Development and validation of a clinical scoring system for predicting endoscopic failure in upper gastrointestinal bleeding ulcer: a case control study.

To develop a validated scoring system to predict endoscopic failure among adult patients with upper gastrointestinal bleeding ulcers. This is a case-control study done in a tertiary hospital in Southern Malaysia using secondary data of 443 cases where 98 cases had rebleeding within 30 days of successful initial therapeutic endoscopy and 345 were controls. Logistic regression was used to evaluate association between patient-related, clinical and endoscopic findings with endoscopic failure. These predictors were then used to develop the MARSH (Medication, Age, Renal impairment, Site and Size of ulcer, Stigmata, Heart disease) scoring system. Internal validation was done via bootstrapping, and the model was presented as a score chart. Logistic regression identified chronic kidney disease (CKD) (adjusted OR = 2.04, p = 0.025), ulcer location (adj. OR = 3.15, p < 0.001), and ulcer size > 2 cm (adj. OR = 2.59, p < 0.001) as significant predictors of endoscopic failure. These predictors were incorporated into MARSH scoring system, which underwent internal validation using bootstrapping. The scoring system demonstrated an accuracy of 70.1%, with a specificity of 87% and a sensitivity of 44%, supporting its potential utility in clinical practice. Duodenal ulcers, especially ulcers >2 cm in diameter, and CKD are significant predictors of endoscopic failure. This scoring system is a preliminary predictive model that will require prospective validation in an attempt to identify high risk patients who are likely to require repeated intervention or surgical intervention.