Mortality risk stratification is crucial for managing Eisenmenger syndrome, a severe form of congenital heart disease with pulmonary arterial hypertension. Some mortality determinants have been identified but a risk score remains to be validated.

This study aimed to develop and validate the AI-MUSES risk score, an artificial intelligence (AI)-assisted tool for mortality risk stratification in these patients.

This study included 1,634 adult patients (> 16 years) with Eisenmenger syndrome from two independent multicentre international cohorts (development cohort: 1,098 patients; validation cohort: 536 patients). The primary outcome was all-cause mortality. Using traditional Cox regression models, the MUSES-1 and MUSES-2 scores incorporated variables such as age, oxygen saturation (SaO ₂ ), pericardial effusion, shunt type, cardiac rhythm, and 6-minute walk test (6MWT) distance.

An AI-driven model (AI-MUSES) was developed with variables selected using the Random Survival Forest tool, and the fractional polynomial method applied to refine and transform predictors. Four key predictors were retained: pre-tricuspid shunt, SaO ₂ < 78%, presence of pericardial effusion and 6-minute walk test distance < 434 m. The AI-MUSES score demonstrated excellent discrimination, with a c-statistic of 0.756 in the development cohort and 0.762 in the validation cohort. Cross validation in the develoment cohort demonstrated excellent discrimination with a c-statistic of 0.741 (0.060) on test-train. Kaplan–Meier survival curves confirmed its ability to stratify patients into three distinct risk categories: low risk (requering follow up), intermediate risk (referral to transplantation centers) and high risk (transplantation) ( Fig. 1 ).

The AI-MUSES score provides a simple, reliable, and externally validated risk stratification tool for patients with Eisenmenger syndrome. By identifying patients at high risk of mortality, the score could facilitate timely referral for transplantation and enhances personalized management strategies.