings from the review include the following:
- A wide variety of methodologies have been employed to study the effects of integrated care plans. Many studies are descriptive in nature and do not have an explicit statistical identification strategy necessary to measure causal effects of integrated plans.
- Many studies use administrative data to measure outcomes, including medical care utilization and spending, and employ various regression-based designs with control variables intended to account for observable differences between integrated care plan enrollees and nonenrollees. However, information on enrollees’ demographic features in administrative data is very parsimonious, leading to concerns about the ability to adequately control for differences across groups and sufficiently eliminate bias.
- Difference in differences (DID) is the most common quasi-experimental research design used in recent evaluations of integrated care plans. DID research designs mitigate selection bias by comparing outcomes over time between enrollees and nonenrollees, but they have other important limitations.
- The evaluation literature would benefit from studies that employ additional quasi-experimental designs where possible. Such work could shed light on the robustness of existing studies that have used DID designs or more descriptive approaches (e.g., matching or regression adjustment). For example, a well applied regression discontinuity design could be more compelling than a DID design, because there would be less demand on the data to sufficiently minimize or eliminate bias.
- Many studies lack hypotheses for how integrated plans directly or indirectly affect the outcomes studied, making it difficult to interpret empirical results on how the plans may or may not have caused a change in a given outcome. Future studies would benefit from outlining hypothesized causal pathways, from plans’ features to the outcomes studied.
- Findings about the effects of integrated care plans on medical care utilization and spending tend to be mixed. Though enrollees’ experiences are generally found to be positive, few changes to quality of care or health outcomes have been identified, and program enrollment and participation have been found to be lower than expected. Together, these mixed findings suggest underlying heterogeneity in the design of integrated models, the populations targeted or enrolled, and/or the providers delivering care under the models and underscore the importance of further research in this area.