Predictive score for positive upper endoscopies outcomes in children with upper gastrointestinal bleeding

Upper gastrointestinal bleeding (UGIB) is a common pediatric emergency. Esophago-gastro-duodenoscopy (EGD) is the first line diagnostic procedure to identify the source of bleeding. However etiology of UGIB remains unknown in 20% of cases. Furthermore, emergency endoscopy is unavailable in many hospitals in our country.

Identify clinical predictors of positive upper endoscopy outcomes and develop a clinical prediction rule from these parameters.

Retrospective study of EGDs performed in children with first episode of UGIB, in the endoscopic unit of Children's Hospital of Tunis, during a period of six years. Statistical analysis used SPSS20. Univariate analysis was performed and multivariate logistic regression was then modelled to derive a clinical prediction rule.

We collected 655 endoscopies (23.2% normal, 76.8% pathological). We found that time to EGD within 24 hours from the onset of bleeding (p=0.027; Adj OR: 3.30 [1.14 - 9.53]), rebleeding (p=0.009; Adj OR: 6.01 [1.57 - 23.02]), positive gastric lavage outcome (p=0.001; Adj OR: 4.79 [1.95 - 11.79]) and non steroidal anti-inflammatory drugs intake (p=0.035; Adj OR: 5.66 [1.13 - 28.31]) were predictors of positive upper endoscopy outcomes. By assigning each factor, the adjusted odds ratio (Adj OR), we developed a score with four items, ranging from 4 to 20. Using the receiver operating characteristic (ROC) curve the best cut off ≥ 9 was defined (sensitivity 88.2%, specificity 60.6%, positive predictive value 92.7% and negative predictive value 47.6%). The score discriminated well with a ROC curve area of 0.837 (95% confidence interval [0.769 - 0.905]).

This clinical prediction rule is a simple measure that may identify children who needed emergency endoscopy. A prospective study is required to validate our results and evaluate other clinical features that were insufficient for this analysis.