Background Failure to rescue (FTR) is the rate of death amongst patients with postoperative complications, and is a quality indicator for surgical care. However, FTR has been inconsistently defined in the literature, with potential impacts on the reported rates and hospital rankings. We empirically examined the impact of differences in the FTR definition on hospital rankings, as a proxy measure of quality. Methods All patients undergoing gastrointestinal or hepatopancreatobiliary cancer resection from 2005-2020 were identified from the New Zealand Cancer Registry and National Minimum Dataset. FTR was defined as the death rate amongst patients with any of 19 postoperative complications. We also selected five FTR definitions commonly used in the literature for comparison. Risk-adjusted FTR rates were compared between hospitals using each definition, as well as for in-hospital and 90-day outcomes. Results In total, 31199 patients were included. The 90-day FTR rate varied depending on the definition used; with all 19 complications included this was 10.4% (1517/14646), however the literature definitions ranged from 10.5-18.7%. The FTR definition impacted hospital rankings, with 4-11 hospitals having a different quartile with the literature definitions. There were 847 in-hospital deaths (56% of 90-day), and the in-hospital FTR rate was 5.8% (847/14516). Hospital rankings were affected by the timing of outcome measurement; 9 centres had a different in-hospital FTR quartile compared to 90-day outcomes. Conclusion The complications in the FTR definition and the timing of outcome measurement impact hospital rankings. This has important ramifications for the use of FTR as a quality indicator when comparing institutional performance.