Fig. 5: Simulated final cohorts (dark blue dots and line) with EHR-site-based policy for target points on the representativeness-coverage pareto optimal frontier (red line).

The All of Us final cohort is shown as a black dot, with black dashed lines extending its KLD(C∣∣P) and KLD(C∣∣U) values outwards. Cyan lines show the relationship between a cohort’s final representativeness, coverage, and target point. The pink line and pink-outlined point indicate the final selected target point of (0.04, 0.909).