Development of Bayesian Mortality Categories for Congenital Cardiac Surgery in Japan

Abstract

Surgery requires a complexity-based ranking system that provides critical information for surgeons to perform strategic operations. However, we still use professional panel systems such as the Risk Adjustment for Congenital Heart Surgery category and the Aristotle Basic Complexity score for this purpose, both of which are subjective. The present study, inspired by more recent development of The Society of Thoracic Surgeons-European Association for Cardiothoracic Surgery mortality scores and categories, applied a Bayesian statistical method to the Japanese nationwide congenital heart registry by estimating inhospital mortality to construct a data-driven, more scientific rating system based on complexity. The study used a 5-year dataset from the Japan Cardiovascular Surgery Database congenital section to construct a Bayesian estimation model. There were 25,968 operations with 186 cardiovascular procedures. To validate the model, we used an independent 2-year dataset with 14,904 operations. The model-based inhospital mortality estimation provided a complexity rating system that replicated the past study that had proposed a five-category system based on the estimated mortality scores. The C-index with the validation dataset for the mortality score and category was 0.80 and 0.79, respectively. The data-driven approach to complexity rating systems for congenital cardiovascular surgery is recommended, as it has better scientific advantages and more convenient updating features.