Machine learning-based prediction of depression severity among medical students

Abstract

Depression is a significant global cause of disability, yet it often remains undetected and untreated. Medical students have a higher likelihood of experiencing depression compared to many other age-matched groups. Previous studies have shed light on factors contributing to depression among college students. However, medical students face understudied unique experiences and demands. This gap hinders the ability to apply previously developed models to predict depression in medical students, limiting the effectiveness of mental health interventions in this population. Our study addresses this gap by presenting a machine learning model that captures the complex interplay of factors influencing the severity of depression in this population. Using data from the Healthy Minds Study, we found that a sequence of binary XGBoost models performed best at predicting the severity of depression as recorded through the Patient Health Questionnaire-9. Our modeling approach identified that prior diagnosis of depression and eating disorder, current financial stress, and younger age had the highest influence on predicted depression severity. The findings of this study hold great promise for enhancing medical education by more effectively addressing students’ depression. Our predictive model may enable educators and administrators to identify students at higher risk, facilitating timely interventions and support.