ABSTRACT
BACKGROUND/AIMS: Peptic ulcer bleeding (PUB) is the most common cause of upper gastrointestinal bleeding in Korea but there has been no research done using big data. This study evaluates the optimal operational definition (OD) for big data research by analyzing clinical characteristics of PUB. METHODS: We reviewed the clinical characteristics of 92 patients with PUB confirmed on endoscopy in Wonkwang University Sanbon Hospital (January 2013 to December 2014). We calculated sensitivity and positive predictive value (PPV) to detect confirmed PUB patients using ODs developed by combining clinical features of patients with PUB. RESULTS: The mean patient age was 63 years. Men had higher prevalence of PUB than women. Bleeding gastric ulcer was proportionately common in the age range of 40s to 60s in men, while a significantly higher rate of bleeding occurred in women older than 70s. The rate of drug-induced ulcer was 28.2%, whereas the prevalence of Helicobacter pylori was 47.8%. Among the hospitalized patients with diagnostic code of PUB, we ruled out patients with endoscopic removal of gastric adenoma or peritonitis, and selected patients who had been administered intravenous proton pump inhibitor. The sensitivity in this setting was 82.6%, and PPV was 88.4%. CONCLUSIONS: PUB was more common in older patients, and there was a clear gender difference in gastric ulcer bleeding by age. With a proper OD using PUB diagnostic codes, we can identify true patients with sufficiently high sensitivity and PPV.