547 - Yield of Chest X-Rays among children with asthma presenting to the ED with chest pain and respiratory symptoms

Publication Number: 2532.547

Alek C. Jorge, Connecticut Children's Medical Center, Berlin, CT, United States; Alexander H. Hogan, Connecticut Children's, Hartford, CT, United States; Assaf Landschaft, Boston Children's Hospital, Bergisch Gladbach, Nordrhein-Westfalen, Germany; Amir Kimia, Connecticut Children's Medical Center, Boston, MA, United States

Background: Providers caring for children with asthma in the emergency department (ED) must determine whether chest pain reflects cough-related irritation, musculoskeletal strain, or true pulmonary pathology. When obtained, chest radiographs (CXRs) are typically used to rule out pneumothorax, pneumomediastinum, or pneumonia. Asthma exacerbations increase the risk of the former two, yet data on their prevalence are limited. Such data could inform pre-test probability and guide imaging decisions.

Objective: (1) To assess the prevalence of pneumothorax and pneumomediastinum among children with asthma undergoing CXR, and (2) to evaluate whether chest pain increases the likelihood of pneumonia.

Design/Methods: We applied natural language processing (NLP), combining a support vector machine (SVM) classifier with a fine-tuned Clinical-Longformer model, to identify children aged 2-18 years presenting with asthma and respiratory symptoms to a tertiary pediatric ED during the second half of 2024. Inclusion criteria are summarized in Table 1. For Aim 1, patients with known CXR results were excluded; for Aim 2, we excluded those treated with antibiotics or with documented viral etiology. NLP sensitivity was set at ≥95%. Outcomes were pneumothorax/pneumomediastinum and pneumonia (defined as CXR interpreted as pneumonia and treated with antibiotics). Two reviewers blinded to aims and outcomes validated NLP outputs. Descriptive statistics summarized prevalence, and binary logistic regression identified predictors of pneumonia.

Results: NLP identified 1,622 children for Aim 1 and 1,585 for Aim 2 (sensitivity 97.6% [95% CI: 95.3-99.9%], specificity 66.9% [95% CI: 58-75%]). Providers considered chest pain an important symptom describing it as positive or negative in 52% of encounters, and 7% cited chest pain as the main reason for CXR ordering.
In Aim 1, 546 patients were imaged, including 92 (17%) with chest pain. None had pneumothorax or pneumomediastinum (0%, 95% CI: 0-5%). In Aim 2, 529 patients met inclusion criteria, of whom 85 (16%) had chest pain. Chest pain was not a significant predictor of pneumonia (Table 2).

Conclusion(s): Among pediatric asthma patients imaged for respiratory symptoms, we did not identify a pneumothorax. For a rare event such as pneumothorax, NLP provides a scalable method to refine these estimates and guide evidence-based imaging use in pediatric asthma care. Among pediatric asthma patients imaged for respiratory symptoms, chest pain does not significantly predict pneumonia.

Table 1. Clinical presentation of asthma where CXR may be considered, and exclusion criteria.
* Positive Pertussis, RSV, Influenza, or other respiratory pathogens

Table 2. Predictors of Pneumonia*, among asthma patients that had respiratory symptoms and were imaged.
* A chest radiograph (CXR) was considered positive for pneumonia if the treating provider initiated antibiotic therapy based on the imaging findings.