A Comparison of Four Severity-Adjusted Models to Predict Mortality After Coronary Artery Bypass Graft Surgery

Abstract

The currently used coronary artery bypass graft predictive models, although generally accurate, have significant shortcomings and should be used with caution. The predicted mortality rate following coronary artery bypass graft surgery varied by a factor of 3.3 from lowest to highest, making the choice of model a critical factor when assessing outcome. The use of these models for individual patient risk estimations is risky because of the marked discrepancies in individual predictions created by each model.

Comparative analysis of risk-adjusted bypass surgery stratification models in a community hospital

“…The area under the ROC curve in our population for the NNE model (0.772) is slightly higher than reported by O'Conner (0.76), 5 Peterson (0.72), 15 or Orr (0.72). 16 In our population, the C-statistic for the PA model was 0.752. The largest PA C-statistics have been reported by Martinez-Alario at a tertiary referral center 12 and Pilam from the San Francisco Heart Institute 11 as 0.857 and 0.80, respectively, whereas the lowest values were reported by Weightman 9 and Pons from a tertiary cardiac center 32 of 0.70 and 0.67, respectively.…”

“…Pilam 11 also reported the highest C-statistic for the CL model (0.80), whereas both Weightman 9 and Pons 32 reported their CL C-statistic of 0.68. The NY C-statistics range from 0.72 15,16 to 0.787 4 and the CA C-statistics range from 0.60 to 0.755. 3,[9][10][11]15,16,32,33 These mortality models may be used to conduct comparisons of risk-adjusted cardiac surgical outcomes among adult cardiac surgical institutions.…”

“…Because patient populations differ significantly between institutions and countries, it is intuitively evident that comparison of absolute mortality rates are not justified. 1,3,16,31 These risk-adjusted indices were developed to correct for differences in patient population and allow for comparison of actual outcome to predicted outcome. There are key differences between models in the patient population from which these models were derived and validated, in the number and type of institutions involved, and in whether the data were retrospective or prospective.…”

“…The NY index has been used extensively and has performed well as an external benchmark, an algorithm developed in 1 population and applied to the evaluation population, in subsequent analyses. [15][16][17] We used the 14-variable mortality model and multivariate odds ratios as given in Table 1. 17 The NNE study group model was developed using data from 3055 patients from 5 clinical centers.…”

Comparison between an empirically derived model and the EuroSCORE system in the evaluation of hospital performance: the example of the Italian CABG Outcome Project☆

“…In order to assess their appropriateness in different clinical settings these models were often applied to surgical procedures for which they were not originally designed [9][10][11][12][13]. The overall conclusion was that the tested models are generally accurate and perform a useful service, but their applicability to different health systems cannot be warranted.…”