The study recruited 2,378 nursing students from four medical colleges across Hubei Province, collecting data between January 2025 and March 2026. Each participant completed two well-established psychometric instruments. The first was the 25-item Connor-Davidson Resilience Scale, a questionnaire that probes multiple facets of resilience, including personal competence, trust in one’s instincts, tolerance of negative affect, acceptance of change, control and spiritual influences. The second was the 21-item Depression Anxiety Stress Scales, known as DASS-21, which yields separate scores for each of the three emotional states as well as a composite measure of overall distress. The research team received ethics approval from the Institutional Review Board of Hubei University of Medicine, conducted all procedures in accordance with the Declaration of Helsinki, and obtained written informed consent from every participant, with responses collected anonymously and participation entirely voluntary.

The analytical centerpiece of the study was latent profile analysis, a person-centered statistical method that differs fundamentally from the variable-centered approaches most readers will be familiar with. Where a traditional correlation asks how scores relate across a whole sample, latent profile analysis asks whether the sample actually contains hidden subgroups, or profiles, of individuals who answer the resilience items in similar ways. The method fits a series of statistical models, each proposing a different number of profiles, and uses fit indices such as the Akaike information criterion, the Bayesian information criterion and bootstrapped likelihood-ratio tests to decide how many profiles the data genuinely support. A key quality metric is entropy, which ranges up to 1.0 and indicates how cleanly individuals are sorted into their assigned profiles. In this study the three-profile solution achieved an entropy of 0.958, an exceptionally high value that signals near-unambiguous classification.

That three-profile solution revealed a striking asymmetry in the distribution of resilience. The largest group, comprising 52.2 percent of students, showed a lower adversity-adaptation profile, meaning their item-level responses reflected comparatively limited capacity to adapt to hardship. A second group of 40.6 percent occupied an intermediate position, while only 7.2 percent of students belonged to a higher adversity-adaptation profile marked by robust resilience across the measured dimensions. The proportions themselves are a finding worth pausing on: fewer than one in twelve future nurses in the sample displayed the highest resilience pattern, while more than half fell into the lowest. For a profession whose daily work demands sustained psychological endurance, that distribution is a sobering portrait of the pipeline supplying the clinical workforce.

The emotional health consequences of profile membership followed a clean graded pattern. Depression, anxiety, stress and total DASS-21 scores all differed significantly across the three profiles, with all comparisons reaching the p < 0.001 threshold. Students in the lower adversity-adaptation profile reported the highest distress, those in the intermediate profile fell in the middle, and those in the higher profile reported the lowest. When the researchers quantified these differences against the higher profile as the reference group, the lower profile was associated with total DASS-21 scores 18.661 points higher, and the intermediate profile with scores 5.775 points higher. In psychometric terms, an 18-point gap on the DASS-21 total represents a clinically meaningful difference in overall emotional distress, not a statistical curiosity.

To test whether resilience profiles carried information beyond basic demographics, the team built hierarchical regression models. A first model incorporating six demographic, family and relational covariates explained 8.7 percent of the variance in total DASS-21 scores. Adding profile membership to that model raised the explained variance to 18.1 percent, an increase of 9.4 percentage points. In other words, knowing a student’s resilience type nearly doubled the predictive power of the model, demonstrating that the profiles capture psychological information that conventional background variables miss. The researchers also employed multinomial logistic regression and the Bolck-Croon-Hagenaars procedure, a technique that corrects for classification uncertainty when relating latent profiles to external outcomes, to characterize which students ended up in which profiles and to confirm the profile-outcome associations.

The profiles were not randomly distributed across the student body. Differences in sociodemographic and relational characteristics distinguished the groups, indicating that family circumstances, interpersonal relationships and demographic factors shape which resilience pattern a student develops. The study’s cross-sectional design means these associations cannot be read as causal chains; it is equally plausible that chronic distress erodes resilience, that low resilience amplifies distress, or that shared upstream factors drive both. What the design does establish, with considerable statistical confidence, is that the two constructs are tightly intertwined and that resilience profiles function as a powerful marker of who is most likely to be suffering.

Why does the person-centered framing matter so much for practice? Traditional screening in nursing schools typically relies on either a single resilience score or direct symptom questionnaires. A single score conceals the architecture of a student’s psychological resources: two students with identical total resilience scores may arrive there through entirely different patterns of strengths and vulnerabilities. By identifying discrete profiles, the latent profile approach gives educators a typology rather than a number. The authors suggest that nursing schools could combine universal support programs, offered to everyone, with direct symptom assessment and mentoring targeted to students’ needs, using profile membership as one signal for allocating attention. The graded relationship between profiles and distress means the intermediate group, which is large and often overlooked because its members are neither the most nor the least distressed, may deserve particular attention.

The study also carries implications for the nursing profession at large. Burnout, emotional exhaustion and attrition among nurses are persistent global problems, and the psychological habits formed during training are widely believed to carry forward into professional practice. If more than half of entering nursing students show the lowest resilience pattern, and if that pattern is associated with markedly elevated depression, anxiety and stress, then the classroom and the clinical placement become critical intervention windows. Resilience, unlike fixed traits, is considered modifiable, and profile-informed support could in principle be designed to strengthen the specific adaptive capacities that a given profile lacks, rather than applying generic wellness programming uniformly.

The authors are careful about the limits of their evidence. Because the data were collected at a single point in time, the direction of the relationship between resilience profiles and emotional symptoms remains undetermined, and the sample, drawn entirely from medical colleges in one Chinese province, may not generalize to nursing students elsewhere. The researchers explicitly call for longitudinal and intervention studies to determine whether profile-informed support actually improves outcomes. Still, the technical achievement is substantial: a near-perfect classification solution, a graded and highly significant association with distress, and a demonstration that resilience profiles add nearly ten percentage points of predictive variance beyond demographics. As mental health screening in professional education moves toward precision approaches, this study offers a template for how hidden psychological subgroups can be surfaced from ordinary questionnaire data, and a reminder that the students caring for tomorrow’s patients need their own care designed with the same sophistication.

Subject of Research: Resilience profiles and their association with depression, anxiety and stress among Chinese nursing students

Article Title: Latent profile analysis of resilience and its relationship with depression-anxiety-stress among Chinese nursing students: a cross-sectional study

Article References: Wang, Y., Zhang, T., Wu, M., Li, X., Xiao, L., & Ke, L. (2026). Latent profile analysis of resilience and its relationship with depression-anxiety-stress among Chinese nursing students: a cross-sectional study. BMC Psychology. https://doi.org/10.1186/s40359-026-05720-x

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05720-x

Keywords: resilience, nursing students, latent profile analysis, depression, anxiety, stress, mental health, DASS-21, Connor-Davidson Resilience Scale, cross-sectional study, China, psychological assessment