Answers are grounded in this paper and cite their sources.
Objective: To develop a machine learning-based clinical prediction model for assessing the risk of pulmonary arterial hypertension (PAH) in patients with systemic sclerosis (SSc), with the aim of enabling early intervention and improving patient prognosis and quality of life. Methods: A retrospective study was conducted, including 65 SSc patients and 20 SSc patients with PAH diagnosed between January 2018 and June 2023. A total of 333 clinical and laboratory parameters were collected as potential predictors. Feature selection was performed using the Boruta algorithm and LASSO regression, with overlapping variables used to construct a multivariable logistic regression model, which was visualized using a nomogram. Model performance was evaluated using the C-index, ROC curve, calibration curve, and decision curve analysis. Internal validation was conducted using the K-nearest neighbors (KNN) algorithm, with ROC AUC, precision-recall curve (PR curve), and confusion matrix as evaluation metrics. Results: Five key predictive factors—age, TNF-α, interstitial lung disease (ILD), impaired pulmonary function, and chest tightness—were identified. The model demonstrated excellent performance with a C-index of 0.94 and an AUC of 0.943. Calibration curves showed good consistency, and decision curve analysis indicated maximal net benefit within a threshold probability range of 0.1 to 0.75. As part of internal validation, KNN validation yielded an AUC of 0.9362 and an F1-score of 0.79, with stable performance observed across 5-fold cross-validation. Conclusion: A highly effective clinical prediction model was successfully developed, capable of identifying SSc patients at risk of developing PAH, particularly suitable for early screening in subclinical cases. The model offers significant clinical utility for targeted interventions, though further validation with larger cohorts and integration of novel biomarkers is warranted.
No takes yet. Share an insight, caveat, or question.
Yan et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: