BACKGROUND: The Society of Thoracic Surgeons (STS) Adult Cardiac National Database has recently completed the development of the 1995 risk model to be used to estimate the risk of operative death for isolated coronary artery bypass graft (CABG) procedures. This article describes the detailed methodology used, as well as a new Expert Advisory Panel review mechanism that was initiated by The Society. METHODS: Placing emphasis on clinical relevance, data quality, data completeness, and univariate analyses, a logistic regression analysis was used to develop the 1995 CABG-only risk model. The STS National Office invited an Expert Advisory Panel (composed of nationally recognized, independent biostatisticians) to review the modeling process used. RESULTS: The 1995 CABG-only model details are reported. Standard performance measures indicated the model had high predictive power and an acceptable level of calibration. The Expert Advisory Panel reviewed the 1995 CABG model and concluded that the current modeling techniques were adequate. Suggestions for future model development and reporting were proposed by the Panel. CONCLUSIONS: The most current STS risk model of CABG operative mortality is a reliable and statistically valid tool. Its development and performance have been critically examined and approved by an independent panel of experts.
BACKGROUND: The Society of Thoracic Surgeons (STS) Adult Cardiac National Database has recently completed the development of the 1995 risk model to be used to estimate the risk of operative death for isolated coronary artery bypass graft (CABG) procedures. This article describes the detailed methodology used, as well as a new Expert Advisory Panel review mechanism that was initiated by The Society. METHODS: Placing emphasis on clinical relevance, data quality, data completeness, and univariate analyses, a logistic regression analysis was used to develop the 1995 CABG-only risk model. The STS National Office invited an Expert Advisory Panel (composed of nationally recognized, independent biostatisticians) to review the modeling process used. RESULTS: The 1995 CABG-only model details are reported. Standard performance measures indicated the model had high predictive power and an acceptable level of calibration. The Expert Advisory Panel reviewed the 1995 CABG model and concluded that the current modeling techniques were adequate. Suggestions for future model development and reporting were proposed by the Panel. CONCLUSIONS: The most current STS risk model of CABG operative mortality is a reliable and statistically valid tool. Its development and performance have been critically examined and approved by an independent panel of experts.