Simplified bedside nomogram for predicting long-term mortality and major adverse cardio-cerebrovascular events following on-pump coronary artery bypass graft

Abstract

Abstract Background When facing patients asking their physicians how their risk of mortality and morbidities will be in the distant future after their cardiac surgery, the lack of predictive models on long-term outcomes is noticed in the clinical practice. Current risk scores, such as the Society of Thoracic Surgeons (STS) and the European System for Cardiac Operative Risk Evaluation (EURO-Score), were designed to predict short-term mortality and need modifications to be used for long-term mortality. Furthermore, these scores involve several factors that might increase the prediction models' complexity. Simplified bedside-friendly models can overcome these complications and render the predictive model more feasible to be used by clinicians. Purpose This study aimed to determine predictors of in-hospital mortality, all-cause mortality, and major adverse cardio-cerebrovascular events (MACCE) among patients undergoing on-pump coronary artery bypass graft (CABG). Subsequently, we aimed to design a simplified bedside nomogram for each of the outcomes using pre-operative predictors. Methods From 2007 to 2017, this single-center, retrospective cohort included 22846 post-CABG patients. Predictors of in-hospital mortality and long-term outcomes were identified using logistic and Cox regression models, respectively. Subsequently, nomograms were established using pre-operative predictors based on parametric survival regression (multiple Weibull models) coefficient. Model performance was measured using the area under the curve (AUC), Harrell’s concordance index (C-index), and calibration curves. Results The study population aged 65.06±9.95 years (74% males). During a median follow-up of 96.85 [70.12-121.17] months, 249 in-hospital mortality, 4172 all-cause mortality, and 7508 MACCEs occurred. Predictors of each study outcome are demonstrated in Figure 1. In-hospital mortality nomogram included glomerular filtration rate (GFR) ≤60, hypertension, valvular heart disease, age, and diabetes (Figure 2) and demonstrated acceptable discrimination ability (AUC: 0.74, 95% confidence interval (CI): 0.70-0.77). Variables in all-cause mortality and MACCE nomograms were consistent, comprising age, heart failure with reduced ejection fraction, GFR≤60, and diabetes (Figure 2). The C-index for all-cause mortality and MACCE prediction was 0.71 (95%CI: 0.70-0.73) and 0.69 (95%CI: 0.68-0.72), respectively. Furthermore, visual assessment of calibration curves indicated good prognostic predictive accuracy for all the three nomograms. Conclusions The designed simplified nomograms had acceptable internal (according to C-index) and external (according to calibration curves) validity for predicting post-CABG in-hospital mortality, all-cause mortality, and MACCE. These nomograms could be implemented in the routine cardiovascular practice after validating in other centers.