Nuance. It’s often a difficult concept to account for in the evolving landscape of evidence-based clinical medicine. We have more information at our fingertips than ever before, and the quality of that information is improving by the day. Yet for many diagnoses, including community-acquired pneumonia (CAP), there often remains ambiguity in the next best step in management. There may be multiple acceptable approaches, but when an incorrect decision can expose a child to unnecessary radiation or foster antibiotic resistance from overuse, it becomes more important to sharpen the line between what is best and what is acceptable. In this issue of Hospital Pediatrics, Geanacopoulos et al. aim to clarify decision-making regarding CAP in children through the use of standard thresholds that connect clinical expertise with the most up-to-date research evidence (10.1542/hpeds.2024-007848).
In this cross-sectional study, the authors designed a survey to analyze physician responses to eight clinical vignettes with varying probabilities of radiographic CAP. It is important to distinguish that the study is not designed to identify bacterial CAP, but rather radiographic CAP for which physicians commonly prescribe antibiotics in clinical practice. The probability of CAP was established using the Pneumonia Risk Score, which is a clinical prediction model that has shown prediction capabilities superior to clinical judgment alone in the initial study of its validity. Respondents were asked to complete the survey by analyzing each vignette, using their clinical judgment to determine their suspected likelihood of CAP and whether they would obtain a chest x-ray or administer empiric antibiotics. After completion of this initial survey, respondents were then provided with the Pneumonia Risk Score clinical prediction model to analyze each vignette and were asked once again to determine their suspected likelihood of CAP and whether they would obtain a chest x-ray or administer empiric antibiotics.
The survey data was used to determine thresholds for testing and treatment in CAP among respondents where the thresholds were defined by the disease probability at which half of the physicians would take each action. The threshold model had two cutoff points: one threshold for diagnostic testing and one threshold for initiating empiric antibiotics. The results were impressive. Knowledge of the clinical prediction model’s estimated disease probability decreased the absolute testing threshold by 4% and the treatment threshold by 8% among physicians in the study, suggesting that there are certain clinical history and/or physical exam elements that physicians place greater importance on than the clinical prediction model.
Of course there are limitations to this study, such as potential selection bias inherent to a survey-based analysis, the majority of the respondents being pediatrics-trained pediatric emergency medicine physicians who work in settings with routinely available imaging modalities, and the fact that vignettes did not include viral testing results that could impact clinical decision-making regarding treatment for radiographic CAP. Nonetheless, this article offers an intriguing look at how evidence-based prediction models can influence and potentially improve physician decision-making in pediatric CAP by balancing clinical expertise with objective data to reduce unnecessary testing and antibiotic use.
