Sampling strategies for selecting general population comparison cohorts

Abstract

When index persons were aged >50 years, ie, had high outcome risks, sampling 5:1 without replacement introduced bias. Sampling with replacement or 1:1 did not introduce bias.