Developmental trajectories of depression, anxiety, and stress among college students: a piecewise growth mixture model analysis

The study investigates how depression, anxiety, and stress develop over four college years and identifies distinct subgroups of students who deviate from overall trends. Contextualized by increasing prevalence of mental health problems among youth and especially college students, the research emphasizes the transition to adulthood and the heightened academic and career pressures during university. Prior work has often analyzed overall changes, overlooking heterogeneity. Guided by a person-centered approach, the research question is to identify distinct developmental trajectories of depression, anxiety, and stress among Chinese college students and to determine internal (gender, personality, lifestyle) and external (family background, peer relationships) factors associated with membership in these trajectory classes. The sophomore year is posited as a critical turning point (knot) for change as students adapt to college life and clarify academic/career goals.

Prior studies using longitudinal designs have identified heterogeneous mental health trajectories across different populations. Among children and adolescents, classes such as low difficulty, improvers, decliners, and high difficulty have been reported; adolescents often show low vs. high anxiety classes with diverging trends. In adulthood and across the life course, growth mixture modeling (GMM) studies have found multiple trajectories, including absence of symptoms, moderate and severe symptom patterns with adolescent or adult onset, and repeated severe symptoms. During COVID-19, adults exhibited four depression and five anxiety trajectories. For college students specifically, variable-centered studies show elevated strain in the first year and partial recovery later, with some evidence of worsening in the first two years and improvement in the junior/senior years. However, few have examined subgroup heterogeneity among college students, especially in China over an entire 4-year span. The literature also identifies potential correlates of depression/anxiety/stress: internal factors (gender differences; personality traits such as extroversion; lifestyle including BMI, physical activity, and sleep) and external/social factors (family background such as hometown location, siblings, parental education; and peer relationships). Protective factors generally include extroversion, adequate sleep, healthy BMI, supportive peer relationships, urban upbringing, siblings, and higher parental education; risk factors include poor sleep, unhealthy BMI, poor peer relations, rural upbringing, no siblings, and lower parental education.

Design and data source: Longitudinal cohort study using the Beijing College Students Panel Survey (BCSPS), a stratified, multistage, probability proportional to size sample representative of Beijing college students. Cohort: students enrolled in 2008, followed across four consecutive academic years. Baseline N=2473 valid cases; follow-up response rates in years 2–4 were 95.27%, 94.66%, and 90.58%. Measures: Mental health was assessed annually with the DASS-42 (14 items each for depression, anxiety, stress; 0–3 Likert responses). Established Chinese validity/reliability. Depression severity cutoffs: normal (0–9), mild (10–13), moderate (14–20), severe (21–27), extremely severe (28–42). Anxiety: normal (0–7), mild (8–9), moderate (10–14), severe (15–19), extremely severe (20–42). Stress: normal (0–14), mild (15–18), moderate (19–25), severe (26–33), extremely severe (34–42). Missing data rates across waves were low (depression: 5.22%, 0.90%, 3.57%; anxiety: 5.63%, 0.34%, 3.57%; stress: 5.30%, 0.77%, 3.57%). Cronbach’s alphas across four years: depression 0.89–0.94; anxiety 0.82–0.92; stress 0.88–0.93. Correlates (baseline): gender (0=female, 1=male), self-rated extroversion (1–9), BMI categorized as low (<18.5=0), normal (18.5–24=1), high (>24=2); sleep hours categorized as low (<7h=0), normal (7–9h=1), high (>9h=2); relationship with classmates (1=very unfamiliar to 5=very familiar); hometown location (0=rural, 1=urban); siblings (0=yes, 1=no); father’s and mother’s education (years; 0, 6, 9, 12, 16, 19 as typical levels). Analytic strategy: A piecewise Growth Mixture Model (GMM) in Mplus 8.3 identified latent trajectory classes for each outcome (depression, anxiety, stress), with a knot at the sophomore year (piecewise slopes S1 and S2). Model selection considered AIC, BIC, SABIC, entropy (used cautiously), LMR-LRT and B-LRT tests, interpretability, and prior literature. Models with up to six classes were tested; optimal solutions were selected (depression: 4-class; anxiety: 5-class; stress: 5-class). Subsequently, multinomial logistic regression in Stata 16.0 examined baseline predictors of class membership relative to the low and stable reference class, including gender, extroversion, BMI, sleep hours, relationship with classmates, hometown, siblings, and parental education. Reporting includes relative risk ratios (RRR), 95% CIs, and p-values.

Sample characteristics: 47.15% female, 53.85% male; mean age 19.60±0.89. Mean extroversion 5.60±1.65; relationship with classmates 3.61±0.83. BMI: 71.95% normal, 19.19% low, 8.86% high. Sleep hours mean 7.41±0.84; 82.15% normal, 10.42% low, 7.43% high. Trajectory classes and proportions: - Depression (4 classes): Low and stable 79.34%; Increasing 8.61%; Decreasing then stable 8.25%; Increasing then decreasing 3.80%. - Anxiety (5 classes): Low and stable 79.95%; Increasing then decreasing 2.79%; Decreasing then stable 5.82%; Increasing 4.77%; Decreasing and high 6.67% (present only in anxiety and stress). - Stress (5 classes): Low and stable 76.35%; Decreasing and high 13.10%; Increasing then decreasing 4.65%; Decreasing then stable 2.99%; Increasing 2.91%. Significant predictors of class membership (RRR vs. low and stable): Depression: - Decreasing then stable: Urban hometown RRR=1.99 (p=0.01); Low sleep hours RRR=1.92 (p=0.01); Better relationship with classmates protective (RRR=0.75 per level, p=0.01). - Increasing then decreasing: Male RRR=1.42 (p=0.04); Low BMI RRR=1.48 (p=0.05); Low sleep hours RRR=1.58 (p=0.05). - Increasing: High BMI RRR=2.11 (p=0.02); High sleep hours RRR=2.31 (p=0.01). Anxiety: - Decreasing then stable: No siblings RRR=1.77 (p=0.03); Extroversion RRR=1.15 (p=0.02); Father’s education RRR=1.09 (p=0.04). - Increasing: Male RRR=1.83 (p=0.01); High sleep hours RRR=1.87 (p=0.04). - Increasing then decreasing and Decreasing and high: no significant predictors identified. Stress: - Decreasing then stable: Urban hometown RRR=3.41 (p=0.01); Low sleep hours RRR=2.62 (p=0.00). - Increasing: Urban hometown RRR=2.25 (p=0.05). - Decreasing and high: Low sleep hours RRR=1.89 (p=0.00). Overall patterns: Most students fall into low-stable classes, but meaningful subgroups exhibit worsening or high early symptoms that later decline. The “decreasing and high” class is unique to anxiety and stress. Sleep irregularities (both short and long duration) and non-normal BMI are linked to higher-risk depression trajectories; male gender is associated with increasing anxiety; urban background associates with certain stress/depression classes; and social connectedness with classmates is protective for depression.

The study addresses the research question by demonstrating heterogeneous mental health trajectories among college students across four years and identifying internal and external correlates of class membership. The piecewise GMM with a sophomore-year knot captured critical shifts: many students exhibit initial strain upon entering university with subsequent improvement, while others experience increasing symptoms, likely reflecting rising academic and employment pressures toward graduation. The identification of a sizable low-stable class supports that most students maintain normal levels, yet the presence of increasing and high early-symptom classes underscores the need for targeted interventions. Internal factors such as male gender, sleep deviations (short and long), and non-normal BMI, and external factors such as urban hometown, lack of siblings, father’s higher education, and weaker peer relationships, differentially predict riskier trajectories. These findings validate a person-centered approach, moving beyond average trends to inform prevention efforts: proactively support students at risk of increasing trajectories, monitor those expected to improve to ensure symptoms decline as anticipated, and pay special attention to the “decreasing and high” anxiety/stress group whose symptoms remain elevated into later years. The results highlight the importance of sleep health, healthy weight, and peer integration programs, and suggest that family background dynamics may shape early adjustment to university life.

- The study classifies developmental trajectories across four college years: depression (low and stable; increasing; decreasing then stable; increasing then decreasing), anxiety (low and stable; increasing; increasing then decreasing; decreasing then stable; decreasing and high), and stress (same five-class structure as anxiety). The “decreasing and high” class appears only in anxiety and stress. - Baseline correlates distinguish trajectory membership. For depression, gender, peer relationships, hometown, BMI, and sleep hours are relevant; for anxiety, gender, extroversion, siblings, father’s education, and high sleep hours matter; for stress, urban hometown and low sleep hours are key. - Implications: Early identification of students at risk for increasing and persistently high trajectories is crucial. Interventions should be tailored by trajectory class, emphasizing sleep regulation, weight and lifestyle management, peer relationship building, and support for students with risk-related family backgrounds. Future research should expand beyond Beijing to broader Chinese and cross-cultural samples, incorporate multimethod assessment beyond self-report, and refine measurement of contextual factors.

- Generalizability: The sample represents Beijing college students; findings may not generalize to all Chinese college populations. - Measurement: DASS-42 is self-reported, introducing potential response and reporting biases. - Single-item measures: Some predictors (e.g., extroversion, peer relationships) were assessed with single items, warranting cautious interpretation and replication with multi-item validated scales.