Abstract
Clinical experience during the COVID-19 pandemic may change provider attitudes and behavior, particularly surrounding initial antibiotic choices. This study integrated quantitative analysis of practice data and qualitative analysis of clinician interview to understand changes in antibiotic use for pneumonia that have emerged from the pandemic. Methods: We identified antibiotics received within 24 hours of all emergency department (ED) encounters at 134 VA Medical Centers (1/1/2015-4/30/2023) with an initial pneumonia diagnosis based upon ED-assigned diagnosis code or natural language processing and a confirmatory chest imaging report within 24 hours. We developed two mixed effects logistic regression models to predict treatments trained on two periods – early (2015-2016) and late (2022-2023) – that incorporated 60 patient factors, including presenting demographics, comorbidities, vital signs, and laboratory results. We visualized trends in observed versus predicted treatment based upon both models. We then compared patient factors that were most predictive of treatment in the early versus the late model. Qualitative interviews with 29 clinicians from eight VA facilities (2023-2024) explored experiences and influences on pneumonia management and were coded by two trained qualitative analysts utilizing the approach by Crabtree and Miller. Results: Of 337,414 identified ED encounters, 299,335 (87%) received antibiotics within 24 hours, while 46,509 (13%) did not. In the early period, the strongest predictors of withholding an antibiotic for the first 24 hours were elevated brain natriuretic peptide (aOR 1.69[95%CI 1.4-2.0]) and history of congestive heart failure (1.35[1.23-1.49]). In the late period, a positive COVID-19 test became the strongest predictor (OR 2.13 [1.88-2.41]). A positive influenza test had no association with withholding antibiotics in the early period and (aOR 0.95 [0.71-1.27] but became a positive predictor in the late period 1.36 [1.13,1.63]). Interviewed clinicians reported improved viral detection, peer practices, protocols, and stewardship programs as influences on practice. Varying mental models of disease ranging from models of microbial invasion to ecosystem and active hosts. Conclusion: The practice of withholding antibiotics for community-acquired pneumonia increased following the COVID-19 pandemic, which may be influenced by changing approaches to viruses. Further research is needed to evaluate the clinical impact of this practice change.