Machine learning–based prediction of 12-month change in 6MWD in pulmonary arterial hypertension: a retrospective prediction model development study

Abstract

The six-minute walk distance (6MWD) is a key prognostic marker in pulmonary arterial hypertension (PAH), but predicting individual changes remains challenging. This study aimed to evaluate whether machine learning (ML) models can predict the relative 12-month change in 6MWD using baseline data. In this retrospective, single-centre prediction model development study, we included 181 patients with PAH assessed between 2010 and 2022. Ninety-two baseline variables were used to train six ML algorithms. Data were split into training (80%) and test (20%) sets. Models were developed using five-fold cross-validation with grid search. Performance was assessed using RMSE, MAE, R2, and Pearson correlation (r). Random forest showed the highest correlation and R2 (r= 0.76, R2 = 0.58), while gradient boosting machine achieved a similarly high correlation (r = 0.72). These findings provide preliminary evidence that baseline clinical data may support estimation of the relative 12-month change in 6MWD in patients with PAH. External validation and evaluation of clinical utility are required before clinical implementation.