Explainable Machine Learning for Predicting 12-Month Relapse in Compulsory Drug Rehabilitation: A Multicenter Study

DOI: https://doi.org/10.62517/jbdc.202601125

Author(s)

Qingjiao Zeng*

Affiliation(s)

Zhejiang Police Vocational College, Hangzhou, Zhejiang, China *Corresponding Author

Abstract

Relapse following compulsory drug rehabilitation remains a critical barrier to sustained recovery, yet routine discharge assessments often lack the prospective precision required for targeted interventions. This multicenter study developed an interpretable prediction framework using data from 4,697 individuals discharged from four rehabilitation centers in Eastern China (2018–2022). Four supervised machine learning models were evaluated via nested cross-validation. A transparent Logistic Regression model provided the most stable discrimination (AUC = 0.735) and calibration, outperforming complex ensemble algorithms. SHapley Additive exPlanations (SHAP) revealed that relapse risk was primarily driven by modifiable psychosocial factors—specifically elevated impulsivity, emotional dysregulation, and injection drug use history—while stronger family support consistently reduced relapse probability, regardless of addiction chronicity. These findings demonstrate the utility of explainable AI in facilitating risk stratification within resource-constrained correctional settings, promoting a shift from generalized supervision to personalized, evidence-based discharge planning.

Keywords

Substance Use Disorder; Relapse Prediction; Machine Learning; Logistic Regression; SHAP Analysis

References

[1] Rosenstock I M. The Health Belief Model and Preventive Health Behavior. Health Education Monographs, 1974, 2(4): 354-386. [2] Ajzen I. The theory of planned behavior. Organizational Behavior and Human Decision Processes, 1991, 50(2): 179-211. [3] Savickas M L. Career adaptability: An integrative construct for life-span, life-space theory. Career Development Quarterly, 1997, 45(3): 247-259. [4] McCraty R, Atkinson M. Resilience training program reduces physiological and psychological stress in police officers. Global Advances in Health and Medicine, 2012, 1(5): 44-66. [5] Parletta N, Milte C M, Meyer B J. Nutritional modulation of cognitive function and mental health. The Journal of Nutritional Biochemistry, 2013, 24(5): 725-743. [6] Reuter P R, Forster B L. Student health behavior and academic performance. PeerJ, 2021, 9: e11107. [7] Jia-Yuan Z, Xiang-Zi J, Yi-Nan F, et al. Emotion Management for College Students: Effectiveness of a Mindfulness-Based Emotion Management Intervention on Emotional Regulation and Resilience of College Students. Journal of Nervous & Mental Disease, 2022, 210(9): 716-722. [8] Foster K, Roche M, Giandinoto J, et al. Workplace stressors, psychological well‐being, resilience, and caring behaviours of mental health nurses: A descriptive correlational study. International Journal of Mental Health Nursing, 2020, 29(1): 56-68. [9] Rossomanno C I, Herrick J E, Kirk S M, et al. A 6-Month Supervised Employer-Based Minimal Exercise Program for Police Officers Improves Fitness. Journal of Strength and Conditioning Research, 2012, 26(9): 2338-2344. [10] Jay Dawes J, Lopes Dos Santos M, Kornhauser C, et al. Longitudinal Changes in Health and Fitness Measures among State Patrol Officers by Sex. Journal of Strength and Conditioning Research, 2023, 37(4): 881-886. [11] Lin P Y, Tseng P, Liang W M, et al. The mediating effect of health behaviors on the association between job strain and mental health outcome: a national survey of police officers. Scientific Reports, 2024, 14(1): 10015. [12] Fekedulegn D, Burchfiel C M, Charles L E, et al. Shift work and sleep quality among urban police officers. Journal of Occupational and Environmental Medicine, 2016, 58(3): e66-e71. [13] Finney C, Stergiopoulos E, Hensel J, et al. Organizational stressors associated with job stress and burnout in correctional officers: a systematic review. BMC public health, 2013, 13: 82. [14] Baker L D, Berghoff C R, Kuo J L, et al. Associations of Police Officer Health Behaviors and Subjective Well-Being: The Role of Psychological Flexibility. European Journal of Health Psychology, 2020, 27(3): 98-108. [15] Purba A, Demou E. The relationship between organisational stressors and mental wellbeing within police officers: A systematic review. BMC Public Health, 2019, 19(1): 1286. [16] Gong Z, Yang J, Gilal F G, et al. Repairing Police Psychological Safety: The Role of Career Adaptability, Feedback Environment, and Goal-Self Concordance Based on the Conservation of Resources Theory. SAGE Open, 2020, 10(2) [17] Violanti J M, Owens S L, Fekedulegn D, et al. An Exploration of Shift Work, Fatigue, and Gender among Police Officers: The BCOPS Study. Workplace Health and Safety, 2018, 66(11): 530-537. [18] Lambert E G, Qureshi H, Frank J, et al. Job Stress, Job Involvement, Job Satisfaction, and Organizational Commitment and Their Associations with Job Burnout Among Indian Police Officers: a Research Note. Journal of Police and Criminal Psychology, 2018, 33(2): 85-99.