“…Some models have been developed with data from a large mixed surgical population, such as patients undergoing noncardiac surgeries [13,32,34,35,37] or various thoracic [40], cardiac [2], or upper abdominal procedures [14,18,33]. Others focus on specific surgical settings, such as hysterectomy [22], partial hepatectomy [15,24], pancreaticoduodenectomy [23 & ], colorectal surgery [52], esophagectomy [38], bariatric surgery [30], uvulopalatopharyngoplasty [53], spine surgery [26], transoral odontoidectomy and posterior fixation for craniovertebral junction [16], microsurgical clipping of ruptured intracranial aneurysms [17], lower extremity amputation [19], open infrarenal abdominal aortic aneurysm repair [48], lung resection [1,40,54], prostatectomy [25], and radical cystectomy [27]. The reason why it is important to consider the database used to develop a risk model is that some variables included in these models are specific to the procedure, which accounts for a large proportion of postoperative morbidity measures as seen in the analysis of the US Medicare database [10 & ]; thus, the findings may not be generalizable to other populations or clinical settings.…”