Refining The Society of Thoracic Surgeons Congenital Heart Surgery Database Mortality Risk Model With Enhanced Risk Adjustment for Chromosomal Abnormalities, Syndromes, and Noncardiac Congenital Anatomic Abnormalities

• Published in: The Annals of thoracic surgery Vol. 108; no. 2; pp. 558 - 566
• Main Authors: Jacobs, Jeffrey P., O’Brien, Sean M., Hill, Kevin D., Kumar, S. Ram, Austin, Erle H., Gaynor, J. William, Gruber, Peter J., Jonas, Richard A., Pasquali, Sara K., Pizarro, Christian, St. Louis, James D., Meza, James, Thibault, Dylan, Shahian, David M., Mayer, John E., Jacobs, Marshall L.
• Format: Journal Article
• Language: English
• Published: Elsevier Inc 01.08.2019
• ISSN: 0003-4975; 1552-6259
• DOI: 10.1016/j.athoracsur.2019.01.069

The Society of Thoracic Surgeons Congenital Heart Surgery Database Mortality Risk Model adjusts not only for procedure and age group pairings but also for additional patient factors, including the binary presence or absence of a chromosomal abnormality (CA), syndrome (S), and/or a noncardiac congenital anatomic abnormality (NCAA). This analysis refines case-mix adjustment by adding more granular adjustment for individual conditions (CA, S, and NCAA), consistent with a hypothesis that associated risk of mortality differs between individual conditions. CA/S corresponding to the same condition were merged to a single condition code. Odds ratios were estimated for all CA/S. For CA/S associated with at least 10 deaths in neonates and infants and at least 10 deaths in children and adults, odds ratios were estimated for the effect of the CA/S separately in neonates/infants and in children/adults. In addition to these condition/age interactions, condition/age/procedure interactions were explored (eg, effect of Down syndrome was estimated based on age and procedure subgroups, including atrioventricular canal repair and single-ventricle palliation). Bayesian modeling was used to create 5 maximally homogeneous groups of CA/S from 81 candidate CA/S variables. A standard logistic regression model then incorporated indicator variables for the 5 categories of CAs/Ss, 7 unique NCAAs, and all other covariates in the previously published Society of Thoracic Surgeons Congenital Heart Surgery Database Mortality Model. Analysis included 107,062 operations in 100 centers (2010 to 2015). Operative Mortality was 3,629 (3.4%). In the development sample, the C statistics of the original nonaugmented model and the augmented model were 0.872 and 0.875, respectively. The Society of Thoracic Surgeons Congenital Heart Surgery Database Mortality Risk Model has been augmented by addition of covariates representing individual CAs, Ss, and NCAAs.