Catheterization for Congenital Heart Disease Adjustment for Risk Method II (CHARM-II)

Tuesday, May 21, 2019
Belmont Ballroom 2-3 (The Cosmopolitan of Las Vegas)
Brian Quinn, M.D. , Children's Hospital - Boston, Dedham, MA
Kimberlee Gauvreau, ScD , Children's Hospital Boston, Boston, MA
Dana R Janssen, M.D., FSCAI , Ayers Children's Medical Center, Nashville, TN
David T. Balzer, M.D., FSCAI , St. Louis Children's Hospital, St. Louis, MO
Susan R. Foerster, M.D., FSCAI , Children's Hospital of Wisconsin, Milwaukee, WI
Wendy Whiteside, M.D. , University of Michigan C.S. Mott Children’s Hospital Congenital Heart Center, Ann Arbor, MI
Bryan H. Goldstein, M.D., FSCAI , Cincinnati Children's Hospital Medical Center, Cincinnati, OH
Darren P. Berman, M.D., FSCAI , Nationwide Children's Hospital, Columbus, OH
Fatima ALI, M.B.B.S. , The Aga khan university and hospital, Karachi, Pakistan
Michael Hainstock, M.D. , University of Virginia, Division of Pediatric Cardiology, Charlottesville, VA
Sara M. Trucco, M.D., FSCAI , The Children's Hospital Of Pittsburgh, Pittsburgh, PA
Alejandro J. Torres, M.D., FSCAI , Morgan Stanley Children's Hospital of New York - Presbyterian Columbia University, New York, NY
Lauren Shirley , Children's Hospital Boston, Boston, MA
Lisa Bergersen, M.D. , Children's Hospital Boston, Boston, MA

Background:
This study sought to improve upon a previously developed risk adjustment methodology for case mix complexity in pediatric cardiac catheterization, allowing for the comparison of adverse event (AE) rates among institutions.

Methods:
Patient and procedural data were collected for all cases, age ≤18 years, performed at sites participating in C3PO-QI (Congenital Cardiac Catheterization Project on Outcomes – Quality Improvement). Sites with verified complete case capture were included in final analysis. All individual AE were validated using expert opinion to ensure universal AE reporting practices. Hemodynamic vulnerability was defined by 6 variables which have previously been shown to be independent risk factors in large datasets. A multivariable logistic regression model with high-severity AE outcome was built using a random sample of 75% of cases within the multicenter cohort; this model was then validated with the remaining 25%. Model discrimination was assessed by the C-statistic and calibration with Hosmer-Lemeshow test.

Results:
Between January 2014 and December 2017, 22,536 cases were recorded among 14 participating institutions, of which high-severity events occurred in 1,126 cases (5%). Overall, 92% of cases were able to be categorized into 1 of 33 unique case types and further stratified into 1 of 5 case type risk categories, maximizing discrimination in the model. The final multivariable model included case type risk category (odds ratio [OR] for category: 2 = 2.5, 3 = 3.9, 4 = 5.5, 5 = 7.5, all p<0.001), number of abnormal hemodynamic indicators (OR for 1 indicator = 1.5, ≥2 = 2, all p<0.001), and age <1 month (OR: 1.3, p = 0.02), C-statistic 0.729, and Hosmer-Lemeshow test p = 0.17. This model performed well in the validation dataset with a C-statistic of 0.717.

Conclusions:
Improving on the prior success of CHARM (Catheterization for Congenital Heart Disease Adjustment for Risk), this updated methodology will allow for the accurate comparison of AE outcomes in the modern era among institutions performing pediatric cardiac catheterization.