Comparative Predictive Performance of Upper Gastrointestinal Bleeding Risk Scores and the CALLY Index in Geriatric Patients

Abstract

Background/Objectives: Elderly patients with acute upper gastrointestinal bleeding (UGIB) are a high-risk group with significant morbidity and mortality. Various risk scoring systems have been devised to predict clinical outcome in UGIB, but their effectiveness in geriatric patients is unclear. The purpose of the present study was to evaluate the predictive performance of established UGIB risk scores in older adults and to compare the predictive performance of the C-reactive protein–albumin–lymphocyte (CALLY) index with conventional UGIB risk scores. Methods: This study was a retrospective single-center analysis of a cohort of patients aged ≥65 years presenting with acute non-variceal upper gastrointestinal bleeding (UGIB) who had esophagogastroduodenoscopy from January to December 2025. Scores for Glasgow-Blatchford Score (GBS), Rockall, AIMS65, ABC, MAP(ASH), CANUKA, T-score, and CALLY index were calculated. The primary management endpoint was the requirement for endoscopic hemostatic treatment. Secondary outcomes were blood transfusion necessity, rebleeding, intensive care unit (ICU) admission, length of hospital stay, and 30-day death. The predictive performance of each scoring system was assessed by receiver operating characteristic (ROC) curve analysis. Results: We enrolled 113 patients (mean age 77.7 ± 8.3 years; 59.3% male), of whom 42 (37.2%) required endoscopic hemostatic treatment. No single risk score consistently outperformed the others across all clinical outcomes. GBS showed the highest observed AUC for predicting the need for transfusion (AUC = 0.892), whereas Rockall showed the highest observed AUC for rebleeding (AUC = 0.739) and ICU admission (AUC = 0.668). In the exploratory mortality analysis based on seven deaths, the directionally corrected AUC was 0.846 for CALLY, 0.756 for MAP(ASH), and 0.742 for CANUKA; no pairwise AUC comparisons were performed for this outcome. The CALLY index showed modest discrimination for endoscopic hemostatic therapy (AUC = 0.664). In the exploratory rebleeding analysis, CALLY yielded an AUC of 0.724; however, this estimate was based on only eight events. Lower CALLY levels were associated with increased 30-day mortality, whereas the CALLY index was not significantly associated with transfusion requirements or length of hospital stay. After correction for multiple comparisons, CALLY showed significantly lower discriminatory performance for transfusion requirement than GBS, T-score, CANUKA, and MAP(ASH), whereas no significant differences were observed for endoscopic intervention or ICU admission. Conclusions: No single risk score was able to predict all clinically important outcomes in older adults with acute UGIB. Conventional scoring systems remained useful for several established clinical outcomes, whereas the CALLY index showed modest discrimination for endoscopic hemostatic therapy. Its findings for rebleeding and 30-day mortality were exploratory because of the limited number of events and require external validation. More prospective multicenter studies are needed to confirm the clinical value of the CALLY index in older individuals with UGIB.