To study the impact of the Affordable Care Act (ACA) on adults with depression, we utilized the Medical Expenditure Panel Survey (MEPS) for the years 2010-2019. MEPS is a nationally representative survey that provides a source of data on the cost, use, and insurance coverage of healthcare across the United States. Our analysis involved a comprehensive data preparation phase followed by a rigorous statistical analysis.
The initial data processing was conducted using SAS. This phase involved systematically importing and integrating multiple MEPS public use files—including the Full-Year Consolidated and Medical Conditions files—for each year from 2010 to 2019. A crucial step was identifying the study cohort of adults with depression by screening the Medical Conditions files using the relevant ICD-9 codes for years prior to 2016 and ICD-10 codes thereafter. To create a single, cohesive dataset for longitudinal analysis, the yearly files were harmonized by standardizing variable names and formats before being stacked. This process resulted in a clean, analysis-ready panel dataset that forms the foundation for all subsequent statistical work.
The statistical analysis was performed in Stata, employing a quasi-experimental Difference-in-Differences (DID) research design to estimate the causal impact of the ACA's Medicaid expansion. The script begins by setting up the complex survey design of MEPS using svyset to ensure that all estimates are nationally representative and that standard errors are correctly calculated. Key interaction terms for the DID model and a Triple-Differences (DDD) model—used to examine heterogeneous effects across racial and ethnic subgroups—were generated. The analysis includes a full suite of descriptive statistics and multiple advanced regression models for key outcomes like total healthcare expenditures, office-based spending, and prescription drug costs. To ensure robustness, Ordinary Least Squares (OLS), Generalized Linear Models (GLM) with a gamma distribution, and two-part models were estimated, accounting for the common challenges of modeling healthcare cost data.