Developing and validating machine learning-based prediction models for frailty occurrence in those with chronic...

Area of Science:

• Gerontology
• Pulmonology
• Medical Informatics

Background:

• Frailty is a syndrome of decreased physiological function, increasing vulnerability in patients with chronic obstructive pulmonary disease (COPD).
• COPD patients face unique physical and psychological burdens that exacerbate frailty risk.
• Early identification and prediction of frailty are crucial for managing COPD patients.

Purpose of the Study:

• To determine the prevalence of frailty in COPD patients.
• To develop and evaluate a reliable prediction model for frailty risk in COPD.
• To improve clinical identification and prediction of frailty in this population.

Main Methods:

• Utilized data from the 2018 China Health and Retirement Longitudinal Study (CHARLS).
• Employed machine learning techniques including XGBoost, Random Forest, and Logistic Regression.
• Developed an online predictive risk modeling website with SHAP interpretations for customized risk assessment.

Main Results:

• Identified depression, smoking, gender, social activities, dyslipidemia, asthma, and residence type as key predictors of frailty in COPD.
• The XGBoost model demonstrated high predictive power with an AUC of 0.942, accuracy of 0.915, sensitivity of 0.873, and specificity of 0.911.
• The developed model effectively distinguished between frail and non-frail COPD patients.

Conclusions:

• The machine learning predictive model serves as a valuable tool for assessing frailty risk in COPD patients.
• The model can assist clinicians in screening and identifying high-risk individuals for timely intervention.
• This approach enhances the management of frailty in the context of chronic obstructive pulmonary disease.