Blood Marker-Based Machine Learning Model for Survival Prediction in Patients with Pneumoconiosis: Construction and...

Area of Science:

• Pulmonary Medicine
• Occupational Health
• Biomarkers

Background:

• Pneumoconiosis poses a significant health risk, particularly in occupational settings.
• Effective short-term prognosis prediction is crucial for timely patient management.
• Inflammatory markers are increasingly recognized as potential prognostic indicators in respiratory diseases.

Purpose of the Study:

• To investigate the association between blood inflammatory markers (CRP, IL-6) and short-term pneumoconiosis prognosis.
• To develop and validate a multifactorial prediction model for pneumoconiosis prognosis.
• To identify key clinical and inflammatory predictors of adverse outcomes in pneumoconiosis patients.

Main Methods:

• Retrospective collection of clinical data from 813 pneumoconiosis patients.
• Utilized LASSO regression for variable screening and logistic regression for model construction.
• Validated the model using Cox regression, Kaplan-Meier curves, ROC analysis, and decision curve analysis.

Main Results:

• Age, dust exposure duration, dyspnea, SpO2, neutrophil count, CRP, and IL-6 were identified as predictors.
• CRP and IL-6 were independent risk factors for poor short-term prognosis (P<0.001).
• The prediction model demonstrated high discrimination (AUCs 0.905-0.920) and clinical utility.

Conclusions:

• A prediction model incorporating CRP, IL-6, and clinical factors effectively identifies high-risk pneumoconiosis patients.
• The model offers a reliable basis for early intervention strategies.
• High CRP and IL-6 levels are significantly associated with reduced short-term survival in pneumoconiosis.