Understanding drug abstinence self efficacy through statistical analysis, machine learning and explainable AI
Abstract Objective This study explores the socio-demographic and psychological factors influencing Drug Abstinence Self-Efficacy (DASE) through a combined Statistical and Machine Learning (ML) framework, aiming to enhance understanding and improve intervention strategies for individuals with substan...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2025-08-01
|
| Series: | Discover Public Health |
| Subjects: | |
| Online Access: | https://doi.org/10.1186/s12982-025-00879-x |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Abstract Objective This study explores the socio-demographic and psychological factors influencing Drug Abstinence Self-Efficacy (DASE) through a combined Statistical and Machine Learning (ML) framework, aiming to enhance understanding and improve intervention strategies for individuals with substance use disorders. Methods Socio-demographic data sheet and psychological tools are used for collection of data. Statistical analysis is conducted to examine any significant differences in DASE across groups. Various classifiers are used for developing the predictive model that determines DASE. Furthermore, Explainable AI (XAI) technique called Shapley Additive Explanations is used. Results Statistical analyses identified significant associations between DASE and factors such as Locus of Control (LOC), Perceived Social Support (PSS) from family, residence type, abstinence period, family relationships, drug type, education, and employment status. Among the ML models, the best-performing model achieved a classification accuracy of 96.70%, suggesting strong potential for accurately identifying individuals with low DASE who may be at higher risk of relapse. SHAP values enhanced model interpretability by highlighting the relative influence of both expected and less apparent predictors, including Family history of substance use, onset of drug use, and overall social support, thereby offering deeper insights beyond traditional statistical analysis. Conclusion The dual approach—integrating statistical and explainable ML techniques—provides a robust and multidimensional understanding of the factors influencing DASE. The high predictive accuracy and interpretability of the models indicate practical applicability in clinical and community-based addiction recovery settings. This approach supports early identification of high-risk individuals, personalized intervention planning, and aligns with public health goals aimed at promoting long-term recovery and mental well-being. |
|---|---|
| ISSN: | 3005-0774 |