International Open Access Journal Platform

精准分诊与资源优化：基于UPI 决策树模型的医校协同路径构建

Toward a Medical-school Collaborative Pathway: A UPIBased Decision Tree Model for Precise Psychological Triage and Resource Optimization

Psychology of China / 2026,8(3): 286-293 / 2026-03-25 13 7

View Download PDF

• Authors: 干瑜璐¹ 马露瑶¹ 袁松²
• Information:

  1．浙江国际海运职业技术学院学生处，舟山；
  2．舟山市第二人民医院精神科，舟山

• Keywords:
  University student mental health; University personality inventory (UPI); Decision tree analysis; Medicalschool collaboration; Risk stratification
  大学生心理健康; UPI量表; 决策树分析; 医校协同; 风险分层
• Abstract: Objective: To systematically evaluate the efficacy of the University Personality Inventory (UPI) in psychological screening among university students, identify key items with significant predictive value, and develop an optimized screening strategy for precise triage and resource allocation that can serve the medical-school collaboration mechanism. Methods: A UPI survey was administered to all incoming first-year students at a university from 2023 to 2025. Students with a UPI total score≥25 or a positive response to item 25 (suicidal ideation within the past month) underwent semistructured clinical interviews for risk level assessment. Key predictive items were identified using binary logistic regression and cross-tabulation analysis, followed by systematic analysis incorporating Receiver Operating Characteristic (ROC) curves and decision tree algorithms. Results: The UPI total score demonstrated excellent predictive ability for high psychological risk status (Area Under the Curve [AUC] = 0.948, N=7,767), with the current cutoff score of 25 being statistically optimal. Binary logistic regression selected 14 key items into the final model, among which items Q62 (Have you ever felt that you had psychological problems) and Q63 (Have you ever received psychological counseling or psychological treatment) showed the strongest predictive power (p<0.001). Decision tree analysis (conducted on the initial screening-positive subsample, N=1,065) identified Q63 as the most important predictor, forming effective predictive pathways in combination with other items. Conclusion: The constructed two-stage screening model (“Total Score Initial Screening + Multidimensional Triage”), when applied to the UPI initial screening-positive group, not only maintained a high detection rate for high-risk students (sensitivity = 87.10%) but also reduced the number of students requiring one-on-one in-depth interviews by approximately 70.58%. This significantly improves the efficiency of interview resource allocation and provides empirical evidence for enhancing the scientific approach to university psychological screening and establishing a medical-school collaborative support mechanism. 目的：系统评估UPI量表在高校心理筛查中的效能，探索关键条目的预测价值，构建一套可服务于医校协同机制的精准分诊与资源优化策略。方法：对某高校2023—2025三年间的所有大一新生进行UPI普查，对UPI总分≥25分或第25题（近一个月有自杀意念）阳性的学生进行专业心理访谈并确定风险等级，采用二元Logistic回归、交叉表分析筛选关键条目，结合ROC曲线、决策树等方法进行系统分析。结果：UPI总分对心理高危状态具有优秀预测能力（AUC=0.948，N=7767），现行25分标准为统计最优值。二元Logistic回归筛选出14个关键条目进入最终模型，其中Q62（是否曾觉得自己存在心理问题）、Q63（是否曾接受心理咨询或心理治疗）等条目预测力最强（p<0.001）。决策树分析（在初筛阳性子样本N=1065中）发现Q63为最重要的预测因子，能结合其他条目形成有效预测路径。结论：基于此构建的“总分初筛+多维分诊”双阶段筛查模型在UPI初筛阳性群体的应用中，不仅能兼顾对高危学生较高的覆盖率（敏感度87.10%），而且能将需进行一对一深度访谈的学生数量减少约70.58%，从而显著提升访谈资源的配置效率，为高校心理筛查工作的科学化与医校协同机制的构建提供实证依据。
• DOI: https://doi.org/10.35534/pc.0803044 (registering DOI)
• Cite: 干瑜璐, 马露瑶, 袁松. (2026). 精准分诊与资源优化: 基于UPI决策树模型的医校协同路径构建. 中国心理学前沿, 8 (3), 286-293.