The clinical internship is a pivotal and stressful rite of passage for nursing students. It marks the transition from classroom theory to the realities of patient care: long shifts, exposure to suffering and death, demanding supervisors, and the constant fear of making mistakes with real lives at stake. Chinese nursing interns face additional pressures, including intense competition in the healthcare job market and the high-stakes structure of clinical placements that can shape career prospects. Previous studies have repeatedly flagged psychological distress among this group, but no quantitative synthesis of positive-screening prevalence during the internship itself had been attempted until now.

To fill that gap, the research team searched nine databases, four Chinese and five English, from their inception to 28 June 2026, hunting for cross-sectional studies that reported positive-screening rates of anxiety or depressive symptoms among nursing interns in mainland China at validated cut-offs. The search covered major Chinese repositories such as CNKI, SinoMed, and the VIP database alongside international sources including PubMed, CINAHL, and others. Study quality was appraised with the Agency for Healthcare Research and Quality checklist, and the review was prospectively registered on PROSPERO in June 2026, a step that strengthens transparency in how the analysis was planned before the data were seen.

The statistical machinery behind the synthesis is worth understanding, because it shapes how the results should be read. The team pooled prevalence estimates using the Freeman–Tukey double-arcsine transformation, a technique that stabilizes the variance of proportions, particularly when rates are close to zero or one, before combining them in random-effects models fitted with restricted maximum likelihood. Crucially, the researchers went beyond the familiar confidence interval and reported 95 percent prediction intervals, which describe the plausible range of true screening rates in a new cohort drawn from the same distribution of settings. When heterogeneity is extreme, as it was here, the prediction interval is often the more honest summary of what the evidence actually shows.

That heterogeneity was, in a word, enormous. Thirty-three studies reported anxiety screening at a standard cut-off, covering 10,157 interns, and pooled to 28.4 percent with a 95 percent confidence interval of 22.8 to 34.4 percent. But the I-squared statistic, which quantifies the proportion of variation attributable to real differences between studies rather than chance, stood at 97.6 percent. The 95 percent prediction interval stretched from roughly 4 percent to 64 percent, meaning that in any single cohort of interns the true positive-screening rate could plausibly sit anywhere from a small fraction of the group to nearly two thirds of it. The pooled figure is a centre of gravity, not a national prevalence.

Depressive symptoms told a similar story with even thinner evidence. Only six studies reported depression screening at a standard cut-off, covering 2,674 interns, pooling to 37.7 percent with a confidence interval of 27.3 to 48.8 percent and a prediction interval of roughly 13 to 66 percent. A technical wrinkle compounded the problem: most studies using the Self-Rating Depression Scale scored positive from 50 rather than the standard threshold of 53, a choice that inflates positive-screening rates. When the team pooled all 22 reported depression rates regardless of threshold, the summary was 34.4 percent. The authors are explicit that the depression estimate rests on few studies that used a standard threshold, and that the summaries describe a specific cohort rather than a national figure.

Robustness checks gave the anxiety estimate firmer footing. The team re-ran the pooling with logit and generalized linear mixed models, performed leave-one-out analyses to test whether any single study was driving the result, restricted analyses to higher-quality studies, and examined funnel-plot asymmetry with Begg’s and Egger’s tests, which probe whether smaller studies with larger effects are systematically missing, a signature of publication bias. The anxiety summary survived all of these probes, with no detectable funnel-plot asymmetry. Exploratory subgroup analyses, including one testing whether screening rates differed by internship stage, found no significant differences; the internship-stage comparison yielded a p-value of 0.468. At this level of variation, the authors note, the prediction intervals carry more information than the point summaries themselves.

One part of the review came up empty, and that emptiness is itself informative. The team had hoped to synthesize the factors associated with anxiety and depressive symptoms, such as demographic characteristics, internship conditions, or coping resources, but the associated factors were too inconsistently defined across studies to pool quantitatively. Different studies measured different exposures with different instruments and reported different multivariable models, making any combined odds ratio unreliable. The authors published the underlying multivariable models in supplementary tables, but the headline conclusion is that the field lacks standardized measurement of what puts interns at risk, a gap that prospective research must close.

The authors are careful to draw a bright line between screening and diagnosis. The figures describe positive screens on validated instruments such as the Generalized Anxiety Disorder 7-item scale, the Self-Rating Anxiety Scale, the Patient Health Questionnaire-9, and the Self-Rating Depression Scale at stated cut-offs. A positive screen is not a clinical diagnosis of an anxiety disorder or major depression; it flags symptoms that warrant further assessment. Even so, the sheer scale of the numbers, with screening rates centred near 28 percent for anxiety and between one third and two fifths for depressive symptoms, and plausibly running from a few percent to about two thirds in any given cohort, suggests that psychological distress during clinical placement is not a marginal problem but a structural feature of the internship experience.

What should be done with these numbers? The authors argue that the summaries serve as a benchmark for a specific cohort, not a national figure, and that the priority now is prospective measurement with standard instruments and explicitly stated thresholds. That means following interns over time with consistent tools, rather than relying on one-off cross-sectional snapshots with shifting cut-offs. For nursing schools and teaching hospitals, the practical message is that mental health support during clinical placement, from accessible counselling to workload adjustments and supervisor training, deserves systematic attention. For the research community, the message is sharper still: without standardized measurement, the true burden of distress among the nurses of tomorrow will remain hidden inside a wide, setting-dependent distribution that no single pooled percentage can capture.

Subject of Research: Mental health screening prevalence among nursing interns during clinical practice in China

Article Title: Prevalence of anxiety and depressive symptoms and associated factors among nursing interns during clinical practice in China: a systematic review and meta-analysis

Article References: Chen, X., Zhang, Y., Zhu, X., & Zhang, F. (2026). Prevalence of anxiety and depressive symptoms and associated factors among nursing interns during clinical practice in China: a systematic review and meta-analysis. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08777-5

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08777-5

Keywords: nursing interns, anxiety symptoms, depressive symptoms, meta-analysis, systematic review, China, clinical practice, mental health, screening prevalence, nursing students, prediction interval, BMC Psychiatry