EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

Predicting Depressive Symptoms From Cross‐Sectional and Longitudinal Data Using XGBoost–SHAP and Cross‐Lagged Model

Jing Jie iD, Ming Yin iD, Jiayue Zeng iD, Songlin Yang iD

Depression and Anxiety · 2026

DOI: 10.1155/da/4401725 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Depression and Anxiety, 2026. Verfasst von Jing Jie und 3 weiteren Autor:innen. Beteiligte Einrichtung: State Key Laboratory of Digital Medical Engineering Key Laboratory of Biomedical Engineering of Hainan Province School of Biomedical Engineering Hain…. Hauptthemen laut MEDLINE: Depression, COVID-19, Students. Die Arbeit nennt 71 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Natural Science Foundation of Hainan Province, Key Research and Development Project of Hainan Province.

AbstractSchlagwörterBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

University students face significant academic, social, and career‐planning challenges. Under such stressors, deficient emotion regulation may increase vulnerability to depressive states, in which persistent symptoms could progress to clinical depression. This study investigates key risk and protective factors for depressive symptoms among university students, including negative life events, neuroticism, and social support. A longitudinal framework was further employed to assess changes in depressive symptoms before and after the COVID‐19 pandemic, with the goal of informing early identification and intervention strategies.

Methods

The dataset included 18 024 valid responses from baseline assessments. Follow‐up assessments conducted 1 year later included 6129 participants. The extreme gradient boosting (XGBoost) model with SHapley Additive exPlanations (SHAP) was used to quantify the relative importance of specific factors (e.g., negative life events and social support) on depressive symptoms, while a cross‐lagged panel model examined longitudinal directional associations between these factors and depressive symptoms.

Results

The prevalence of depressive symptoms was 15.2% among the student sample. Significant correlations were observed between symptom severity and adverse family relationships, childhood trauma, prior suicide attempts, and psychiatric history. Neuroticism and negative life events showed significant positive associations with depressive symptoms, while subjective support and support utilization demonstrated negative associations with symptom severity. Longitudinal analysis revealed an increase in both symptom severity and depression detection rates following the COVID‐19 pandemic. The cross‐lagged panel model identified neuroticism as a significant positive longitudinal predictor of depressive symptoms, whereas subjective support showed a significant negative longitudinal association with subsequent depressive symptoms.

Conclusion

This study identifies neuroticism as an important longitudinal risk indicator for depressive symptoms and subjective support as a protective indicator, while negative life events showed significant cross‐sectional associations with symptom severity. The findings demonstrate how integrating longitudinal designs with interpretable machine learning elucidates complex mental health dynamics in student populations, providing an evidence‐based framework for targeted interventions within campus mental health policies.

Schlagwörter

MeSH-Schlagwörter (MEDLINE)

DepressionCOVID-19StudentsHumansLongitudinal StudiesFemaleCross-Sectional StudiesMaleNeuroticismBoosting Machine Learning AlgorithmsYoung AdultSocial SupportAdultUniversitiesRisk FactorsPredictive Learning Models

Stichwörter der Autor:innen

COVID-19 pandemiccross-lagged panel modeldepressive symptomsmachine learningneuroticismsocial support

Autor:innen und Institutionen

  • State Key Laboratory of Digital Medical Engineering Key Laboratory of Biomedical Engineering of Hainan Province School of Biomedical Engineering Hainan University Sanya Hainan 572024 China hainu.edu.cn.
  • School of Biomedical Engineering Sichuan University Chengdu China scu.edu.cn.

ORCID-Profile: Jing Jie, Ming Yin, Jiayue Zeng, Songlin Yang

Förderung

  • National Natural Science Foundation of China
  • Natural Science Foundation of Hainan Province
  • Key Research and Development Project of Hainan Province

Zitieren

APA

Jie, J., Yin, M., Zeng, J., & Yang, S. (2026). Predicting Depressive Symptoms From Cross‐Sectional and Longitudinal Data Using XGBoost–SHAP and Cross‐Lagged Model. Depression and Anxiety. https://doi.org/10.1155/da/4401725

Vancouver

Jie J, Yin M, Zeng J, Yang S. Predicting Depressive Symptoms From Cross‐Sectional and Longitudinal Data Using XGBoost–SHAP and Cross‐Lagged Model. Depression and Anxiety. 2026. doi:10.1155/da/4401725

BibTeX
@article{Jie2026,
  author = {Jie, Jing and Yin, Ming and Zeng, Jiayue and Yang, Songlin},
  title = {Predicting Depressive Symptoms From Cross‐Sectional and Longitudinal Data Using XGBoost–SHAP and Cross‐Lagged Model},
  journal = {Depression and Anxiety},
  year = {2026},
  doi = {10.1155/da/4401725}
}
Kostenlos starten →