The practice of temperature measurement at well-child visits and normal temperature percentiles in children

Abstract
Objective
To determine: 1) the frequency and characteristics of routine temperature measurement at well-child visits, 2) the proportion of incidental fever and rates of interventions by temperature measurement and incidental fever detection, and 3) the median and extreme (1st and 99th) temperature percentiles by patient characteristics and thermometer route.
Study Design
We addressed the 1st 2 aims with 2 datasets: 1) In a cross-sectional study, we analyzed 22,518 sampled well-child visits from the National Ambulatory Medical Care Survey (NAMCS) between 2003 and 2015, and 2) In a retrospective cohort study, we used electronic medical record data and progress note review from the Stanford Research Repository (STARR) to analyze 274,351 well-child visits between 2014-2019. Progress notes reviewed included well-child visits with documented fever (n=270); 2 independent reviewers categorized fever into the probability of being incidentally detected: unlikely, possible, or probable. For both datasets, exposures for aim 1 were characteristics of visits associated with temperature measurement and exposures for aim 2 were routine temperature measurement and documented fever (≥100.4˚F, ≥38.0˚C). The outcome for aim 1 was the occurrence of temperature measurement and the outcomes for aim 2 were interventions [antibiotic prescriptions, diagnostic tests, ER/hospital admission (for the NAMCS analysis) and vaccine deferral (for the STARR analysis)]. For aim 3, we used the same STARR dataset (dataset #2) to conduct a cross-sectional descriptive analysis of pediatric temperature percentiles.
Results
Temperature was measured at 48.5% (95% CI 45.6-51.4) of well-child visits in NAMCS and at 58.9% (155,527/274,351) of visits in STARR. While measurement was more common during visits by non-pediatric providers, in Hispanic, and Black patients, and in patients with government insurance using NAMCS, the practice was found to be driven by individual clinics using STARR, with 16/24 (67%) clinics measuring temperature at >90% of visits (“routine measurement clinics”) and 8/24 (33%) clinics measuring temperature at <20% of visits (“occasional measurement clinics”). After adjusting for age, ethnicity, race, and patient insurance, antibiotic prescriptions were more common in routine measurement clinics (aOR 1.21, 95% CI 1.13-1.29) but diagnostic testing was less common (aOR 0.76, 95% CI 0.71-0.82). Fever was detected at 270/155,527 (0.2%) well-child visits at routine measurement, 47 (17.4%) of which were classified as probable incidental fever. Antibiotic prescription and diagnostic testing were more common at visits with probable incidental fever than visits without fever (7.4% vs 1.7% and 14.8% vs 1.2%; p<0.001 for both comparisons) and vaccines were deferred at 50% of these visits. Current 50th and 99th temperature percentiles (rectal temperatures in infants ≤12 months old: 99.0˚F and 100.1˚F, respectively; axillary temperatures in children 1-18 years old: 98.0˚F and 99.2˚F, respectively) were lower than the historical standard of “normal” (98.6˚F) and the current definition of fever (≥100.4˚F). Rectal thermometers yielded temperature values 0.8˚F (95% CI 0.8-0.8; p<0.001) higher than axillary thermometers in infants. Temperature did not vary by sex, age, or hour of temperature measurement.
Conclusions
Temperature measurement occurs in slightly more than half of well-child visits, with some clinics measuring temperatures in nearly all visits and others in very few visits. Given the potential impact on clinic efficiency, parental anxiety, subsequent diagnostic and management interventions, and vaccine deferral, the harm-benefit profile of this practice warrants further consideration. We also confirmed that current pediatric temperatures are lower than historical standards and provide us with a more informed approach to the interpretation of temperature values.