Chronotypes in the US - Influence of age and sex

Abstract

An individual's chronotype reflects how the circadian system embeds itself into the 24-h day with rhythms in physiology, cognition and behavior occurring accordingly earlier or later. In view of an increasing number of people working at unusual times and linked health and safety risks, the wide range in human chronotypes may provide opportunities to allow people to work (and sleep) at times that are in synch with their circadian physiology. We aimed at estimating the distribution of chronotypes in the US population by age and sex. Twelve years (2003-2014) of pooled diary data from the American Time Use Survey were used to calculate chronotype based on mid-point of sleep on weekends (MSFWe, n = 53,689). We observed a near-normal distribution overall and within each age group. The distribution's mean value is systematically different with age, shifting later during adolescence, showing a peak in 'lateness' at ~19 years, and shifting earlier thereafter. Men are typically later chronotypes than women before 40, but earlier types after 40. The greatest differences are observed between 15 and 25 for both sexes, equaling more than 50% of the total chronotype difference across all age groups. The variability in chronotype decreases with age, but is generally higher in males than females. This is the first study to estimate the distribution and prevalence of individual chronotypes in the US population based on a large-scale, nationally representative sample. Our finding that adolescents are on average the latest chronotypes supports delaying school start times to benefit their sleep and circadian alignment. The generally wide range in chronotypes may provide opportunities for tailored work schedules by matching external and internal time, potentially decreasing long- and short-term health and safety risks.

Fig 1. Chronotype and sleep duration.

a) Distribution of sleep duration on weekends (min. 3h, max. 14h) in the total sample of 53,689 respondents. b) Relationship between chronotype (mid-sleep on weekends, MSF^We) and sleep duration on weekends (mean values and standard errors). No significant association was observed in a univariate regression model (b = -0.0004, p > 0.05), indicating that short and long sleepers are equally frequent among early and late types. c) Chronotype distributions did not statistically differ when restricting sleep duration on weekends to 5–11 hours (n = 47,435) vs. 3–14 hours (n = 53,689). n.s. = non-significant, i.e. p > 0.05.

Fig 2. Chronotype distributions (based on weighted frequencies of mid-sleep times on weekends, MSF^We) in a) the total sample of 53,689 respondents, b) for the age groups 20–24 years (n = 2,479) and 70–74 years (n = 2,266), c) separately for females (n = 30,226) and males (n = 23,463), and d) for males (n = 1,809) and females (n = 1,698) in the age group 15–19 years.

Although bell-shaped at all ages, older age groups show leptokurtic distributions with the distribution average shifted to the left (earlier chronotypes) (panel b). Extreme early (MSF^We < 2:00) and extreme late (MSF^We > 6:00) chronotypes are more frequent in men than women (panel c). While frequencies are higher for girls than boys (15-19y) among early chronotypes (i.e., MSF^We < 3:00), they are higher for boys among late types (i.e., MSF^We > 6:00), overall resulting in a slightly advanced chronotype for girls in this age group (MSF^We_females = 4:24 ±1.97h vs. MSF^We_males = 4:31 ±1.99h, p = 0.051) (panel d).

Fig 3. Influence of age and sex on chronotype.

Although very early (based on mid-sleep on weekends, MSF^We < 2:00) and very late (MSF^We > 6:00) chronotypes are observed at all ages and for both sexes, the variability (assessed as sd = standard deviation, panel a) as well as the average of chronotype (panel b) changes non-linearly with age and sex. The inlay graph in panel b shows the number of respondents at each age by sex; note that one data point is outside the axis limits: women aged 80+ (n = 1602). c) The distinct age- and sex-pattern is observed in every year of survey (2003–2014), with a non-significant trend towards slightly earlier distribution means in later years (see inlet in panel c).