Integrated risk prediction for opioid overdose: combining genetic and socioeconomic factors

A multidimensional prediction model integrating genetic and socioeconomic factors was developed and validated to assess the risk of opioid overdose, aiming to facilitate individualized risk estimation and inform early clinical identification and preventive strategies. In this retrospective study, 340 patients with a history of opioid use were enrolled. Participants were randomly allocated to a training set ( n = 238) or a validation set ( n = 102) in a 7:3 ratio. Candidate variables, including genetic variants, socioeconomic determinants, clinical parameters, and comorbidities, were initially screened using univariate analysis. Subsequently, least absolute shrinkage and selection operator regression and multivariable logistic regression were applied for variable selection and identification of independent predictors. Four machine learning models—random forest, support vector machine, k-nearest neighbors, and gradient boosting machine –were constructed using the final seven predictors. Model performance was evaluated based on the area under the receiver operating characteristic curve (AUC), calibration curves, and decision curve analysis (DCA). Model interpretability was enhanced using SHapley Additive exPlanations (SHAP) values. Seven independent risk factors for opioid overdose were identified ( P < 0.05). Among these, a history of prior overdose showed the strongest association (odds ratio = 3.192). The random forest model demonstrated superior performance, with an AUC of 0.879 in the training set and 0.774 in the validation set. Its calibration curve was closest to the ideal line. Furthermore, DCA indicated a higher net benefit for the random forest model across a 10–30% risk threshold range. SHAP analysis confirmed daily morphine milligram equivalents and prior overdose history as the most contributory features, aligning with the logistic regression results. An integrated risk prediction model for opioid overdose was successfully developed and validated. The random forest algorithm exhibited optimal discriminative ability, calibration, and clinical utility, providing a tool for individualized risk assessment to guide precision prevention.