Aims
Non-variceal upper gastrointestinal bleeding (NVUGIB) is a frequent emergency requiring early risk stratification to identify patients needing urgent endoscopy and hospitalization. This study evaluated the predictive performance of the AS, modified Glasgow-Blatchford (GBm), AIMS65, and H3B2 scores in suspected NVUGIB.
Methods
A retrospective observational study included adults undergoing upper endoscopy for suspected NVUGIB between May and December 2024 at Hospital General de México. Clinical, laboratory, and endoscopic data were collected. Pre-endoscopic scores (AS, AIMS65, GBm, H3B2) were calculated. Diagnostic accuracy for predicting endoscopic treatment and hospitalization was assessed using ROC curves and AUC analysis.
Results
A total of 122 patients were analyzed (63.1% male; mean age 57.7±16.3 years). Melena (54.1%), coffee-ground emesis (25.4%), and hematemesis (16.4%) were the main presenting symptoms. Endoscopic therapy was required in 16.4% of cases, mostly with epinephrine or hemoclips. Gastropathy (40.2%) and peptic ulcer disease (35.3%) were the predominant findings. Hospitalization was required in 38.5% of patients.
For predicting endoscopic therapy, the AS score had the highest AUC (0.636; p=0.055), followed by H3B2 (0.629) and AIMS65 (0.615). For hospitalization prediction, GBm (AUC 0.745; p<0.001) and H3B2 (AUC 0.738; p<0.001) performed best, while AS also showed significant discrimination (AUC 0.664; p=0.002). Optimal cutoffs were AS >2, AIMS65 >1, H3B2 >2, and GBm >6.
| Outcomes | Scores | AUC (95% CI) | P value | Cutoff point | Sensitivity | Specificity |
|---|---|---|---|---|---|---|
| Need for endoscopic treatment | AS | 0.636 (0.502–0.770) | 0.055 | >2 | 0.700 | 0.578 |
| AIMS65 | 0.615 (0.462–0.768) | 0.104 | >1 | 0.700 | 0.657 | |
| H3B2 | 0.629 (0.509–0.750) | 0.068 | >2 | 0.850 | 0.588 | |
| GBm | 0.600 (0.483–0.716) | 0.159 | >7 | 0.850 | 0.608 | |
| Need for hospitalization | AS | 0.664 (0.564–0.764) | 0.002 | >2 | 0.723 | 0.520 |
| AIMS65 | 0.595 (0.492–0.697) | 0.079 | >1 | 0.766 | 0.600 | |
| H3B2 | 0.738 (0.649–0.827) | <0.001 | >2 | 0.830 | 0.507 | |
| GBm | 0.745 (0.655–0.836) | <0.001 | >6 | 0.894 | 0.613 |
Conclusions
The AS score, based on clinical variables available at first contact, showed comparable performance to laboratory-dependent scores—particularly for predicting hospitalization. Its simplicity supports its potential use as a practical triage tool in emergency settings. Further validation could consolidate its role in early NVUGIB management.