Identifying depression in the National Health and Nutrition Examination Survey data using a deep learning algorithm

Area of Science:

• Computational psychiatry
• Epidemiology
• Machine learning in healthcare

Background:

• Depression is a leading global cause of disability, necessitating accurate epidemiological assessments.
• Estimating depression prevalence and risk factors remains challenging using traditional survey methods.
• Deep learning offers potential for analyzing complex health data to understand depression.

Purpose of the Study:

• To assess the efficacy of deep-learning and machine-learning classifiers in identifying depression.
• To evaluate the predictive performance of these algorithms across different datasets and timeframes.
• To explore the cross-temporal and cross-national generalizability of deep learning for depression detection.

Main Methods:

• Utilized customized deep-neural-network and machine-learning classifiers.
• Analyzed survey data from the US National Health and Nutrition Examination Survey (NHANES) (1999-2014).
• Assessed data from the South Korea National Health and Nutrition Examination Survey (K-NHANES) in 2014.

Main Results:

• Deep learning achieved high accuracy in detecting depression (AUCs of 0.91 in NHANES, 0.89 in K-NHANES).
• The algorithm accurately predicted future depression prevalence (AUC of 0.92).
• Machine learning models, including logistic regression (AUC 0.77), also showed predictive capabilities in cross-national data.

Conclusions:

• Deep neural networks effectively identified depression from health and demographic factors in both datasets.
• The deep-learning algorithm demonstrated strong predictive performance across time and national borders.
• Further research is warranted to explore clinical applications of AI in mental illness detection and risk factor analysis.