Development and Validation of a Machine Learning‑Based Predictive Model for Assessing the Risk of Comorbid Depression...

Area of Science:

• Medical Informatics
• Computational Psychiatry
• Pulmonology

Background:

• Asthma is frequently associated with comorbid depression, impacting patient quality of life.
• Early identification of depression risk in asthma patients is crucial for timely intervention.
• Machine learning offers potential for developing predictive models for complex health conditions.

Purpose of the Study:

• To develop and validate a machine learning model for predicting comorbid depression risk in asthma patients.
• To identify key predictors of depression in this population.
• To provide a tool for early clinical risk assessment.

Main Methods:

• A retrospective analysis of 2464 asthma patients from NHANES data was performed.
• Feature selection utilized Boruta and LASSO algorithms.
• Eight machine learning models were trained and evaluated using 5-fold cross-validation, with XGBoost selected as the top performer.

Main Results:

• The XGBoost model achieved an AUC of 0.750, with 69.1% accuracy, 68.2% sensitivity, and 73.8% specificity.
• Key predictors identified included hypertension, COPD, stroke, sleep questionnaire scores, smoking status, PIR, and education level.
• SHAP analysis highlighted sleep questionnaire, PIR, and education as primary predictive factors.

Conclusions:

• The developed XGBoost model effectively predicts depression risk in asthma patients.
• This model serves as a valuable reference for early clinical identification and intervention strategies.
• Machine learning approaches can enhance the management of comorbid conditions in asthma care.