A Machine learning model to predict substance abuse relapse in Nairobi County, Kenya

Abstract

Substance use disorder (SUD) remains a critical public health issue in Nairobi County, Kenya, where high relapse rates undermine efforts to achieve long-term recovery for individuals affected by addiction. Despite the availability of treatment programs and rehabilitation centers, relapse rates continue to exceed 50% within the first year of recovery, indicating the limitations of current intervention strategies. The complexity of relapse, driven by psychological, social, and biological factors, requires a more sophisticated approach to prediction and prevention. This research explored how machine learning could be applied as a predictive tool to identify individuals at risk of relapses. Specifically, the study aimed to enhance intervention approaches and the recovery results in Nairobi County. The data used in the study was collected from a secondary data source (Kaggle) with features that include several variables, such as demographic information, substance education, behavioral factors, psychological evaluations, and socioeconomic indicators. The data collected was then used to train and validate the machine learning model, which utilized predictive algorithms such as decision trees, support vector machines, random forest, Artificial Neural Network and the K-nearest neighbors to determine the probability of relapse. Eventually, the study evaluated the model's performance by utilizing the standard performance metrics. By developing this predictive model, the study anticipated handling numerous major challenges in the management of substance abuse disorder in Nairobi County. First, by identifying individuals at high risk of relapse, the model will enable healthcare providers to implement targeted interventions, thereby reducing relapse rates and improving the chances of sustained recovery. Second, the model's predictions can inform resource allocation, ensuring limited healthcare resources are directed toward those most in need. Third, the research will contribute to public health policy by providing data-driven insights that can be used to design more effective prevention and treatment programs. The developed models in this research performed well with the adopted and implemented model, Artificial Neural Network achieving an accuracy of ~0.9774. Keywords: Substance Use Disorder (SUD), Machine Learning Model, Relapse.