Agreement between Self-Reported and Actigraphy-Derived Sleep Duration in Rural South African Adults

Background Self-reported sleep duration is widely used in epidemiological research, but validation against objective measures remains limited in African settings. This study assessed agreement between self-reported and actigraphy-derived sleep duration in rural South African adults. Methods A convenience sample of 145 adults aged ≥40 years from the Dikgale Health and Demographic Surveillance System, Limpopo Province, was included. Self-reported weekday and weekend sleep duration was calculated from usual bed and wake times. Objective sleep duration was measured using wrist-worn ActiGraph wGT3X-BT accelerometers for nine days. Agreement was evaluated using Bland–Altman analysis, intraclass correlation coefficients, Cohen’s kappa, diagnostic accuracy indices, and robust regression. Results Participants were 72% female, with mean age 52.6 years. Self-reported sleep exceeded actigraphy-derived sleep on weekdays, weekends, and overall. Mean self-reported sleep was 9.0 h on weekdays and 9.9 h on weekends, compared with approximately 7.6 h by actigraphy. Mean overestimation was 1.40 h on weekdays, 2.23 h on weekends, and 1.63 h overall. Agreement was poor, with individual intraclass correlation coefficients of 0.14, −0.02, and 0.05, respectively. Self-report classified 52.4% as long sleepers compared with 15.2% by actigraphy, while objectively short sleep was common but rarely identified by self-report. Greater actigraphy-derived sleep duration was associated with larger discrepancy between methods. Conclusions Self-reported sleep duration substantially overestimated actigraphy-derived sleep and poorly classified short and long sleep. These findings indicate potential exposure misclassification in South African sleep epidemiology and support cautious interpretation of questionnaire-derived sleep duration.