| dc.description.abstract | This dissertation details the methodology and application of analytical methods using observational data to assess the impact of health policies. We first focus on the Difference-in-differences (DID) method and its extensions, particularly in the context of staggered treatment adoption over multiple years. We describe these concepts within the context of Medicaid expansion and retention in care among people living with HIV (PWH) in the United States. We highlight the identification and estimation of the average treatment effect among the treated, emphasizing the necessary assumptions for valid estimation. We then introduce an extension of the DID method capable of estimating average, quantile, probability, and Mann-Whitney treatment effects among the treated under a single approach and a universal parallel trends assumption. Our approach uses a semi-parametric cumulative probability model (CPM) to handle complicated, often difficult-to-model outcome distributions. We demonstrate our approach with a simulation study and an application to Medicaid expansion and CD4 cell count at enrollment into care among PWH in the United States. We then shift away from DID and focus on a real-world application to assess the impact of COVID-19-related stay-at-home orders on the reporting of child maltreatment and whether this impact was modified by sociodemographic characteristics. We find that the numbers and rates of reporting after versus before the stay-at-home orders vary by county-level poverty, unemployment, median annual household income, health insurance coverage, and education. These results offer insights for policymakers on how pandemic-related policies may have varied effects across different sociodemographic groups. | |
