Do Night Owls Actually Handle Night Shifts Better?

The premise behind chronotype rostering is that late types belong on nights. The largest recent measurement of shift preference does not support that in its strong form. In the Nightingale cohort of 37,731 Dutch female nurses, definite evening types rated night shifts 3.7 out of 7, below their own rating of evening shifts at 5.4 (de Bruijn et al., 2025). Definite evening types were the only group whose top-rated shift was not the day shift, and what they preferred was the evening shift, not the night.

The ranking across chronotypes holds in the parts of the paper that do not disagree with themselves. Compared with intermediate types, definite evening types had 0.17 times the odds of a one-point-higher day-shift preference score (95% CI 0.16 to 0.18) and 0.22 times the odds on morning shifts (95% CI 0.21 to 0.24), while definite morning types had 2.17 times the odds on day shifts (95% CI 2.02 to 2.32) (de Bruijn et al., 2025). Those figures point the same way as the means: chronotype sorts what people say they want, and it sorts it most sharply on the shifts that start early.

One odds ratio is contested inside the source itself. The abstract of de Bruijn et al. (2025) gives definite evening types 2.68 times the odds of a one-point-higher night-shift preference score (95% CI 2.48 to 2.90) and 2.20 times the odds on evening shifts (95% CI 2.03 to 2.38), while the Results section attaches 2.20 to night shifts instead, in the same sentence that puts definite morning types at 0.58 on nights (95% CI 0.54 to 0.62). The paper never says which pairing is the error, so both are reported here rather than one being quietly preferred. Whichever holds, the number is odds of a higher liking score on a questionnaire, not odds of being assigned to night work, so writing that evening types are twice as likely to work nights would be a different and unsupported claim.

The sleep-problem pattern is U-shaped rather than a straight line from lark to owl. With intermediate types as the reference, sleep problems in 2017 were more frequent among definite evening types (IRR 1.25, 95% CI 1.20 to 1.30) and also among definite morning types (IRR 1.11, 95% CI 1.07 to 1.15) (de Bruijn et al., 2025). Intermediates reported the fewest. A reading in which evening types are the poor sleepers and morning types the good ones is not what this cohort shows, and it points at the same conclusion the factory intervention reached: the extremes at both ends are the group worth protecting.

Exposure ran opposite to preference in places. Recent shift work including nights between 2011 and 2017 was reported by 22.2% of definite morning types and 44.1% of definite evening types, so some sorting already happens informally (de Bruijn et al., 2025). Substantial mismatch survives it. Among the nurses who rated night shifts at 2 or below on the 7 point scale, a group making up 63% of those answering the item, 20% worked nights during the period.

The population limits are not cosmetic. These are female Dutch nurses with a median age of 55.7 years, 65.6% of them postmenopausal, reporting their own sleep timing and sleep problems. Nurses who could not tolerate nights may already have left, which would flatten the differences that remain visible. The authors conclude only that future studies should consider chronotype as an effect modifier. Nothing in the design can show that chronotype caused the sleep problems, or that reassigning people would remove them.

Self-reported chronotype carries a circularity problem, because how someone answers can reflect the roster they already work. Akimova et al. (2023) tried to sidestep that in 53,211 UK Biobank workers by using a polygenic score for eveningness, which does not shift with the schedule. Note that this is a genetic propensity estimate, not a person telling you they are a night owl, and it is not something any scheduler can measure. Workers who always worked nights reported 12 minutes 49 seconds less sleep per night than those who never or rarely did (95% CI -17:01 to -8:36, p < 0.001), a 3.47% reduction the abstract rounds to 13 minutes.

The chronotype signal appeared only in interaction with night work. Among workers who never or rarely worked nights, a one standard deviation higher eveningness score was associated with no difference in sleep duration (0:00, 95% CI -0:29 to 0:29). Among those who always worked nights, the same one standard deviation was associated with 3 minutes 37 seconds more sleep per night (95% CI 0:10 to 7:04), which the authors describe as recovering 28% of the night-work sleep penalty. Genetic eveningness showed no association with sleep duration among people not working nights.

Three limits should arrive before anyone acts on that. The always-nights group is 1,295 people, 2.43% of the sample, and the main interaction has a lower confidence bound of 10 seconds, which is a whisker from zero. The larger interaction of 11 minutes 17 seconds (95% CI 5:50 to 16:44) belongs only to workers logging more than 44 hours a week; the two shorter-hours strata were not significant and one pointed the other way. When sleep was measured by wrist accelerometer instead of self-report, the trend looked similar but the interaction no longer reached significance. The analysis is cross-sectional, and the authors state they cannot tell whether chronotype moderates the penalty or instead reflects who copes well enough to stay on nights.

The only intervention evidence here is also the thinnest. Vetter et al. (2015) ran a controlled field intervention at a single German factory. The chronotype-adjusted schedule removed the most strenuous shifts for extreme chronotypes, meaning morning shifts for late types and night shifts for early types, across a rotation of morning (06:00 to 14:00), evening (14:00 to 22:00) and night (22:00 to 06:00) work. Intermediate chronotypes in the middle two quartiles served as the control group and kept all three shift types, with two of their twelve strenuous shifts per month replaced by evening shifts.

The schedule change was followed by a significant increase in self-reported sleep duration and quality among extreme chronotypes, higher wellbeing ratings on workdays, a 1 hour reduction in overall social jetlag, no significant change in stress levels, and greater satisfaction with leisure time among early types only (Vetter et al., 2015). That 1 hour is the only number this review can quote, and both the figure and the schedule description above were confirmed against the published abstract rather than the full text, which was not obtainable, so no sample size or per-outcome effect size is available. The stress result is reported only as a non-significant comparison, with no test statistic, sample size or confidence interval available to judge how much power it had, and a non-significant difference at one small site is not the same as showing the schedule left stress untouched. Allocation was not randomized, the setting is one industrial employer, and the finding is now eleven years old. Call it promising and suggestive, not established.

None of the three studies measured a health, safety, or error outcome. Between them they measured sleep duration, sleep-problem scores, social jetlag, wellbeing ratings and shift preference. The intervention authors say the long-term health and economic effects still need to be established by future work. Chronotype rostering is therefore defensible today as a fairer way to hand out the least wanted shifts and a plausible route to better sleep at the extremes, not as a health or safety program.