Predicting In-Hospital Mortality With Coronary Bypass Surgery Using Hospital Discharge Data: Comparison With a Prospective Observational Study

Abstract

The aim was to determine the usefulness of the hospital discharge Minimum Basic Data Set (MBDS) for predicting in-hospital mortality with coronary bypass surgery by using data from a prospective observational study as a reference. The observational study involved collecting data on all patients undergoing first coronary bypass surgery at five hospitals in Catalonia, Spain between November 2001 and November 2003. In addition, data covering the same period and hospitals were obtained from the MBDS for procedure code 36.1. We investigated the concordance between the information from the two data sources and logistic regression was used to derive predictive models for in-hospital mortality. The model derived using MBDS data was validated using data from the prospective observational study and MBDS data for the years 2004-2006. Model validity was evaluated using discrimination and calibration indices. Some 4.1% of cases in the observational study could not be found in the MBDS. The concordance between the two data sources was highly variable and generally low (kappa values ranged from 0.16 to 0.79). The discriminative ability of the MBDS model was equivalent to that of the observational study model (c=0.80 vs. c=0.79), but when the validity of the former was tested using prospective data and MBDS data for 2004-2006, the discrimination c-index decreased to 0.76 and 0.65, respectively, and the calibration worsened significantly (P< .001). The risk of in-hospital mortality following coronary surgery cannot be accurately evaluated using MBDS data. However, our results indicate that their use as a predictive tool could be improved.