Designing Cluster RCT of Health System Interventions: with PCORI Opioid Trial as an Example

의료 체계 인터벤션에 대한 군집 무작위 배정 시험의 설계: PCORI Opioid Trial을 중심으로.

There are many issues with designing cluster RCT (randomized controlled trial) of health system interventions. In particular, determining the effect of health system interventions can be challenging due to the ongoing change in the outcome distribution resulting from the events or policy changes external to the study. We discuss these issues using an example trial (Provider-Targeted Behavioral Interventions to Prevent Unsafe Opioid Prescribing for Acute Pain in Primary Care) that was recently funded by PCORI (Patient-Centered Outcomes Research Institute). We introduce a piecewise logistic regression model to estimate the effect of the health system intervention on binary outcome given the background change in the outcome distribution. We conducted simulations to estimate the power of the piecewise logistic regression model with different effect sizes and sample sizes. Investigators can use this simulation result to design cluster RCTs of health system interventions.