Introduction: Pulmonary Endarterectomy (PEA) is the gold-standard treatment for chronic thromboembolic pulmonary hypertension (CTEPH). There remains an unavoidable operative risk with PEA, even in the best centres. Very limited data exists on factors which influence post-PEA mortality and patient-reported outcomes (PRO). The ability to identify those at greatest operative risk has the potential to inform both patient selection and choice

Methods: Consecutive CTEPH patients undergoing PEA between 2007 and 2018 were included (n=1266). Predictors included an intentionally broad array of demographic, functional and physiological measures. Three statistical models were considered for each outcome: 90-day mortality (90DM), 5-year mortality (5YM) and PRO improvement. Models were compared to PEA surgeon-derived predictions and those of EuroSCORE II, the most widely implemented cardiac surgery risk tool. Best fit models were incorporated into our risk tool and validated using a separate prospective PEA cohort (2018 ? 2021; n = 443).

Results: Random forest models had the greatest predictive accuracy for all three outcomes. Our models had good discriminatory ability for 90DM (AUROC 0.817) and 5YM (AUROC 0.810), outperforming surgeon-derived (AUROC 0.587) and EuroSCORE II (AUROC 0.653) predictions. Pre-operative haemodynamics were important in 90DM but not 5YM where cardiovascular disease risk factors dominated. PRO improvement was predicted less well from standard pre-operative measures (AUROC 0.470).

Conclusion: Outcomes from PEA can be predicted pre-operatively to a clinically useful degree. Our validated models enable individualised risk stratification to better inform shared decision making