Instructor Massachusetts General Hospital Boston, Massachusetts, United States
Background: Rapid vital sign assessment, including temperature measurement, is critical among pediatric patients presenting to the emergency department (ED). Abnormal temperature readings often trigger a cascade of interventions (e.g., fever workup in the < 1 month old). In light of recent attention to racial bias introduced by pulse oximetry, critical attention to error and bias introduced by other skin-based devices is warranted. While error rates in temporal thermometry are well documented, potential for differential error rates by race are not well established. Objective: To evaluate whether error rates in error rates in temporal thermometry measurement differ by race. Design/Methods: Pediatric patients (≤18 years old) identified as either Black or non-Hispanic White in the medical record presenting to the ED in a five-hospital system between January 2020 and December 2022 who received at least one paired temperature measurement were eligible for inclusion. Paired measurements were defined as a temporal and oral/rectal temperature within 30 minutes. Temperature measurements were coded as discordant if one modality met fever (≥100.4ºF) or hypothermia ( < 36 ºC below 28 days and ≤35ºC older than 28 days) criteria while the other did not. Generalized estimating equations were used to model the adjusted effect of race on discordant temperature measurements and account for correlation at the patient level. Results: The final study sample included 1,887 paired temperatures (1,678 patients). Among all paired measurements, 25% had discordant measurements (23% in Black patients vs. 25% in White patients). When stratified by race, patients had similar average absolute measurement error (1.4°F (SD 1.2°F) in both groups). This pattern persisted after adjusting for age, sex, year of presentation, and pediatric complex chronic conditions. In the final model, age was significantly associated with measurement error, with younger children (age ≤12 years) more likely to have discordant measurement than older children (>12 years) [Age < 1y aOR 2.1 (95% CI [1.52, 2.87]), 1-4y aOR 3.2 (95% CI [2.25, 4.50]), 5-12y aOR 2.5 (95%CI [1.79, 3.41])].
Conclusion(s): While race was not statistically significantly associated with error in temporal thermometry measurement, age ≤12 years old was associated with increased odds of missed fever by temporal thermometry. Future work should seek to investigate if utilization of temporal thermometry is associated with clinically significant outcomes, such as delayed time to antibiotics. These findings underscore the importance of accurate temperature measurement, particularly in younger children.
