Mortality Risk Prediction in Patients With Antimelanoma Differentiation-Associated, Gene 5 Antibody-Positive,...

Area of Science:

• Medical research
• Machine learning applications in healthcare
• Pulmonology

Background:

• Antimelanoma differentiation-associated gene 5 antibody-positive dermatomyositis-associated interstitial lung disease (anti-MDA5+DM-ILD) is linked to rapidly progressive interstitial lung disease (RP-ILD) and significant mortality.
• A reliable and accessible prediction model is crucial for evaluating mortality risk in these patients.

Purpose of the Study:

• To develop and validate a machine learning-based risk prediction model for 3-month mortality in anti-MDA5+DM-ILD patients.
• To create an easy-to-use web-based tool for assessing this mortality risk.

Main Methods:

• Retrospective enrollment of 509 patients with anti-MDA5+DM-ILD from 6 Chinese hospitals.
• Application of six machine learning algorithms (XGBoost, LR, LightGBM, RF, SVM, KNN) to construct and evaluate the prediction model.
• Identification of key predictive variables including RP-ILD, ESR, ALB, age, CRP, AST, LDH, and NLR.

Main Results:

• Logistic regression (LR) was selected as the optimal algorithm, demonstrating excellent performance with an AUC of 0.866 in validation and 0.90 in training datasets.
• The model showed strong generalizability with high AUC values in external validation cohorts (0.836 and 0.915).
• A web-based tool was developed integrating the prediction model for practical clinical use.

Conclusions:

• A robust clinical prediction model for 3-month mortality in anti-MDA5+DM-ILD patients was successfully developed.
• The accompanying web tool provides an accessible method for risk assessment and clinical decision-making.