This project accompanies the manuscript Designing a quasi-experiment to study the clinical impact of adaptive risk prediction models. Advances in clinical risk prediction have made it possible to target interventions based on patient-level risk, but evaluating these systems using randomized trials is often impractical. Existing quasi-experimental designs rarely allow decision rules—such as risk thresholds or prediction models—to adapt over time, even though such adaptation is common in practice.
To address this gap, we introduce an adaptive regression discontinuity (RD) framework and demonstrate its properties using simulation. In the clinical context of cardiovascular risk and lipid-lowering therapy, this demo illustrates how evolving treatment rules can be evaluated using quasi-experimental principles.