FRIDAY, Feb. 6, 2026 (HealthDay News) -- Adults with congenital heart disease (CHD) have a persistently high risk for cardiac reoperation, according to a study presented at the annual meeting of the Society of Thoracic Surgeons (STS), held from Jan. 29 to Feb. 1 in New Orleans.Elaine Griffeth, M.D., from the Mayo Clinic in Rochester, Minnesota, and colleagues validated previously derived machine learning (ML) and logistic regression models using data from a high-volume adult CHD surgery practice in a national database. In a prior study, an accurate and clinically applicable alternative to the previously published STS adult congenital mortality risk model was created. These same models were then validated externally in the STS Adult Cardiac Surgery Database (STS ACSD) for adult patients with CHD undergoing cardiac reoperation.The researchers found that operative mortality was 6.6 percent: 4.4 and 12.9 percent in elective and urgent/emergent procedures, respectively. The composite outcome occurred in 16.7 percent of patients. Compared with the institutional database, which had operative mortality of 1.8 percent and composite outcome of 8.8 percent, morbidity/mortality was higher in STS ACSD. In STS ASCD, the model performance/area under the curve was 0.674 and 0.651 for the institutional multivariable logistic regression model and the institutional ML model, respectively. Systematic underprediction was observed, with event rates exceeding mean predicted probabilities in every decile. In model performance, five of seven predictors remained influential. Between the datasets, the key differences were a greater number of urgent/emergent procedures and fewer patients with functionally univentricular physiology."We want to have high reliability in the surgeries we are offering, and we are trying to tailor this model with data from past patients," Griffeth said in a statement. "The more informed patients are about their risks for surgery, the better."Press ReleaseMore Information.Sign up for our weekly HealthDay newsletter