Machine learning models using baseline clinical data predicted WHO functional class at 12 months (r=0.401) to estimate healthcare costs in pulmonary arterial hypertension.
Observational (n=181)
No
Machine learning models using routinely collected baseline clinical data can predict 12-month WHO functional class and estimate healthcare costs in patients with pulmonary arterial hypertension.
Estimación del efecto: r = 0.401
Objective: To develop and validate machine learning (ML) models that forecast healthcare costs in pulmonary arterial hypertension (PAH) by predicting World Health Organization functional class (WHO-FC) at 12-month follow-up using baseline clinical data. Design and method: This pilot study analyzed 181 patients with invasively confirmed PAH. A total of 56 baseline variables, including demographics, echocardiographic and hemodynamic parameters, laboratory tests, and functional assessments, were used to train six ML models comprising regularized and tree-based approaches. The primary outcome was WHO-FC at 12 months after therapy initiation. Predicted WHO-FC values were linked to published real-world per-patient-per-month (PPPM) healthcare cost data, stratified by WHO-FC (I–IV) for pharmacy, medical, and total expenditures, and inflation-adjusted to 2024. Models were trained using an 80/20 train-test split and validated with five-fold cross-validation. Results: Observed WHO-FC improved from 2. 9 ± 0. 4 at baseline to 2. 3 ± 0. 5 at 12-month follow-up, while mean total PPPM costs remained stable (12, 329 vs. 12, 332). Among all models, k-nearest neighbors demonstrated the best predictive performance with r = 0. 401, R2 = 0. 161, RMSE = 0. 125, and MAE = 0. 106. Predicted total costs across ML models ranged from 12, 481 to 12, 614. Conclusions: ML-based prediction of WHO-FC enables early estimation of healthcare costs and supports economic planning in PAH using routinely collected baseline data. This proof-of-concept approach aligns cost forecasting with clinical treatment targets. Multicenter validation is required to assess generalizability before integration into decision-support workflows.
Tilmann et al. (Fri,) conducted a observational in pulmonary arterial hypertension (n=181). Machine learning models using baseline clinical data was evaluated on WHO-FC at 12 months after therapy initiation (r = 0.401). Machine learning models using baseline clinical data predicted WHO functional class at 12 months (r=0.401) to estimate healthcare costs in pulmonary arterial hypertension.