Detecting depression using an ensemble classifier based on Quality of Life scales

Area of Science:

• Computational psychiatry
• Machine learning in healthcare
• Psychosocial factor analysis

Background:

• Major Depressive Disorder (MDD) affects 350 million globally, with traditional diagnosis relying on symptom identification.
• Emerging research explores the link between quality of life (QoL) factors and mental well-being for earlier MDD prediction.
• Current diagnostic tools may benefit from enhanced predictive performance through advanced analytical methods.

Purpose of the Study:

• To propose an ensemble binary classifier for analyzing health survey data to predict MDD.
• To improve machine learning performance on large datasets by identifying associations between QoL scale items and mental illness.
• To enhance the early diagnosis and prediction of MDD through improved predictive accuracy.

Main Methods:

• An ensemble binary classifier was developed to analyze National Health and Nutrition Examination Survey (NHANES) data.
• The classifier was trained using health survey data, with ground truth established by the SF-20 Quality of Life scales.
• Performance was evaluated against the PHQ-9 mental screen inventory.

Main Results:

• The ensemble classifier achieved an F1 score of 0.976 and 95.4% accuracy in predicting MDD.
• Only 4% of instances were misclassified as depressed cases.
• The proposed classifier outperformed baseline algorithms in all measured aspects.

Conclusions:

• The ensemble binary classifier demonstrates superior performance in predicting MDD compared to existing methods.
• This approach offers a promising avenue for earlier and more accurate detection of MDD.
• Future research should explore further applications and refinements of this predictive model.