Circadian rhythm types and shift work demands shape sleep quality and depressive symptoms in shift-working nurses

Atypical work schedules are increasingly common globally, and shift work is known to disrupt sleep–wake patterns and circadian rhythms, elevating health risks. Nurses, required to provide continuous patient care, face substantial shift-related burdens and report high rates of poor sleep quality and depressive symptoms. Poor sleep is linked to impaired nurse well-being, performance, and patient safety. The Job Demands–Resources (JD-R) theory posits that high job demands can impair health, whereas personal resources may buffer adverse effects. Circadian rhythm types (flexibility–rigidity and languidness–vigorousness) reflect individual adaptability to shift work and may function as personal resources shaping tolerance to shift demands. Prior research suggests circadian traits relate to sleep outcomes and possibly psychological health, but studies often rely on self-reports and rarely integrate objective work demand data. This study aims to integrate objective shift work demand indicators with self-reported circadian types to (1) identify predictors and moderators of sleep quality and depressive symptoms, and (2) examine combined and dose–response effects, informing individualized scheduling and health interventions for shift-working nurses.

Prior studies document high prevalence of sleep problems and mood symptoms among shift nurses, with 41–60% reporting poor sleep quality and over half showing depressive or anxiety symptoms in some cohorts. The JD-R framework links high job demands (e.g., irregular night shifts, heavy workloads) to health impairment processes, and evidence shows associations between workloads and sleep disturbances, depression, and anxiety. Circadian typology research indicates that greater languidness increases vulnerability to shift work disorder and poor sleep, while flexibility may relate to better sleep and lower shift work disorder risk. Links between circadian traits and depression are less established, though chronotype and circadian stability have been associated with depressive symptom severity, and circadian traits may moderate stress–sleep relationships. Existing work frequently relies on self-reported exposures; objective shift demand measurement and analysis of threshold or dose–response patterns remain limited, motivating the present study.

Design and setting: Cross-sectional study conducted from May 1, 2024, to May 31, 2025, at a tertiary general teaching hospital in Guangxi, China. Recruitment occurred during routine physical examinations at the hospital’s health examination center. Ethical approval: Ethics Committee of the First Affiliated Hospital of Guangxi Medical University (No.: 2024-S711-01); written informed consent obtained. Participants: Inclusion criteria: age 20–45 years; full-time registered nurses; ≥6 months clinical experience; worked in a fixed ward and ≥1 rotating shift/month over the past 6 months. Exclusion: major illnesses (e.g., cancer, cardiovascular/cerebrovascular disease, diabetes), major surgery, major psychiatric disorders, sickness absence ≥3 days, major family life events in past 6 months; pregnancy or breastfeeding in past 6 months; time-zone travel in past month; unavailable shift work data. Measures: Self-report questionnaire captured demographics (gender, age, ethnicity, marital status, children, education, work years, professional title, years of shift work, shift model) and instruments: - Circadian Type Inventory (CTI): FR (5 items) and LV (6 items) on 5-point Likert scale; higher FR = greater flexibility; higher LV = greater languidness (vulnerability to drowsiness/sleep loss). Chinese version reliability: Cronbach’s α = 0.843 (FR), 0.719 (LV) in this study. - Sleep quality: Pittsburgh Sleep Quality Index (PSQI), total 0–21, higher = poorer sleep; Chinese version α ~0.82–0.83; insomnia screening cutoff >7; α = 0.748 in this study. - Depressive symptoms: Patient Health Questionnaire-9 (PHQ-9), total 0–27; cutoff ≥10 for probable depression; α = 0.871 in this study. Objective shift work demands (past 4 weeks) extracted from nursing management system: - Work quantity: total evening shift count; total night shift count; total shift count (evening + night); shift work hours (cumulative evening + night hours); day work hours; total work hours (day + shift). Definitions: shift work between 18:00–08:00; evening shift ≥4 consecutive hours between 18:00–00:00; night shift ≥4 consecutive hours between 00:00–08:00; long night (18:00–08:00 with 3-h nap) counted as night shift. - Work intensity: workload and workload exposure for day and shift work. Workload = expected nurse-to-patient ratio (NPR) / actual NPR. Actual NPR computed as active primary nurses to assigned patients; expected NPR derived from unit patient severity (CMI-based) using exponential function: Expected NPR = 0.1154 × exp(1.0791 × patient severity). CMI normalized to 0–2. Workload exposure = workload × corresponding work hours. Due to missing data, participants from emergency (n=7), anesthesiology (n=1), and hemodialysis (n=2) were excluded from work intensity analyses. Statistical analysis: Python-based. Two-tailed p<0.05 considered significant. Descriptive statistics used mean±SD or median (IQR) and counts (%). Univariate analyses: Mann–Whitney U or Kruskal–Wallis H for group differences in PSQI, PHQ-9, FR, LV; Spearman correlations among continuous variables. GLMs for PSQI and PHQ-9 outcomes: preprocessing included outlier removal, missing handling, Yeo–Johnson transformation, and z-standardization. Predictors included (I) variables significant in univariate analyses; (II) key shift demand variables (total night shift count, shift work hours, shift workload exposure; total shift count excluded for multicollinearity) and circadian types (FR, LV); (III) confounders (age, sex, day workload exposure; years of work excluded due to collinearity with age). Three candidate moderator models (night shift count, shift work hours, shift workload exposure) were compared via AIC, BIC, pseudo-R²; diagnostics included residuals and VIF<5. Nonlinear analysis: GAMs with smooth terms for continuous predictors and factor terms for categorical variables; nonlinearity assessed via EDoF>1 and p<0.05. Piecewise linear regression identified breakpoints for shift demand variables with significant nonlinearity; segment slopes compared. Monte Carlo simulations: Predictive functions derived from best GLMs; 100-day iterative simulation with exponentially weighted smoothing (α=0.2; sensitivity analyses varied α). Two strategies: (1) Population-based: 1,000 virtual individuals bootstrapped from sample distributions to estimate population mean ± SD trajectories. (2) Scenario-based groups by circadian adaptability profiles: Moderate (FR & LV between 25th–75th percentiles), High (FR ≥75th, LV ≤25th), Low (FR ≤25th, LV ≥75th). Baseline shift demand scenario: total night shifts=4, shift work hours=44, shift workload exposure median=147; demands varied to assess dose–response patterns.

Sample: 288 nurses analyzed (after excluding 24); 10.42% male; median age 33.5 (24–44). Shift models: 65.28% day-night (DN), 34.72% day–evening–night (APN). Median PSQI=8 (IQR 4; range 1–17); median PHQ-9=7 (IQR 5; range 0–22). Insomnia prevalence (PSQI>7): 51.39%; probable depression (PHQ-9≥10): 24.31%; both: 20.14%. Univariate correlations: Poorer sleep quality correlated with higher depressive symptoms (0.560, p<0.001), greater languidness (0.356, p<0.001), and lower BMI (−0.122, p=0.038). Depressive symptoms correlated negatively with flexibility (−0.179, p=0.002) and positively with languidness (0.412, p<0.001). GLMs (all VIFs 1.05–3.76; acceptable diagnostics): - Sleep quality (best model with shift work hours moderator: AIC 1377.37; BIC 1428.05; pseudo-R² 0.4403): Significant predictors: PHQ-9 (β=0.245, 95% CI 0.195–0.295), shift work hours (β=0.093, 0.004–0.182), LV (β=0.065, 0.014–0.116), BMI (β=−0.056, −0.105 to −0.007). Marginal interactions: PHQ-9×LV (p=0.090), PHQ-9×shift hours (p=0.082). - Depressive symptoms (best model with shift work hours moderator: AIC 1657.89; BIC 1712.20; pseudo-R² 0.4805): Significant predictors: PSQI (β=0.314, 0.246–0.381), FR (β=−0.129, −0.190 to −0.069), LV (β=0.159, 0.090–0.227), and PSQI×FR interaction (β=0.091, 0.035–0.146). LV×shift hours showed marginal significance (β=0.069, p=0.049). Night shift count and shift workload exposure were not significant in either model. Nonlinear analyses: GAM for PSQI explained 40.5% variance; shift work hours had a significant nonlinear association with PSQI (EDoF=1.6, p=0.010); night shift count and shift workload exposure showed no nonlinear effects. GAM for PHQ-9 (pseudo-R²=0.441) found no nonlinear effects of shift demand variables. Piecewise regression identified a breakpoint at ~24 shift work hours per 4 weeks: below 24 h, a marginal negative slope (β=−0.221, p=0.051); above 24 h, a significant positive slope (β=0.031, p=0.039), indicating worsening sleep with more shift hours. Simulations: Population-based simulation indicated the mean trajectory reached the insomnia threshold around day 30 while psychological status remained relatively good. Scenario simulations under a typical 4-week demand (4 night shifts; shift hours=44; workload exposure=147) revealed distinct dose–response patterns by circadian adaptability: Moderate adaptability sustained healthy sleep and mood at baseline but reached insomnia threshold when demands increased by ~50%. High adaptability maintained favorable outcomes even when demands doubled. Low adaptability showed rapid deterioration in sleep and depressive symptoms under typical demands, requiring ~75% reduction in demands to maintain good sleep; tolerable conditions approximated one night shift per month.

Findings reveal that circadian rhythm types (FR/LV) and shift work hours exert direct and moderating effects on sleep quality and depressive symptoms among rotating-shift nurses, aligning with the JD-R framework wherein high demands impair health and personal resources buffer impacts. The bidirectional link between sleep and depression was confirmed, with poorer sleep exerting a stronger association with depressive symptoms than vice versa, underscoring sleep disturbances as a key mechanism driving mood symptoms in shift workers. Languidness consistently predicted poorer sleep and higher depression, suggesting greater vulnerability to sleep loss and drowsiness, whereas flexibility predicted lower depression and moderated the sleep–depression association, though its protection diminished with worsening sleep. Objective shift demand data clarified exposure–response relationships: total shift hours, not night shift counts per se, were salient, with a threshold around 24 shift hours/4 weeks beyond which sleep quality deteriorated progressively. Simulations demonstrated circadian profile–specific dose–response patterns: high adaptability buffered adverse effects even under heavier demands; moderate adaptability tolerated typical demands but was sensitive to increases; low adaptability rapidly decompensated, indicating a need for substantially reduced demands. Collectively, these results address the research question by delineating how circadian typology and measurable shift demand exposures jointly shape sleep and mood, supporting personalized, circadian-informed scheduling and targeted interventions in nursing workforce management.

Circadian rhythm types and objectively measured shift work hours jointly influence sleep quality and depressive symptoms among shift-working nurses. A potential threshold was identified: cumulative shift work hours exceeding ~24 hours within four weeks were associated with worsening sleep quality. Distinct dose–response patterns emerged across circadian adaptability profiles, indicating that individualized, circadian-informed scheduling and interventions may improve sleep and mental health in shift nurses. Future research should employ longitudinal and multi-center designs, extend observation windows, integrate objective sleep and mood assessments, and experimentally test personalized scheduling strategies based on circadian profiles and quantified shift exposures.

- Cross-sectional design limits causal inference; longitudinal or cohort studies are needed. - Single-center sample with predominantly female participants (89.58%) may limit generalizability, particularly to male shift workers. - No baseline health check; pre-existing depressive symptoms might have been overlooked despite exclusion of major psychiatric disorders. - Work demands assessed over only four weeks; longer observation windows (e.g., 12–24 weeks) are needed to capture cumulative effects. - Simulations assumed relatively stable shift patterns and did not model daily scheduling variability or detailed accumulation/recovery dynamics. - Sleep and depressive symptoms were self-reported (PSQI, PHQ-9), potentially introducing recall/reporting bias; future studies should include objective assessments.