|
|
A New Chapter in Outcome Analysis for Congenital Heart Surgery: Empirically Based Stratification of Mortality Risk
D. R. Clarke1, S. M. O'Brien2, J. P. Jacobs3, M. L. Jacobs4, F. G. Lacour-Gayet5, K. F. Welke6, B. Maruszewski7, Z. Tobota8, W. J. Miller9, L. Hamilton10, C. Mavroudis11, F. H. Edwards12. 1The Children's Hospital, Denver, CO, 2Duke Clinical Research Institute, Durham, NC, 3The Congenital Heart Institute of Florida (CHIF),, Tampa, FL, 4St. Christopher Hospital for Children, Philadelphia, PA, 5The Children's Hospital Heart Institute, Denver, CO, 6Oregon Health and Science University, Portland, OR, 7Memorial Hospital Child's Health Centre, Warsaw, Poland, 8Children's Memorial Health Institute, Warsaw, Poland, 9Rho, Inc, Chapel Hill, NC, 10Freeman Hospital, Tyne, United Kingdom, 11Children's Memorial Hospital, Chicago, IL, 12University of Florida, Jacksonville, FL,
BACKGROUND:
The analysis of congenital heart surgery outcomes requires a method of grouping operations with similar risk of adverse outcomes. The two major systems in current use are based on estimates of risk or complexity that were subjectively derived. Our goal was to create an objective, empirically based index that can be used to identify procedure groups with similar statistically estimated risk of in-hospital mortality.
METHODS:
Mortality risk was estimated for 145 congenital procedures using combined data (77,913 patients) from the European Association of Cardiothoracic Surgeons Congenital Heart Surgery Database (33,447 patients operated between January 2002 and April 2007) and the Society of Thoracic Surgeons Congenital Heart Surgery Database (44,466 patients operated between January 2002 and December 2006). Operations involving two or more procedures done concurrently were classified using the most technically complex procedure as determined by the surgical difficulty component of the Aristotle Basic Complexity score. Ties were adjudicated using the average difficulty rank of six surgeons. Mortality rate estimates were calculated using a Bayesian model that accounted for uncertainty due to rare outcomes and incorporated prior information from an expert panel. Procedures were sorted by increasing risk and then grouped into data driven mortality levels. The chosen levels were optimal in minimizing the within-level variation and maximizing the between-level variation. An unbiased assessment of future model performance was obtained using the method of bootstrap resampling. Discrimination of the final mortality grouping was assessed by the C-statistic.
RESULTS:
Statistically estimated mortality rates ranged across procedures from 0.3% (ASD repair, Patch) to 24.6% (Transplant, Heart and lung). The proposed stratification has 5 levels which capture over 95% of the between-procedure variation in mortality. The average mortality rate within each risk level is depicted in Table I. Bootstrap validation showed good discrimination (bias-corrected C-statistic = 0.756).
CONCLUSIONS:
The proposed mortality index levels have a high degree of within-category homogeneity and high discrimination for predicting mortality. Stratification based on these levels will facilitate equitable comparison of mortality outcomes across institutions with differing case mix.| Index Level | # of procedures | # of patients | # of deaths | Mortality Rate | | 1 | 20 | 24366 | 206 | 0.8% | | 2 | 39 | 22936 | 667 | 2.9% | | 3 | 37 | 12172 | 660 | 5.4% | | 4 | 42 | 15392 | 1679 | 10.9% | | 5 | 7 | 3047 | 719 | 23.6% |
Back to 2008 Annual Meeting
Back to Program Outline
|
