Prevalence and predictors of under or overestimation sleep duration in adults: The ELSA-Brasil study

Objective: Every one-hour change in sleep duration (SDUR) has a dose-response impact on cardiovascular disease and mortality. However, self-reported SDUR may not represent the real sleep time in a significant proportion of the adult population. This study was designed to determine the frequency of potential SDUR under/overestimation when comparing subjective versus objective SDUR data. Methods: Consecutive adults from the ELSA-Brasil study collected subjective SDUR and underwent wrist actigraphy for seven days. We defined SDUR underestimation (underSDUR) and overestimation (overSDUR) when the differences between subjective and objective SDUR reached at least -1/+1-hour, respectively. We performed multinomial logistic regression analyses to identify independent factors of under- and overSDUR. Results: Data from 2,036 participants were used in the final analysis (42.7% males; age: 49±8 years). The frequency of underSDUR and overSDUR were 19.4% and 19.7%, respectively. The predictors of underSDUR included black race (OR: 1.65; 95% CI: 1.79–3.93), mixed-race (OR: 1.69; 95% CI: 1.20–2.38); daytime sleepiness (OR: 1.37; 95% CI: 1.05–1.80); longer objective SDUR (OR: 3.00; 95% CI: 2.54–3.56); longer wake time after sleep onset time, WASO (OR: 2.19; 95% CI: 1.22–3.95), and moderate/severe insomnia (OR: 2.54; 95% CI: 1.78–3.63). Longer WASO (OR: 2.26; 95% CI: 1.33–3.82), and a higher number of awakenings (OR: 1.02; 95% CI: 1.00–1.03) were independently associated with overSDUR. Conclusions: We found a significant rate of SDUR under/overestimation when comparing subjective versus objective data. While underSDUR was independently associated with black/mixed race, daytime sleepiness, longer SDUR and WASO, overSDUR was specifically associated with markers of sleep fragmentation.