A MACHINE-LEARNING MODEL FOR PREDICTING DEPRESSION SYMPTOMS SEVERITY IN A SAMPLE OF YOUNG ADULTS: A CROSS-SECTIONAL STUDY FROM QATAR BIOBANK (QBB)

Abstract

Background: Depression is a major public health issue, particularly in the MENA region, affecting the quality of life and overall productivity of young adults. Recent advances in machine learning (ML) offer new methodologies for predicting mental health outcomes based on large-scale data. Aim: This study utilizes a machine-learning approach to predict the severity of depression symptoms among young adults using data from the Qatar Biobank (QBB). Methods: A cross-sectional analysis of 2,660 young adults aged 18–36 years from the QBB was performed. Depression symptoms severity (DSS) was assessed using the Patient Health Questionnaire-9 (PHQ-9), with scores ≥10 indicating significant symptoms severity. We applied Logistic Regression, Random Forest and Decision Tree ML models, evaluating their performance through accuracy metrics including the area under the curve (AUC) and sensitivity. Results: The study found a (70%) prevalence of DSS among participants, with a significant proportion being Qataris (58%). Risk factors reported by Logistic Regression include employment status, family size, income, and night shift working, participation in intense sports, weekday screen time, receiving phone calls, cohabitation with smokers, consumption of coffee and Karak, physical health indicators, self-rated health, weight changes, the use of vision aids, Whereas lifestyle variables were mainly protective factors. Machine Learning models, particularly Random Forest, showed excellent predictive capabilities (AUC: 0.948 ± 0.013), with variables such as self-rated health, indoor shisha exposure, and caffeine intake playing significant roles in symptoms severity prediction. Conclusions: The use of Machine Learning models in this study demonstrates significant potential in predicting depression symptoms severity, which can aid in early identification and personalized intervention strategies. The findings are particularly valuable for healthcare providers and policymakers in designing and implementing effective mental health strategies that fit with local cultural and social context. Future research should aim to refine these models and explore their integration into broader healthcare systems for better mental health management. More qualitative research is needed to understand the sociocultural and environmental factors related to DSS.