Risk prediction for early in-hospital mortality following heart transplantation in the United States.

Background: Risk factors for early mortality after heart transplant (HT) have not been used for quantitative risk prediction. We sought to develop and validate a risk prediction model for posttransplant in-hospital mortality in HT recipients.

Methods And Results: We derived the model in subjects aged ≥18 years who underwent primary HT in the United States from January 2007 to June 2009 (n=4248) and validated it internally using a bootstrapping technique (200 random samples, n=4248). We then assessed the model's performance in patients receiving an HT from July 2009 to October 2010 (external validation cohort, n=2346). Posttransplant in-hospital mortality was 4.7% in the model derivation cohort. The best-fitting model based on recipient characteristics at transplant had 6 variables: age, diagnosis, type of mechanical support, ventilator support, estimated glomerular filtration rate, and total serum bilirubin. Model discrimination for survivors versus nonsurvivors was acceptable during derivation and internal validation (C statistic, 0.722 and 0.731, respectively) as was model calibration during derivation (Hosmer Lemeshow [HL] P=0.47). Model performance was reasonable in the external validation cohort (predicted mortality, 4.9%; actual mortality, 4.3%; R(2)=0.95; C statistic, 0.68; HL P=0.48). Adding the donor-related variables of age and ischemic time to the model improved its performance in both the model derivation (C statistic, 0.742; HL P=0.70) and the external validation (C statistic, 0.695; HL P=0.42) cohorts.

Conclusions: The proposed model allows risk stratification of HT candidates for early posttransplant mortality and may be useful in counseling patients with regard to their posttransplant prognosis. The model with additional donor-related variables may be useful during donor selection.