Evaluating Federal Policies Using Bayesian Time Series Models: Estimating the Causal Impact of the Hospital...

Area of Science:

• Health Services Research
• Biostatistics
• Public Health Policy

Background:

• Evaluating simultaneous, population-wide policies with delayed effects is challenging.
• Bayesian time series models are used for causal inference but require thorough investigation of assumptions and methods.
• Assessing the impact of interventions like the Hospital Readmissions Reduction Program (HRRP) on population health is crucial.

Purpose of the Study:

• To establish useful estimands for evaluating large-scale policies.
• To discuss and evaluate the key causal assumption for imputing potential outcomes.
• To illustrate an approach for model elicitation using pre-policy data.

Main Methods:

• Utilized Bayesian time series models for causal effect estimation.
• Developed and evaluated a method to assess untestable causal assumptions using observed data.
• Applied classic statistical techniques for model elicitation based on pre-intervention data.
• Examined the Hospital Readmissions Reduction Program (HRRP) impact on elderly mortality across US regions and time windows.

Main Results:

• The HRRP was found to increase mortality for pneumonia and acute myocardial infarction in at least one region and time frame.
• The policy likely had a detrimental effect on public health outcomes.
• Variations in impact were observed across different geographic subregions and time windows.

Conclusions:

• The HRRP's effect on population mortality requires careful consideration of causal assumptions and modeling techniques.
• The findings suggest potential negative public health consequences of the HRRP.
• Further research is needed to refine methods for evaluating large-scale, simultaneous policy interventions.