A machine learning model for the prediction of unhealthy alcohol use among women of childbearing age in Alabama

Abstract

Introduction: This study utilizes a machine learning model to predict unhealthy alcohol use treatment levels among women of childbearing age.

Methods: In this cross-sectional study, women of childbearing age (n = 2397) were screened for alcohol use over a 2-year period as part of the AL-SBIRT (screening, brief intervention, and referral to treatment in Alabama) program in three healthcare settings across Alabama for unhealthy alcohol use severity and depression. A support vector machine learning model was estimated to predict unhealthy alcohol use scores based on depression score and age.

Results: The machine learning model was effective in predicting no intervention among patients with lower Patient Health Questionnaire (PHQ)-2 scores of any age, but a brief intervention among younger patients (aged 18-27 years) with PHQ-2 scores >3 and a referral to treatment for unhealthy alcohol use among older patients (between the ages of 25 and 50) with PHQ-2 scores >4.

Conclusions: The machine learning model can be an effective tool in predicting unhealthy alcohol use treatment levels and approaches.

Keywords: artificial intelligence; depression; machine learning; unhealthy alcohol use; women.