Meta-Analysis Reveals Substantial Variation in Depressive Symptoms Among ICU Nurses Worldwide

A comprehensive systematic review and meta-analysis synthesizing observational studies across twenty-one countries has estimated that 38.6 percent of intensive care unit nurses experience depressive symptoms, while uncovering pronounced variability driven by national economic development and the timing of the COVID-19 pandemic. Conducted by critical care medicine and nursing researchers at Sichuan University and published in Frontiers in Public Health, the investigation evaluated thirty-eight cross-sectional studies incorporating 22,474 critical care nurses to address long-standing discrepancies in occupational mental health literature. Although the calculated summary proportion reflects a substantial psychological burden, the authors reported extreme statistical heterogeneity across the included literature, characterized by an inconsistency index of 98.9 percent and a between-study variance of 0.23. More importantly, the derivation of a 95 percent prediction interval spanning from 2.8 percent to 84.6 percent demonstrates that the actual prevalence of depressive symptoms in any given clinical setting fluctuates across an exceptionally wide spectrum. Consequently, the research team emphasized that this pooled estimate cannot be interpreted as a fixed or universally applicable global baseline, but rather serves as an epidemiological indicator of widespread vulnerability shaped by systemic and situational working conditions.

The intensive care unit represents a uniquely demanding clinical ecosystem wherein nurses routinely confront intense occupational stressors that exceed those typical of general hospital wards. Prolonged exposure to these compounding occupational stressors without adequate recovery periods precipitates adverse psychological sequelae, including generalized anxiety, persistent sleep dysregulation, and depressive symptoms. Clinically, these depressive manifestations encompass enduring depressed mood, pervasive anhedonia, cognitive deficits, chronic physical fatigue, appetite disruptions, diminished feelings of personal worth, and elevated risks of suicidal ideation. Beyond eroding individual well-being, widespread depressive symptoms among critical care nurses correlate with increased clinical error rates, compromised patient safety standards, and heightened institutional turnover intentions.

Despite widespread recognition that intensive care environments pose substantial mental health risks, previously published empirical surveys generated highly disparate prevalence figures, with reported rates ranging from as low as 6.1 percent to as high as 77.9 percent. The authors noted that these pronounced divergences stemmed from unaccounted differences in sample sizes, cultural contexts, regional economic realities, assessment schedules, participant age distributions, and accumulated professional experience. While an earlier systematic synthesis had attempted to consolidate the empirical literature, that prior effort predated the emergence of the COVID-19 emergency and omitted studies conducted after the pandemic began to fundamentally destabilize hospital operations globally. Furthermore, the earlier review did not perform extensive subgroup stratification or meta-regression techniques capable of identifying why estimates varied so dramatically between study cohorts. To rectify these analytical gaps, the investigators structured their updated synthesis to provide a rigorous, source-backed appraisal of depressive symptom prevalence, prospectively registering their protocol with the international prospective register of systematic reviews under registration code CRD420261435951 and executing the study under Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines.

To assemble the evidence base, the research team conducted an exhaustive systematic literature search spanning seven major scientific databases: PubMed, Web of Science, Scopus, Embase, the Cochrane Library, CINAHL, and PsycINFO from their respective inception dates through September 30, 2026. The search architecture utilized combinations of Medical Subject Headings and free-text terms tailored to identify peer-reviewed observational studies examining depression and depressive symptoms specifically among intensive care nursing staff across adult, pediatric, and neonatal units. Eligibility criteria required cross-sectional or longitudinal designs that administered validated self-report screening instruments and either reported numerical prevalence figures or provided sufficient data to derive them. The protocol excluded conference abstracts, study protocols, narrative reviews, case reports, duplicate entries, and non-English publications. Two independent reviewers extracted study descriptors, including first author, publication year, survey location, participant sample sizes, mean participant ages, marital and gender demographics, measurement tools, and diagnostic cutoff scores. To prevent sample duplication across multi-center reports, the reviewers cross-referenced author teams, clinical institutions, recruitment intervals, and demographic compositions, retaining only the largest cohort when overlapping participant groups were detected.

Methodological analysis required specialized statistical handling due to extreme differences in study sizes and the occurrence of proportion values near distributional extremes. The researchers conducted their analytical procedures in Stata version 19.0, applying the Freeman-Tukey double-arcsine transformation to stabilize variance across individual study proportions before synthesizing estimates. Pooled results were subsequently converted back to original proportion scales using inverse transformation methods to permit direct interpretation. Because substantial clinical and methodological diversity was anticipated across international clinical environments, the investigators established a random-effects framework a priori, estimating between-study variance using restricted maximum likelihood. Heterogeneity was evaluated through the inconsistency index alongside tau-squared, with values of 50 percent or higher designating substantial between-study divergence. In addition to standard 95 percent confidence intervals, the investigators derived 95 percent prediction intervals to project the statistical range within which true prevalence rates would be expected to fall in newly evaluated settings. Prespecified subgroup comparisons and exploratory meta-regressions were established to test potential moderating influences, categorizing national economies via United Nations M49 criteria and classifying survey periods relative to World Health Organization declarations regarding the COVID-19 pandemic emergency.

Critical appraisal of the thirty-eight included articles was conducted using the Joanna Briggs Institute critical appraisal checklist for prevalence studies, which evaluates nine distinct methodological quality criteria. Following dual-reviewer assessment with third-party adjudication, thirty studies, representing 78.9 percent of the included literature, were classified as possessing low risk of methodological bias, while the remaining eight studies, or 21.1 percent, exhibited moderate risk. When examining individual quality domains, sample frames were judged appropriate in twenty investigations, accounting for 52.6 percent of the evidence base, although every single included study relied on convenience sampling rather than random epidemiological sampling. Adequate statistical power and clear sample size justifications were documented in twenty-seven studies, representing 71.1 percent of the total cohort, and twenty-one studies achieved acceptable response rates. Notably, all thirty-eight included investigations demonstrated satisfactory performance across four vital appraisal domains: delivering comprehensive descriptions of clinical settings and nursing participants, ensuring adequate data coverage in analytical procedures, employing validated psychometric measurement scales, and applying appropriate statistical calculations to evaluate primary outcomes.

Stratified subgroup evaluations identified statistically significant differences in depressive symptom prevalence across economic classifications and temporal pandemic boundaries. When examining national economic status based on United Nations criteria, studies conducted in developing economies yielded a pooled depressive symptom prevalence of 45.1 percent across twenty-six cohorts, with a 95 percent confidence interval spanning 36.8 to 53.7 percent. In contrast, twelve investigations performed within developed nations demonstrated a significantly lower summary prevalence of 25.4 percent, with a confidence interval between 14.6 and 38.0 percent, establishing a statistically significant subgroup disparity with a p-value of 0.01. Temporal stratification relative to the COVID-19 pandemic demonstrated an equally pronounced divide. Twenty-one studies conducted during the formal pandemic timeframe reported a pooled prevalence of 47.2 percent, whereas seventeen investigations completed outside the pandemic window produced a pooled estimate of 28.3 percent, representing another statistically significant difference with a p-value of 0.01. Conversely, stratifying studies by instrument type revealed no statistically significant divergence, with depression-specific screening inventories yielding a pooled rate of 41.5 percent and multidimensional psychological distress scales producing 36.6 percent, reflecting a non-significant p-value of 0.52.

To substantiate whether these moderating variables exerted independent influences on symptom frequency, the researchers conducted exploratory univariate and multivariable meta-regressions on the transformed Freeman-Tukey scale. Univariate regression confirmed that national economic development and data collection timing were significantly associated with pooled prevalence at p-values of 0.009. Developing economy classification exhibited a positive regression coefficient of 0.4136, accounting for 14.2 percent of between-study variance, while data collection outside the pandemic demonstrated an inverse coefficient of negative 0.3879, explaining 13.8 percent of variance. None of the other demographic or methodological moderators, including total cohort sample size, mean participant age, proportion of married nurses, proportion of female staff, or psychometric instrument category, exhibited statistically significant associations in univariate models. In the multivariable meta-regression combining available complete cases, both economic status and pandemic timing retained their independent statistical significance, with developing country status generating a coefficient of 0.4017 at a p-value of 0.005, and non-pandemic timing generating a coefficient of negative 0.4548 at a p-value of 0.001, together accounting for 31.1 percent of total between-study variance.

The authors carried out a series of rigorous sensitivity and publication bias evaluations to confirm the statistical stability of their synthesized findings. Visual inspection of funnel plot distribution revealed asymmetry, which was corroborated by a statistically significant Egger’s linear regression test indicating potential small-study effects. However, non-parametric trim-and-fill analysis identified zero missing or unpublished studies, leaving the final summary estimate unchanged and suggesting that plot asymmetry reflected profound between-study diversity rather than systemic publication withholding. Leave-one-out sensitivity testing confirmed that omitting any single study did not alter the pooled prevalence estimate. Methodological sensitivity analyses further demonstrated remarkable consistency: restricting analysis to low-risk investigations yielded a pooled prevalence of 41.8 percent; limiting the synthesis to depression-specific screening scales yielded 41.5 percent; excluding studies with unstated threshold cutoffs produced 42.1 percent; and excluding the single largest investigation, an extensive Chinese cohort comprising 13,745 participants, produced a pooled figure of 38.1 percent. Additionally, re-estimating the primary synthesis via a logit-transformed random-effects model yielded a comparable summary estimate of 37.2 percent, demonstrating that the primary findings were not artifacts of analytical modeling choices.

In contextualizing their findings, the research team highlighted several substantive methodological limitations inherent to the synthesized international evidence. The strict restriction to English-language peer-reviewed literature and the exclusion of grey literature may have introduced linguistic selection bias. Furthermore, the geographic distribution of eligible research was heavily concentrated within Asian healthcare systems, potentially constraining the direct generalizability of the findings to clinical environments operating under different cultural dynamics, institutional hierarchies, and nursing labor models. Substantial heterogeneity also persisted in psychometric measurement: studies utilized diverse screening instruments, including the Hospital Anxiety and Depression Scale, Beck Depression Inventory, Patient Health Questionnaire, Center for Epidemiologic Studies Depression Scale, Depression Anxiety Stress Scales, Self-Rating Depression Scale, and General Health Questionnaire, frequently employing differing threshold cutoffs or failing to report specific threshold scores altogether.

Subject of Research: Prevalence and moderating factors of depressive symptoms among intensive care unit nurses

Article Title: Prevalence of depressive symptoms among intensive care unit nurses: a systematic review and meta-analysis of observational studies

Article References: Huang, C.-X., Tang, Q.-X., Lu, Y., Ruan, Y., & Li, X. (2026). Prevalence of depressive symptoms among intensive care unit nurses: a systematic review and meta-analysis of observational studies. Frontiers in Public Health, 14, Article 1956872. https://doi.org/10.3389/fpubh.2026.1956872

Image Credits: AI Generated

DOI: 10.3389/fpubh.2026.1956872

Keywords: intensive care unit, critical care nursing, depressive symptoms, occupational mental health, systematic review, meta-analysis, COVID-19 pandemic, economic status, nurse staffing, workplace stressors, Prevalence, depressive