A Machine Learning-Based Method for Developing the Chinese Symptom Checklist-11 (CSCL-11)

Abstract

The Chinese version of the Symptom Checklist-90 (SCL-90) is excessively lengthy, resulting in extended completion time and reduced respondent compliance. This study aimed to utilize a condensed subset of items from the Chinese SCL-90 to identify individuals at high risk for psychological disorders based on machine learning methods, forming a concise and efficient preliminary psychopathological screening instrument for the Chinese general population. Analyzing data collected from 4808 SCL-90 psychological surveys, this study applied variable clustering to select the most representative items, resulting in an 11-item scale: the Chinese Symptom Checklist-11 (CSCL-11). The CSCL-11 demonstrated high internal consistency (Cronbach’s α = 0.84). The results of factor analysis supported a single-factor model for the CSCL-11, demonstrating an acceptable fit (SRMR = 0.035, RMSEA = 0.064, CFI = 0.935, and TLI = 0.919). The CSCL-11 demonstrated strong predictive performance for the Global Severity Index (GSI; RMSE = 0.11, R2 = 0.92, Pearson’s r = 0.96) and various subscale scores (RMSE < 0.25, R2 > 0.70, Pearson’s r > 0.85). Additionally, it achieved a 96% accuracy rate in identifying individuals at high risk for psychological disorders. The comparison results indicated that the CSCL-11 outperformed SCL-14, SCL-K11, and SCL-K-9 in predicting GSI scores. In identifying high-risk groups, CSCL-11 demonstrated performance similar to that of SCL-14 and surpassed both SCL-K11 and SCL-K-9. The CSCL-11 retains most of the critical information from the original Chinese SCL-90 and serves as a preliminary psychopathological screening tool for the Chinese general population.