Evaluating federal policies using Bayesian time series models: estimating the causal impact of the hospital readmissions reduction program

Researchers are often faced with evaluating the effect of a policy or program that was simultaneously initiated across an entire population units at single point in time, and its effects over targeted can manifest any time period afterwards. In presence data measured Bayesian series models have been used to impute what would happened after initiated, had not taken place, order estimate causal effects. However, considerations regarding definition target estimands, underlying assumptions, plausibility such choice appropriate model thoroughly investigated. this paper, we establish useful estimands for evaluation large-scale policies. We discuss imputation missing potential outcomes relies on assumption which, even though untestable, be partially evaluated using observed data. illustrate approach evaluate key facilitate elicitation based from interval before initiation classic statistical techniques. As illustration, study Hospital Readmissions Reduction Program (HRRP), US federal intervention aiming improve health patients pneumonia, acute myocardial infraction, congestive failure admitted hospital. HRRP mortality among elderly four geographic subregions, different windows. find increased pneumonia infraction least one geographical region horizon, is likely detrimental public health.