Abstract 21567: Development and Validation of a Risk-Prediction Model for 30-day Waiting-List Mortality Among Children Listed for Heart Transplant in the US
Background: Risk factors for waiting list mortality in children listed for transplant (HT) have been previously described. However, a quantitative risk-prediction model based on these factors has never been validated. The purpose of the study was to validate a quantitative risk prediction model for 30-day waiting list mortality, a critical time interval during which two-thirds of wait-list deaths occur.
Methods: We developed and validated a risk-prediction model for 30-day waiting list mortality for children using data from the Organ Procurement and Transplant Network (OPTN). All children <18 years of age who were listed for primary HT in the US between July 2004 and September 2009 were included. Logistic regression analysis was used to identify risk factors for 30-day mortality and to develop a prediction model using two-thirds of the overall cohort. This model was then applied to the remaining cohort for validation.
Results: Of 2,355 children who met the inclusion criteria, the model was developed using a random sample of 1,578 (two-thirds) children (with 10.3% 30-day mortality) and applied to the remaining 777 children (one-third) for validation. The best predictive model had 4 clinical factors retained as categorical variables: Hemodynamic support [ECMO (Odd ratio, OR 6.5), ventilator (OR 2.9), BIVAD (OR 5.3), LVAD (OR 2.2) (reference category, medical therapy), congenital heart disease diagnosis (OR, 2.1), Listing status 2 (OR 0.48) and creatinine clearance <40 ml/min/1.73m2 The C-statistic (0.82) and the Hosmer-Lemeshow goodness-of-fit (P=0.57) statistics in the model-development cohort were replicated in the validation cohort (C-statistic 0.80 and the Hosmer-Lemeshow goodness-of-fit P=0.76) suggesting excellent prediction for 30-day waiting list mortality using the model.
Conclusions: This risk prediction model using 4 baseline clinical factors at the time of listing has excellent prediction characteristics for 30-day mortality. The model may be useful to support decision-making around transplant list, timing of mechanical support and potentially organ allocation policy.
- © 2010 by American Heart Association, Inc.