Forecasting Long-Term Drinking Status for Individuals Treated for Alcohol Dependence: A Bayesian Approach

Abstract Background Alcohol dependence presents a significant challenge, with treatment efforts often falling short in preventing relapse due to the nature of the condition. Ensuring sustained long-term recovery among successfully treated individuals is of paramount clinical importance. Therefore, this study aims to predict the long-term (after 12 months of successful treatment) alcohol consumption status of individuals treated for alcohol dependence, utilizing baseline characteristics and initial treatment response information. Methods A cohort study design was employed to assess both short and long-term alcohol consumption status among male participants who had undergone successful treatment for alcohol dependence at an addiction medicine outpatient department in Southern India from July 2020 to December 2021. Initially, participants were monitored for four months to observe their short-term treatment response, followed by a telephonic follow-up to assess their long-term status. This study employs the application of the Bayesian prediction model for evaluating the long-term binary outcome. Results Among four distinct fitted predictive models, the one encompassing data on initial four-month alcohol consumption, age, marital status, illness duration, and baseline clinical global impression-severity score emerged as predictors for long-term treatment outcomes. This model showed around 70% accuracy on the test dataset, with sensitivity and specificity hovering around 70%. Conclusion The findings emphasize the importance of maintaining alcohol abstinence or controlled drinking in the fourth month after successful treatment to ensure long-term success. Additionally, the results underscore the necessity of exploring variables beyond baseline information to understand factors influencing long-term outcomes comprehensively.