A Comparison of the ABC and AIMS65 Scores in Predicting Outcomes in Patients with Acute Upper Gastrointestinal Bleeding: A Retrospective Multicenter Study.

backgroundUpper gastrointestinal bleeding (UGIB) remains a significant clinical emergency with substantial mortality. Accurate risk stratification is essential for optimal patient triage and management. The ABC score (Age, Blood tests, Comorbidities) and AIMS65 score are prominent pre-endoscopy risk stratification tools, yet direct comparative studies within diverse United States healthcare populations remain limited.

aimsTo compare the predictive accuracy of ABC and AIMS65 scores for in-hospital mortality and secondary clinical outcomes in patients with acute UGIB.

methodsThis retrospective cohort study analyzed 2,009 adult patients admitted with acute UGIB across multiple Northwell Health hospitals between January 2019 and January 2024. Both ABC and AIMS65 scores were calculated for each patient using structured EMR data, ICD-10 diagnosis codes, and anesthesiology procedure documentation. Primary outcomes included in-hospital mortality and 30-day readmission. Secondary outcomes encompassed hospital length of stay, ICU admission, development of complications (shock, sepsis, acute kidney injury), vasopressor use, and need for mechanical ventilation. Univariable logistic regression models assessed predictive accuracy using area under the receiver operating characteristic curve (AUC), with bootstrap internal validation (10,000 resamples) confirming negligible optimism bias. DeLong's test compared discriminative abilities between scores. Sensitivity analyses evaluated score performance across pandemic periods and in a broader AIMS65-computable cohort.

resultsAmong 2,009 patients (56.1% male; median age 70 years), 97 (4.8%) experienced in-hospital mortality and 59 (2.9%) had 30-day readmission. The ABC score demonstrated significantly superior predictive accuracy for mortality compared to AIMS65 (AUC 0.793 vs. 0.661; p < 0.0001 by DeLong's test; optimism-corrected AUCs: 0.793 and 0.661, respectively). Each one-unit increase in ABC score corresponded to a 50.7% increase in mortality odds (OR 1.507; 95% CI: 1.386-1.638). Neither score significantly predicted 30-day readmission. ABC score showed stronger correlations with secondary outcomes including hospital length of stay (r = 0.47 vs. r = 0.33), ICU length of stay (r = 0.35 vs. r = 0.22), and vasopressor requirements (r = 0.30 vs. r = 0.21). ABC's superiority was consistent across pre-pandemic, peak pandemic, and post-peak subgroups. In the broader AIMS65-computable cohort (n = 6,766), AIMS65 demonstrated an AUC of 0.706, confirming that the cohort restriction modestly attenuated its discrimination but that ABC's advantage persisted.

conclusionsThe ABC score demonstrates significantly superior predictive accuracy for in-hospital mortality compared to AIMS65 in patients with acute UGIB, with consistent advantages across secondary outcomes, pandemic periods, and cohort definitions. These findings support the preferential use of ABC score for risk stratification in clinical practice.