Validation of prolonged length of stay as a reliable measure of failure to rescue in colorectal surgery

Failure-to-rescue measures a hospital's response capacity to avoid the death of a patient after a complication. The aim of this study was to validate the use of prolonged length of stay to calculate failure-to-rescue rates as a substitute for traditional coding of complications in colorectal cancer surgery. We performed a cross-sectional between-instruments agreement study. Our study population was comprised of 204 colorectal cancer surgical patients from a public academic hospital during 2017 and 2018. We obtained two failure-to-rescue indicators from administrative data: an indicator using International Classification of Diseases, tenth edition, (ICD-10) codes; and another one using a cut-off point of prolonged length of stay as a predictor of patients with complications. Then, they were compared with a reference indicator from clinical records. Failure-to-rescue rates were between 10 and 13.64 for the study site depending on which indicator was used. A hospital stay ≥10 days had the maximum Youden's index (0.6) and an area under the ROC curve of 0.87. This was used in the failure-to-rescue indicator using prolonged length, which obtained the highest agreement (any coefficient >0.75). ICD-10 codes identified complications poorly. Prolonged length of stay could be a valid replacement of ICD-10 codes when measuring failure-to-rescue in administrative databases for colorectal surgical patients.