Accurate preoperative prediction of unplanned, 30-day postoperative readmission using 8 predictor variables: Methodological issues

We would like to thank Gibula et al for their study developing the accurate preoperative prediction of unplanned, 30-day postoperative readmission using 8 predictor variables. They were pursuing this goal in their study that developed and implemented the Surgical Risk Preoperative Assessment System (SURPAS), which uses just 8 preoperative predictor variables to estimate the risk of 11 adverse outcomes across a broad spectrum of surgical specialties. Preliminary analyses were performed using 28 preoperative variables to determine which factors showed a bivariable association with unplanned, related 30-day hospital readmission. The bivariable association between each of these variables and unplanned, related 30-day readmission was tested using a χ² test for categorical predictor variables or an unpaired t test for continuous variables. This model, as the full model, was compared with the 8-variable model. They used a logistic regression model for both subsets of variables and a comparison of results was performed on the basis of using the C-index as a measure of discrimination, the Hosmer-Lemeshow observed-to-expected plots as a measure of calibration, and the Brier score, a combined metric of discrimination and calibration. The result showed that an 8 variable SURPAS model detects patients at risk for postoperative, unplanned, related readmission as accurately as the full model developed from all 28 nonlaboratory preoperative variables.