A simple prediction score for in-hospital mortality in patients with nonvariceal upper gastrointestinal bleeding

Background No prediction scores for the mortality of both inpatients and outpatients who developed nonvariceal upper gastrointestinal bleeding (UGIB) without endoscopic findings have been established. We aimed to derive and validate a novel prediction score for in-hospital mortality. Methods We conducted a three-stage, multicenter retrospective study. In the derivation stage, patients with nonvariceal UGIB at six institutions were enrolled to derive the prediction score by logistic regression analysis. External validation of the score was performed to analyze discrimination by patients at six other institutions. Then the performance of this score was compared with that of four existing scores. Results We enrolled 1380 and 825 patients in the derivation and validation cohorts, respectively. A prediction score (CHAMPS-R Score) comprising seven variables (Charlson Comorbidity Index ≥ 2, in-hospital onset, albumin < 2.5 g/dL, altered mental status, Eastern Cooperative Oncology Group performance status ≥ 2, steroids, and rebleeding) with equal-weight scores was established, with high discriminative ability in both derivation and validation cohorts ( c statistic, 0.91 and 0.80, respectively). When rebeeding was excluded from the score (an onset model; CHAMPS Score), this score also achieved high discriminative ability ( c statistic, 0.90 and 0.81, respectively). The prediction scores had significantly higher discriminative ability than the Glasgow Blatchford Score, AIMS65, ABC Score, and clinical Rockall Score in both cohorts (all, p < 0.05). Conclusions We derived and externally validated prediction scores for in-hospital mortality in patients with nonvariceal UGIB. The CHAMPS Score might be optimal for managing such patients. Its mobile application is freely available ( https://apps.apple.com/app/id1565716902 for iOS and https://play.google.com/store/apps/details?id=hatta.CHAMPS for Android).