Participants
The study participants were all medical students at a medical university in Japan. In 2018 and 2019, a beneficial sleeping habits survey questionnaire was distributed to each participant and collected during their annual physical examination in April. The physical examination was conducted based on the School Health and Safety Act in Japan [18]. The participants were informed of this study's purpose and methods, and were asked to consent to participation by ticking a box in the survey. The consent of the participants who missed ticking the box in the questionnaire was confirmed later by email. In 2020, the survey was conducted online in July because the students could not visit the university campus amid the COVID-19 pandemic.
Usually, the students from the first to the fourth year attend around five or six classes a day, starting at 8:50 in the morning. Conversely, the fifth- and sixth-year students participate in clinical practice in the university hospital or community-based hospitals from the morning as per each hospital's schedule. In July 2020, a quarter of the students in the first grade went to school in rotation, and the remaining first-year students and the second- and third-year students were instructed to take remote classes. The students reported living at their homes or their parents' homes. The schedule for the remote classes was the same as usual. As the fifth- and sixth-year students could not go to community-based hospitals, they attended clinical practice at the university hospital, thereby reducing the time and sharing the opportunity.
Measures
A few second-year students created the questionnaire used in this study as part of a practical assignment for their curriculum in 2016—it was adapted from a questionnaire developed by Shirakawa [19]. The questions were designed to collect information about sleeping habits and living situations, including the following: the hours at which one goes to bed and wakes up during weekdays and weekends; the time spent lying awake in bed before falling asleep and after waking up; how comfortable one feels with the sleep duration; frequency of naps; a self-assessment of one's sleeping depth; one's ease in waking up from bed; what one does before going to bed; the time spent watching TV or using a mobile phone; how one gets up; how often one wakes up early in the morning; arousal frequency during sleep; frequency of going to the bathroom during sleep time; whether there is a need to use hypnotics or minor tranquilizers for sleeping; how frequently sleep paralysis occurs; how frequently one dreams, snores, experiences sleep apnea; whether one performs any extracurricular activity, has a part-time job; whether one lives alone or with family; and whether one smokes or drinks. The Aichi Medical University Hospital Ethics Board approved this study's design (approval number: 2018-M005, 2020-M015).
Statistical Analysis
Data were analyzed for 644 students in 2018, 649 students in 2019, and 392 students in 2020. Among them, 199 observations for weekdays and 181 for weekends, were treated as missing data. They were as follows: (A) the bedtime or wake-up time was absent (12 observations for weekdays and 21 for weekends in 2018; 12 observations for weekdays and 18 for weekends in 2019), (B) the sleep duration was less than 240 minutes (7 observations for weekdays and 6 for weekends in 2018; 5 observations for weekdays and 4 for weekends in 2019; 6 observations for weekdays and 4 for weekends in 2020), or (C) it was more than 900 minutes, and their bedtime and wakeup time were (C-1) between 6:00 to 10:00 and 23:00 to 4:00 (30 observations for weekdays and 16 for weekends in 2018; 45 observations for weekdays and 26 for weekends in 2019; 2 observations for weekdays and 2 for weekends in 2020), (C-2) between 10:00 to 15:00 and 5:00 to 12:00 (2 observations for weekends in 2018; 66 observations for weekdays and 39 for weekends in 2019; 9 observations for weekdays and 9 for weekends in 2020), (C-3) between 22:00 to 4:00 and 18:00 to 0:00 (1 observation for weekdays and 23 for weekends in 2018; 2 observations for weekdays and 18 for weekends in 2019; 1 observation for weekdays in 2020), or other three data (between 8:30 and 8:15 on weekdays in 2020, between 22:00 and 14:00 on weekends in 2019, and between 7:45 and 6:45 on weekends in 2020), respectively. These data were considered bedtime and wake-up time transposed, mixed up 12-hour and 24-hour notations, or some sort of mistake. For the multilevel analyses, 684 participants were included, who enrolled in the university during 2018-2020. Among the responses collected, 173 observations for weekdays and 161 for weekends were treated as missing data. Such observations included the following: (A) the bedtime or wake-up time was absent (12 observations for weekdays and 21 for weekends in 2018; 6 observations for weekdays and 12 for weekends in 2019); (B) the sleep duration was less than 240 minutes (7 observations for weekdays and 6 for weekends in 2018; 3 observations for weekdays and 2 for weekends in 2019; 3 observations for weekdays and 3 for weekends in 2020); or (C) it was more than 900 minutes, and their bedtime and wakeup time were (C-1) between 6:00 to 10:00 and 23:00 to 4:00 (30 observations for weekdays and 16 for weekends in 2018; 40 observations for weekdays and 24 for weekends in 2019; 2 observations for weekdays and 2 for weekends in 2020); (C-2) between 10:00 to 15:00 and 5:00 to 12:00 (2 observations for weekends in 2018; 60 observations for weekdays and 29 for weekends in 2019; 8 observations for weekdays and 7 for weekends in 2020); or (C-3) between 22:00 to 4:00 and 18:00 to 0:00 (1 observation for weekdays and 23 for weekends in 2018; 1 observations for weekdays and 14 for weekends in 2019).
All analyses were performed using SAS 9.4 (SAS Institute Inc., Cary, NC, USA). The mean sleep duration during weekdays and weekends was compared using a paired t-test. Multilevel analyses were performed to determine whether there was any difference among each sleep duration data for weekdays and weekends across the 3 years, using the SAS PROC MIXED procedure. We did not exclude participants for whom bedtime or wake-up data were missing, as SAS PROC MIXED automatically handles missing data using restricted maximum likelihood.
We used the following model to explore the differences in the slopes for each interval of 3 years.
Sleep durationiy= ꞵ1 + ꞵ2×Genderi + ꞵ3×Gradeiy + ꞵ4×Place of stayiy + ꞵ5×Smokingiy + ꞵ6×Drinkingiy + ꞵ7×Extracurricular activitiesiy + ꞵ8×Part-time jobiy + ꞵ9×Doing something always before going to bediy + ꞵ10×Taking any hypnotics or minor tranquilizeriy + ꞵ11×Apneaiy + ꞵ12× Arousal during sleepiy + eiy
where i repsents the individual, y repsents year, ꞵ1-12 repsents parameters, and e is the error term.
The level of statistical significance was set at 5% (p<0.05).