Statistical inference for time‐to‐event data in non‐randomized cohorts with selective attrition