The Diagnostic Accuracy of an Electronic Medical Record-Based Screening Tool for Pediatric Pneumonia.

Background:  Pneumonia accounts for 1-4% of all Emergency Department (ED) visits in children and leads to over 100,000 hospitalizations per year. This high burden of disease makes pneumonia an important area of research in the busy ED and highlights the need for timely and accurate diagnosis of these patients. The electronic medical record (EMR) offers an expedient and standardized approach to this problem. Our study aimed to evaluate the diagnostic accuracy of an EMR-based screening tool in the diagnosis of pediatric pneumonia. Methods:  Our study was part of a larger, pragmatic randomized cross-over study regarding antibiotic stewardship in pediatric community-acquired pneumonia (CAP). We conducted a cross-sectional cohort study of all children ages 6 months – 18 years who presented to a large, tertiary care pediatric ED between October 1, 2019 to March 31, 2020. The electronic screening tool automatically alerted the treating clinician to consider a diagnosis of pneumonia in any patient with a completed chest radiograph (CXR) and a triage chief complaint of "respiratory problem" (Figure 1). The primary outcome (gold standard) was an International Classification of Diseases, tenth revision (ICD-10) diagnosis code of pneumonia upon ED or hospital discharge. We determined the parameters of diagnostic accuracy of this electronic tool in screening for pediatric pneumonia. Results:  34,970 unique ED encounters were included (52% male; median age 5 years, interquartile range 2 - 11 years) (Figure 2). 6,094 (17%) encounters had a CXR completed and 2,969 (8%) encounters had “respiratory problem” as a chief complaint. A total of 1,243 patients triggered the screening tool. 1,366 patients met the outcome criteria for pneumonia. Our study identified 454 true positives, 32,815 true negatives, 789 false positives, and 912 false negatives. The sensitivity and specificity of the screening tool were 33.2% (95% confidence interval [CI] 30.7 - 35.8) and 97.6% (95% CI 97.5 - 97.8), respectively. The positive and negative predictive values of the screening tool were 36.5% (95% CI 34.2 - 38.9) and 97.3% (95% CI 97.2 - 97.4), respectively.  The diagnostic accuracy of this screening tool, or the overall probability that a patient was correctly classified as positive or negative for pneumonia, was 95.1% (95% CI 94.91 - 95.36). Conclusion: A completed CXR with a chief complaint of "respiratory problem" had a high specificity, negative predictive value and overall accuracy for pediatric pneumonia. While this approach should mitigate alarm fatigue among providers, this strategy demonstrated low sensitivity. These results will inform future iterations of the electronic screening tool as part of a larger study with the goal of improving its sensitivity and positive predictive value by the next respiratory season, thereby enhancing a quality of care outcome for pediatric pneumonia.