Generalizing Health Insurance Plan Effects on Medicaid Spending with Randomized and Observational Data

• Activity Number: 259 - SPEED: Missing Data and Causal Inference Methods, Part 2
• Type: Contributed
• Date/Time: Monday, July 29, 2019 : 3:05 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Irina Degtiar* and Francesca Dominici and Sherri Rose
• Companies: Harvard T.H. Chan School of Public Health and Harvard T.H. Chan School of Public Health and Harvard Medical School
• Keywords: generalizability; transportability; causal inference

Abstract

Randomized and observational studies each have their own strengths and limitations for estimating causal effects in a target population. Randomized data have internal validity but are often not representative of the target population. Observational data are affected by confounding bias but are often a representative sample of the target population and hence have external validity. Our goal is to estimate the causal effect of health insurance plans on cost among New York City Medicaid patients. We explore methods for generalizing inference to a target population represented by a combination of both randomized and observational data. We found that significant differences exist in baseline covariates and spending (after confounding adjustment) between the Medicaid beneficiaries that are included in the randomized and observational data, respectively. Generalizability methods can help reconcile these discrepancies by specifying a target population for which inference is desired; various methods either extend randomized trial estimates to the target population or combine estimates from randomized and observational data.