Reliability of retrospective self-report of shift work exposure in a large occupational cohort

Objectives: To examine agreement between baseline and retrospective assessments of shift work exposure and test whether this differs in cancer and cardiometabolic disease patients. Methods: A subgroup of the Airwave Health Monitoring study, a cohort of police employees, self-reported shift work status at study entry and retrospectively at follow-up. Agreement statistics were computed overall and stratified by subsequent cancer and cardiometabolic disease status. Adjusted logistic regression tested whether discordant reporting differed by disease status. Results: Overall agreement was 86% (95% CI = 84-88%; κ = 0.72), with no evidence of asymmetric misclassification (McNemar's p = 0.32). Agreement was similar among participants who later developed cancer (κ = 0.76) but lower, although imprecisely estimated, among those who developed cardiometabolic disease (κ = 0.56). Adjusted odds of discordant reporting did not differ significantly by cancer (OR = 0.96, 95% CI= 0.39-2.05) or cardiometabolic disease status (OR = 1.85, 95% CI = 0.81-3.87). Conclusions: Retrospective shift work reports showed good overall reliability, with no evidence of asymmetric misclassification, though reliability varied somewhat by disease outcome. These findings provide an empirical basis for future studies using retrospective shift work exposure in the AHMS and similar cohorts.