Diabetes and its complications are among the largest contributors to healthcare costs, cost growth, and patient morbidity and mortality. Diabetes is often not timely diagnosed. When it is diagnosed, patients often fail to comply with a complex treatment regimen. The health burden of diabetes falls disproportionately on the poor and on ethnic minorities. A central goal of our project is to examine how the ACA expansion of Medicaid will affect diabetes diagnosis, treatment, and outcomes among the newly enrolled. We will do so using the natural experiment that results from some states expanding Medicaid and others choosing not to do so; combined with detailed, longitudinal electronic health records (EHR) for a large population of 9 million persons in four expansion and five non-expansion states. The combination of a multi-state shock to Medicaid eligibility and large-scale access to EHR is unique. Prior studies of adult Medicaid expansions have been small-scale, short-duration, lacked detailed health records, or a combination of these limitations. The best available study, the Oregon Health Insurance Experiment, compared 6,000 new Medicaid recipients to 6,000 controls in a single state, for a two-year period. The Oregon research found large increases in diabetes diagnosis and use of diabetes medication, but no significant change in blood sugar levels. Yet any effect of Medicaid on outcomes will emerge slowly. In contrast, we estimate that we will have 100-150,000 new Medicaid recipients, 5 years of pre-treatment data to establish baseline health and match treated individuals to controls, 6 years of post-treatment data in the study period (with potential followup after that), and several treated states, with different Medicaid programs, and potential to expand to additional states. This will let us study both near- and medium term outcomes, and assess whether treatment effects vary with personal characteristics or the nature of the Medicaid program. A second project goal is to innovate in causal inference methods. We will combine difference-in-differences (DiD), matching, and multiple imputation methods. We will use matching and multiple imputation methods to match control persons in non-expansion states, to match to new Medicaid enrollees in expansion states. We will use distributed lag regressions to map the emergence of any treatment effect over time. DiD, matching, and multiple imputation methods have not to our knowledge been combined in a single study. This combined design will be useful in studying other health insurance expansions.
|Effective start/end date
|9/30/15 → 9/29/21
- National Center for Chronic Disease Prevention and Health Promotion (1U18DP006120-01)
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.