This article quantifies how weekly work hours map onto multiple dimensions of mental health in healthcare workers inside a Chinese medical consortium, with a specific focus on “where the curve bends” rather than whether a relationship exists at all (Tang et al., 2026). Using WHO/ILO cut-points, it tests both the well-known ≥55 h/week risk boundary and the under-studied “subthreshold” band of 49–54 h/week, explicitly asking whether harm begins before the regulatory red line and whether that onset differs by subgroup (Tang et al., 2026).
Methodologically, the study is a cross-sectional online survey conducted June to July 2023 across four public hospitals (tertiary infectious disease, secondary psychiatric/infectious disease, and two primary care hospitals) in Southern China, yielding 983 analyzable responses from 1,067 initial submissions (92.1% response rate) with no item-level missingness for regression variables (Tang et al., 2026). Weekly working hours were computed from average hours/day and days/week and grouped as 35–40 (reference), 41–48, 49–54, and ≥55 h/week, while weekend overtime frequency was measured on a 5-point scale from never to always (Tang et al., 2026). Outcomes were assessed via COPSOQ III across six domains: burnout, depressive symptoms, sleep troubles, stress, somatic stress, and cognitive stress (Tang et al., 2026). The analysis strategy uses hierarchical multiple linear regression with stepwise adjustment: Model A controls individual characteristics, Model B adds organizational factors, and Model C additionally includes work characteristics (quantitative demands and work pace), with HC3 robust standard errors and multicollinearity screened via VIF < 2.5 (Tang et al., 2026).
The core empirical claim is a consistent dose–response relationship between longer weekly hours and worse scores across all six mental health outcomes (Tang et al., 2026). Beyond visual dose–response patterns, the paper reports that each one-unit increase in weekly work hours is associated with higher adjusted scores for burnout, depression, sleep troubles, stress, somatic stress, and cognitive stress (β ranging roughly 2.38 to 3.52, all P < 0.001), reinforcing that the association is not confined to a single symptom cluster (Tang et al., 2026). The threshold analysis then sharpens the message: 49 h/week marks the onset of meaningful risk signals, while ≥55 h/week corresponds to broader deterioration across domains, aligning the mental-health warning system with (but also preceding) the better-established 55-hour benchmark used in international guidance (Tang et al., 2026). In the fully adjusted Model C, ≥55 h/week remains an independent predictor of burnout (β = 4.23), stress (β = 4.25), and somatic stress (β = 5.30), indicating that these associations persist even after accounting for workload quantity and speed of work (Tang et al., 2026). A particularly decision-relevant finding is that work pace is a robust correlate across all six outcomes, suggesting that intensity and cognitive density of work can function as a pervasive upstream driver rather than a minor covariate (Tang et al., 2026). If there is a “quick joke” embedded in the statistics, it is this: “just one more hour” repeated 10 times stops being a personality trait and becomes a measurable exposure (Tang et al., 2026).
The subgroup results translate the average curve into operational risk stratification. Female healthcare workers show significant impairment already at 49–54 h/week, including elevated burnout, depression, somatic stress, and cognitive stress, while men show clearer deterioration predominantly at ≥55 h/week, especially for sleep troubles (Tang et al., 2026). Shift status produces a more complex pattern: non-shift workers show early risk signals at 49–54 h/week, whereas shift workers display a biphasic deterioration, with somatic stress rising in the 49–54 band and then widespread adverse effects at ≥55 h/week (with depression showing the largest effect and sleep troubles also highly sensitive) (Tang et al., 2026). Exposure duration further lowers tolerance, as long-term exposed workers show adverse effects already at 49–54 h/week and more consistent multi-domain impairment at ≥55 h/week, supporting the paper’s emphasis on cumulative burden rather than only weekly snapshots (Tang et al., 2026).
Conceptually, the authors frame the findings as a “triple burden” in which duration (long hours and weekend overtime), intensity (work pace), and workload (quantitative demands) jointly magnify risk, strengthening the interpretation that policies targeting hours alone may underperform if pace and task density remain unaddressed (Tang et al., 2026). This logic drives the practical recommendations: align scheduling with WHO/ILO-oriented limits around 55 h/week, implement monitoring and fatigue early-warning systems (including wearable-based indicators), adopt more physiologically protective shift designs, and institutionalize micro-breaks as an organizational practice rather than an individual coping tactic (Tang et al., 2026). The paper is explicit about constraints typical to cross-sectional designs, including residual confounding and generalizability limits, while still arguing that standardized measurement and comparable work stressors make the evidence informative for similar middle-income settings (Tang et al., 2026).
Mini glossary of important concepts (as used in the article)
Healthcare workers (HCWs): The study population, including qualified healthcare professionals (regular staff, trainees, interns) recruited across four hospitals, analyzed to understand how working time arrangements relate to mental health outcomes (Tang et al., 2026).
Chinese medical consortium: The organizational context comprising four public hospitals within one system, used to examine how institutional setting and shared governance context might shape overtime and mental health risks (Tang et al., 2026).
WHO/ILO cut-points: Standardized weekly-hours categories (35–40, 41–48, 49–54, ≥55 h/week) used to classify exposure and interpret thresholds in relation to international guidance (Tang et al., 2026).
Dose–response relationship: A pattern where mental health risk increases progressively with longer weekly work hours; here it is tested both graphically and via trend tests across ordered hours categories (Tang et al., 2026).
Subthreshold overtime (49–54 h/week): Working hours below the ≥55 h/week benchmark but above standard schedules, treated as a critical “early warning” range where several subgroups already show significant deterioration (Tang et al., 2026).
Work-hour thresholds (49 and 55 h/week): Empirically identified breakpoints where risk becomes evident (49 h/week) and where multi-domain deterioration becomes widespread (≥55 h/week) (Tang et al., 2026).
Weekend overtime frequency: A complementary exposure reflecting how often respondents work overtime on weekends, measured on a 5-point never-to-always scale (Tang et al., 2026).
COPSOQ III: The Copenhagen Psychosocial Questionnaire III, used here to measure six mental health dimensions on a 0–100 metric where higher scores indicate worse status (Tang et al., 2026).
Burnout: One COPSOQ III domain capturing exhaustion-related psychosocial strain; in this study, it is one of the earliest outcomes to show threshold effects and remains associated with ≥55 h/week in the fully adjusted model (Tang et al., 2026).
Somatic stress: A COPSOQ III domain reflecting bodily stress manifestations; it shows sensitivity in subthreshold overtime for key subgroups and remains independently associated with ≥55 h/week after full adjustment (Tang et al., 2026).
Cognitive stress: A COPSOQ III domain related to cognitive overload and sustained mental demands; it increases notably in female workers at 49–54 h/week in subgroup analyses (Tang et al., 2026).
Quantitative demands: A work-characteristics construct operationalizing workload amount and time pressure using multiple items (e.g., tasks piling up, lacking time), included in Model C to separate “how much work” from “how fast work must be done” (Tang et al., 2026).
Work pace: A work-characteristics construct capturing speed and sustained high tempo (e.g., needing to work very fast), emerging as a robust correlate across all six mental health outcomes (Tang et al., 2026).
Shift work status: An organizational factor (including rotating shifts) tested as an effect modifier; shift workers show a biphasic pattern with early somatic stress increases at 49–54 h/week and broad deterioration at ≥55 h/week (Tang et al., 2026).
Exposure duration: Work tenure categorized as short-term (<5 years) vs long-term (≥5 years), used to test whether cumulative exposure lowers tolerance thresholds (Tang et al., 2026).
Hierarchical regression (Models A, B, C): A stepwise adjustment approach that adds covariate blocks sequentially to assess how associations change when demographics, organizational context, and work characteristics are controlled (Tang et al., 2026).
HC3 robust standard errors and VIF: Statistical safeguards used to make inference more reliable under heteroskedasticity (HC3) and to rule out problematic multicollinearity (VIF < 2.5) (Tang et al., 2026).
References
Tang, Y., Zhang, M., Huang, Y., Wang, F., He, C., Fang, X., Wang, X., & Zhang, Y. (2026). The dose-response relationship between work hours and mental health in healthcare workers: A cross-sectional study in a Chinese medical consortium. Safety Science, 196, 107093. https://doi.org/10.1016/j.ssci.2025.107093