Predicting Depression in Community Dwellers Using a Machine Learning Algorithm

Area of Science:

• Public Health
• Computational Psychiatry
• Machine Learning in Healthcare

Background:

• Depression is a major global cause of disability, necessitating effective community screening.
• Socioeconomic factors significantly impact depression prevalence and burden.

Purpose of the Study:

• To develop and validate a machine learning model for depression screening in community-dwelling populations.
• To identify key predictive variables for depression using feature selection techniques.

Main Methods:

• Utilized data from the Korea National Health and Nutrition Examination Surveys (2014 and 2016).
• Employed the synthetic minority oversampling technique (SMOTE) for class imbalance and LASSO for feature selection and classification.
• Validated the model on a hold-out test set of 9488 participants.

Main Results:

• The LASSO model selected 13 variables from an initial 37.
• The model achieved an area under the receiver operating characteristic curve of 0.903 and an accuracy of 0.828 on the test set.
• Perceived stress emerged as the most influential variable in classifying depression.

Conclusions:

• The LASSO method is practical for developing efficient depression screening tools for community dwellers.
• The findings highlight the importance of perceived stress in depression classification.
• Further research is recommended to enhance classification model efficiency and accuracy.