Defining a clinical prediction rule to diagnose bacterial gastroenteritis requiring empirical antibiotics in an emergency department setting: A retrospective review

Abstract

Background: Gastroenteritis (GE) is a non-specific term for various pathologic states of the gastrointestinal tract. Infectious agents usually cause acute GE. At present, there are no robust decision-making rules that predict bacterial GE and dictate when to start antibiotics for patients suffering from acute GE to the emergency department (ED). We aim to define a clinical prediction rule to aid in the diagnosis of bacterial GE, requiring empirical antibiotics in adult patients presenting to the emergency department with acute GE. Methods: A two-year retrospective case review was performed on all cases from July 2015 to June 2017 that included patients with acute GE symptoms referred to the ED, after which their stool cultures were performed. The clinical parameters analyzed included patient with comorbid conditions, physical examination findings, historical markers, point-of-care and radiographic tests and other laboratory work. We then used multi-variate logistic regression analysis on each group (bacterial culture–positive GE and bacterial culture–negative GE) to elucidate clinical criteria with the highest yield for predicting bacterial gastroenteritis (BGE). Results: A total of 756 patients with a mean age of 52 years, 52% female and 48% male, respectively, were included in the study. On the basis of the data of these patients, we suggested using a scoring system to delineate the need for empirical antibiotics in patients with suspected bacterial GE based on six clinical and laboratory variables. We termed this the BGE score. A score 0 – 2 points suggests low risk (0.9%) of bacterial GE. A score of 3 – 4 points confers an intermediate risk of 12.0% and a score of 5 – 8 points confers a high risk of 85.7%. A cut-off of ≥ 5 points may be used to predict culture-positive BGE with a 75% sensitivity and 75% specificity. The area under the receiver operating characteristic (AUROC) for the scoring system (range 0 – 8) was 0.812 (95% CI: 0.780–0.843) p-value < 0.001. Conclusion: We suggest using the BGE scoring system (cut-off ≥ 5 points) to delineate the need for empirical antibiotics in patients diagnosed with gastroenteritis. While this is a pilot study, which will require further validation with a larger sample size, our proposed decision-making rule will potentially serve to improve the diagnosis of BGE and thus reduce unnecessary prescription of antibiotics, which will in turn reduce antibiotic-associated adverse events and save on costs worldwide.