Risk prediction for surgery on adults with congenital heart disease: Nonprocedural clinical and social factors predominate.

Abstract

STAT categories, the STS-ACHS score, and the Brompton PEACH score have been used for adult congenital heart disease (ACHD) surgical risk prediction. We explored the utility of these risk prediction tools in our 30-year experience, with an emphasis on the most recent 5 years, given the evolving nature of ACHD surgery. A total of 1109 ACHD operations performed between 1995 and 2024 at a single institution were analyzed for in-hospital mortality and time-related survival using logistic regression and multiphase parametric modeling, focusing on STAT, ACHS, and PEACH scores, in addition to >400 patient-specific variables. Final variable selection was guided by bootstrap bagging. Subanalyses focused on the current era, 2019 to 2024. In the full cohort (n = 1109), in-hospital mortality was 1.4% (n = 16 of 1109). A late-phase hazard for death emerged at ∼15 years, and 20-year survival was 81 ± 3%. Compared with the PEACH reference cohort, our experience included more revision, Fontan, and heart failure surgeries (including 82 operations related to advanced therapies). Nevertheless, PEACH showed good discrimination (area under the curve [AUC], 0.78; P < .0001) compared to STAT or ACHS, although it significantly overestimated mortality (O/E ∼0.5). PEACH was a strong independent determinant of time-related survival (P < .0001). Other independent predictors included infective endocarditis, heart failure, >4 prior sternotomies, social concerns, preoperative inotropes, and cyanosis. The nature of the surgical procedure did not feature strongly in risk. In the current era (2019-2024; n = 311), patients had worse/more American Heart Association anatomy/physiology grade, single right ventricle, endocarditis, prior sternotomies, urgent surgery, heart failure and extensive surgical reconstructions (all P = .007-.0001), although in-hospital mortality was not different (1.9%; n = 6 of 311; P = .41). PEACH (AUC, 0.79; P = .04) was superior to STAT or ACHS for predicting in-hospital mortality. However, the perioperative mortality estimate (POME) provided by an ACHD surgeon to patients preoperatively provided the best discrimination in the current era. PEACH and POME (P = .0004) predicted time-related survival; other independent predictors-loss to follow-up, infection, heart failure, >4 prior sternotomies, social concerns, inotropes, Fontan physiology, cyanosis, and the nature of surgical procedures-did not predict risk well. ACHD surgery in quaternary centers is increasingly major revision, endocarditis, and heart failure surgery, hampering conventional risk-prediction models based on procedure. Nonprocedural patient-specific circumstances strongly predict risk; as such, individual clinicians' risk prediction can outperform validated risk models.