Social media use and depression across 11 years of U.S. adult data

Abstract

Controversy surrounds the relationship between social media use and depression. Prior research has relied on single datasets, narrower age groups or broad measures of screen time, leaving uncertainty about whether associations of social media use and depression were stable across repeated populations over time. Using 11 repeated cross-sectional surveys (2014–2024) of adults aged 18–70 (total N = 183,000), we examined whether social media use (SMU) was consistently associated with binary self-reported depression (SRD) and whether it contributed to classification of SRD alongside demographics and other types of media use variables. For each year, we trained balanced random forest (BRF) models and evaluated discrimination using Area Under the Receiver Operating Characteristic curve (AUROC), calibration (Brier score), and feature importances. BRF predicted SRD with accuracies between 0.71 and 0.75 and AUROCs between 0.65 and 0.73 for each of 11 datasets. SHAP dependence plots showed that the contribution of social media use to the model classified depression shifted from neutral to positive at approximately 150 minutes per day. These findings identify a stable, repeated association between higher social media use and SRD. This work supports further longitudinal and experimental work on whether high exposure levels represent a clinically meaningful risk.