Suicide is one of the leading causes of death worldwide. Based on the data from the World Health Organization, suicide is the fourth leading cause of death in adolescents and young people (15–29 years) (1). Suicidal ideation (SI) is a strong predictor of suicide attempts and completed suicide (2). In college students, the 1-year incidence of SI ranges from 3.00 to 8.53% worldwide (3, 4) and the prevalence of SI has increased by 138% in 10 years (2007–2017) (5). It is important to identify risk factors of SI within this vulnerable age group, which is critical to its prevention and early intervention.

A total of 381 subjects were followed up during the COVID-19 pandemic, a follow-up rate of 78%, and the median follow-up duration was 9 months. There were no significant differences in baseline age, sex and BMI between the total subjects and the subjects who were followed up (all p-value ≥0.05). Among the 333 subjects who did not have SI in the pre-COVID-19 pandemic period, 19 subjects (5.71%) reported incident SI during the COVID-19 pandemic. Table 1 presents the demographic, sleep, and mood characteristics of the entire sample pre-COVID-19 and stratified by the presence or absence of incident SI during the COVID-19 pandemic.

As shown in Table 1, pre-COVID-19 BMI [21.61 ± 3.08 kg/m² versus (vs.) 20.19 ± 2.46 kg/m², p = 0.031], bedtime (24:07±0:33 vs. 23:51±0:34, p = 0.039), the proportion of self-reported short sleep duration (84.21% vs. 48.09%, p = 0.002), eveningness chronotype (26.32% vs. 8.28%, p = 0.043), and the total scores of BAI (8.32 ± 4.52 points vs. 4.85 ± 5.49 points, p < 0.001) as well as BDI (12.21 ± 6.29 points vs. 7.33 ± 6.09 points, p = 0.001) were significantly higher in students with incident SI compared to those without incident SI during the COVID-19 period.

We found that pre-COVID-19 short sleep duration (AOR = 4.91, 95% CI = 1.16–20.74, p = 0.031) and pre-COVID-19 eveningness chronotype (AOR = 3.80, 95% CI = 1.08–13.30, p = 0.037) were associated with increased risk of incident SI during the COVID-19 pandemic, after adjusting for baseline age, bedtime, BMI, sex, and total scores of BAI and BDI (Model 1 in Table 2). In secondary analysis, when we used the changes of BAI and BDI from follow-up to baseline, instead of the total scores of baseline BAI and BDI as covariates in Model 2 (Table 2), the results were similar and significant (short sleep duration: AOR = 6.79, 95% CI = 1.41–32.74, p = 0.017; eveningness chronotype: AOR = 5.59, 95% CI = 1.43–21.79, p = 0.013). However, the morningness chronotype was not associated with incident SI (AOR = 3.28, 95% CI = 0.57–18.79, p = 0.183). As shown in Figure 2, students who had incident SI during the COVID-19 pandemic had shorter sleep duration compared to those who did not have incident SI (6.35 h vs. 6.63 h, p = 0.026). Figure 3 depicts the comparisons of the proportion of SI among the three chronotypes (Intermediateness 4.56% vs. Morningness 5.13% vs. Eveningness 16.13%, p = 0.043). Moreover, we examined the association between the baseline score of PSQI, instead of short sleep duration, and incident SI. PSQI was not associated with an increased risk of incident SI during the COVID-19 pandemic (AOR = 1.07, 95% CI = 0.84–1.36, p = 0.585).

Supplementary Tables S1, S2 present the sociodemographic, sleep and mood characteristics of students stratified by sleep duration and chronotypes, respectively. Students with short sleep duration had higher proportion of pre-bedtime PESM-E (59.88% vs. 47.59%, p = 0.025), later bedtime (24:08 ± 0:32 vs. 23:36 ± 0:28, p < 0.001), and shorter TIB (6.98 ± 0.66 h vs. 7.49 ± 0.56 h, p < 0.001), whereas no significant difference was found in wake up time (07:07 ± 0:22 vs. 07:05 ± 0:23, p = 0.900) or sleep onset latency (8.98% vs. 8.43%, p = 0.859) (Supplementary Table S1). Furthermore, the proportion of pre-bedtime PESM-E (Intermediateness 53.23% vs. Morningness 38.46% vs. Eveningness 77.42%, p = 0.005), bedtime (Intermediateness 23:53 ± 0:34 vs. Morningness 23:29 ± 0:26 vs. Eveningness 24:08 ± 0:28, p < 0.001), and TIB (Intermediateness 7.23 ± 0.66 h vs. Morningness 7.52 ± 0.62 h vs. Eveningness 6.99 ± 0.59 h, p = 0.001) were significantly different among three chronotypes. Wake up time (Intermediateness 07:07 ± 0:23 vs. Morningness 07:00 ± 0:21 vs. Eveningness 07:08 ± 0:22, p = 0.100), the proportion of short sleep duration (Intermediateness 52.09% vs. Morningness 33.33% vs. Eveningness 54.84%, p = 0.079) and SOL (Intermediateness 8.75% vs. Morningness 2.56% vs. Eveningness 16.13%, p = 0.151) showed no statistical difference among different chronotypes (Supplementary Table S2).

In adolescents and young adults, suicide is one of the leading causes of death (1). Sleep disturbance has been known as a potential risk factor for suicide (36). The prevalence of SI and sleep disturbance both increased during the COVID-19 pandemic (22–24, 37, 38). This is the first study to examine the association between the pre-COVID-19 short sleep duration, eveningness chronotype and incident SI during the COVID-19 pandemic in medical students. Our findings suggest that pre-COVID-19 short sleep duration and eveningness chronotype are significantly associated with an increased risk of incident SI during the COVID-19 pandemic in medical students.

Psychological problems associated with COVID-19, including suicide, have drawn much attention from the public, as well as the medical community (39). Several studies have suggested that incident SI is associated with short sleep duration during the COVID-19 pandemic in adolescents (15, 40). However, no longitudinal study has examined the association between pre-COVID-19 sleep duration and incident SI during the COVID-19 period. In the current study, our findings suggested that medical students who slept less than 7 h per night had a higher risk of incident SI during the COVID-19 pandemic. In our study, the students with short sleep duration had later bedtime and shorter TIB, but not more difficulty in getting or maintaining. Furthermore, we found that the students with short sleep duration spent longer time on electronic screen media entertainment before bedtime. It needs to be noted that each medical student needs to attend the fixed early morning class at 8:00 a.m. in our study. Taken together, it appears that short sleep duration in our sample is mainly due to lake of sleep opportunities, which is consistent with a previous study conducted in China (41). In our study, electronic screen media entertainment was one of the leading reasons for delayed bedtime for medical students. All this evidence suggests that prolonging sleep by increasing sleep opportunities (e.g., advance bedtime or delay school time) might be helpful (42, 43). Moreover, our finding of short sleep duration associated with incident SI is consistent with previous longitudinal studies that were not conducted during the COVID-19 pandemic (11, 13). Taken together, it appears that self-reported short sleep duration is a predictor for future SI, no matter whether in common situations or under major health events.

The mechanisms for short sleep duration with increased risk of SI remain unclear. First, it might be associated with inflammation. Previous findings have suggested that insufficient sleep is associated with inflammation (44−46). Inflammation increases the risk of depression, which is one of the strong predictors of SI (47). Second, it might be associated with decreased serotonergic activity. It has been reported that short sleep is associated with decreased serotonergic activity. Lower concentrations of serotonin have been reported to be associated with suicide (48). However, we do not have biological data enabling us to examine the underlying mechanisms for the association of short sleep duration with a higher risk of SI. Future studies should examine the mechanisms linking the association between short sleep duration and SI.

An individual’s preference for the time of sleeping and being awake is called their chronotype. Those with an eveningness chronotype are likely to go to bed at midnight, wake later, and show high efficiency and activity in the afternoon compared to those with the morningness chronotype. In the current study, we found that students with the eveningness chronotype have a higher risk of incident SI even after adjusting for depression and short sleep duration. In our study, we found eveningness chronotype, independent of short sleep duration, predicted the incidence of SI during COVID-19. This finding is consistent with some studies that were conducted in patients with depressive disorder (49, 50). However, our finding is inconsistent with a previous longitudinal study that was conducted in general adolescents (20). The inconsistent findings might be due to that study investigated individuals with a persistent eveningness chronotype, while our study used baseline eveningness chronotype as a predictor of incident SI. Furthermore, it has been reported that both suicidality and eveningness have genetic variations in the ARNTL and CLOCK genes (51). It appears that suicidality may share the same genetic variations with eveningness.

It needs to be noted that the findings of this study were based on self-reported sleep duration. Self-reported sleep duration estimates are about 0.5–1 h longer than objective measures in general population (52–54). Self-reported sleep duration does not only reflect sleep duration per se, but is also associated with other factors, such as age, sex, socioeconomic status, education, race, sleep quality, physical health problems (e.g., pain and obesity) and mental health problems (e.g., depression and emotional stress) (55). In the current study, when we further controlled the potential confounders that may affect self-reported sleep duration, the results remained similar and significant.

Our findings suggest that pre-COVID-19 short sleep duration and the eveningness chronotype are risk factors for incident SI during the COVID-19 pandemic. In our study, we found self-reported short sleep duration was caused by the later bedtime rather than insomnia symptoms. Sleep is a modifiable factor that can reduce suicide risk and be targeted through public health prevention. Prolonged sleep has been shown to be associated with a decreased risk of SI (56). In Colorado, students’ sleep duration increased by 4.6 min for the school start time delaying 15 min, and the risk of suicidal attempts was also decreased (57). It appears that earlier bedtime for prolonging sleep duration or more flexible time schedules for eveningness may help to decrease incident SI in medical students. More multi-center longitudinal, intervention and experimental studies are needed to further confirm the association between short sleep duration, eveningness chronotype and SI, as well as their underlying mechanisms.

Strengths of this study include the prospective design and careful control for multiple potential confounders that may affect the association between sleep duration, chronotype and SI. Several limitations should be acknowledged and taken into account when interpreting our results. First, given that different majors may affect the understanding of symptoms of depression and anxiety, and be affected differently by the COVID-19 pandemic (58), we only included medical students. Thus, our results cannot be generalized to other populations. The association of sleep duration, chronotype and risk of SI in the general population needs to be examined in future studies. Second, our findings were based on self-report questionnaires, which may lead to recall bias. However, it has been reported that the questionnaires used in this study are valid and reliable for the Chinese population.

Our findings suggest that pre-COVID-19 self-reported short sleep duration and eveningness chronotype predict incident SI during the COVID-19 pandemic in medical students. Prolonging sleep duration by earlier bedtime or flexible school time may help to decrease SI in medical students during major public health crises. However, self-reported sleep might not fully reflect objective sleep states and we only included medical students in one college, restricting our findings to generalize to other populations. Future multi-center studies using objective sleep measurements are needed in the general population.

The datasets used and/or analyzed in the current study are available from the corresponding author on reasonable request.

The studies involving humans were approved by Research Ethics Board of Shantou University Mental Health Center. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.

DZ: Conceptualization, Data curation, Formal analysis, Methodology, Writing – original draft, Writing – review & editing. QQ: Conceptualization, Supervision, Writing – review & editing. YP: Conceptualization, Data curation, Investigation, Writing – review & editing. HZ: Conceptualization, Data curation, Investigation, Writing – review & editing. YH: Conceptualization, Data curation, Investigation, Writing – review & editing. HW: Conceptualization, Data curation, Investigation, Writing – review & editing. QT: Conceptualization, Data curation, Investigation, Writing – review & editing. YL: Conceptualization, Funding acquisition, Supervision, Writing – review & editing.