Frequency and Consequences of Routine Temperature Measurement at Well-Child Visits

In this retrospective cohort study, we used well-child visit data from the Stanford Research Repository.⁶  The Stanford Research Repository contains EMR data from Stanford University hospitals and clinics beginning in May 2014 and includes detailed provider and department information, billing codes, patient demographics, and clinical details such as vital signs, diagnoses, diagnostic testing, and medication prescriptions. The Stanford University Institutional Review Board approved the investigation.

We included well-child visits for children ≤18 years old seen between May 2014 and June 2019 at the Packard Children’s Health Alliance, a network of 24 Stanford-affiliated primary care clinics located throughout the northern California Bay Area. Because of the many potential Current Procedural Terminology and International Classification of Diseases (ICD) codes associated with well-child visits, well-child visits were identified by a field called “visit type” to differentiate preventive care visits from urgent care visits. To determine specificity of the data pull, we performed a random chart review on 30 visits in which a preventive care ICD-code was the primary diagnosis and 30 visits in which a preventive care ICD-code was a secondary diagnosis. All 60 notes were consistent with well-child visits (100% specificity). To assess whether visits were scheduled to address acute concerns but transitioned to well-child visits because the child was due for one, we also reviewed 20 visits scheduled within 3 days of the appointment. All notes were consistent with well-child visits. For aim 1, all well-child visits were included (n = 274 351). For aim 2, we excluded 64 (0.02%) visits with temperature values <94°F and 3 (0.001%) visits with temperatures >105°F, because, after chart review, these values were deemed to be likely inaccurately documented.

Temperature measurement was defined as the presence or absence of a temperature value in the vital signs section of the EMR. On preliminary analysis, we observed that clinics either measured temperatures at nearly all (>90%) well-child visits or at few (<20%) visits (Fig 1). Therefore, we categorized clinics as “routine measurement” or “occasional measurement,” respectively. We investigated differences in patient demographics between routine measurement and occasional measurement clinics.

We defined “incidental fever” as fever (temperature ≥100.4°F [≥38.0°C]) not previously suspected and thus, would not have been detected if not for routine temperature measurement. Because this is a subjective attribution, we classified incidental fevers as “probable” (ie, documentation in the progress note that fever was noted incidentally), “possible” (ie, no mention of preceding symptoms or chief complaint that may explain fever), or “unlikely” (ie, explicit documentation of concern that patient had infectious symptoms and/or fever preceding the encounter). The rubric is included in Supplemental Table 4, and examples of each classification are listed in Supplemental Table 5. “Probable” incidental fever was further classified into “noticed” or “unnoticed.” Visits with “unnoticed” fevers lacked mention of fever in the progress note (except for as a listed vital sign) and were not associated with interventions commonly linked to fever, including vaccine deferral. To classify fever, 2 of 5 board-certified pediatricians (AP, AS, MW, JM, and RD) reviewed each chart associated with fever. We a priori set a Fleiss-Cohen weighted κ at 0.80 agreement and modified the rubric until this level was achieved. Disagreements between the 2 reviewers were resolved by team consensus.

Interventions were defined as those commonly obtained in the management of febrile pediatric patients and included diagnostic tests (C-reactive protein [CRP], complete blood cell count [CBC], rapid influenza swab, rapid strep throat swab, and urine culture) and antibiotic prescriptions. In preliminary analysis, we discovered that some clinics obtained urinalyses more frequently (∼20% of all well-child visits) than expected, so we limited our urinary tract infection testing variable to urine cultures. For well-child visits with fever, we also described rates of and reasons for vaccine deferral. Examples of reasons for deferral are listed in Supplemental Table 6. Deferral was defined as explicit progress note wording to delay vaccination. For patients who had vaccines deferred, we reviewed subsequent well-child visit notes to determine if the deferred vaccines were given.

Based on our preliminary findings that the practice of temperature measurement at well-child visits may be clinic driven, we decided to conduct an informal post hoc survey of the clinic managers via Qualtrics to better elucidate protocols surrounding temperature measurement. Survey questions are detailed in Supplemental Information.

We first determined the frequency of temperature measurement at all well-child visits and examined how patient characteristics differed between routine measurement versus occasional measurement clinics. Bivariable and multivariable logistic regressions were performed to identify patient characteristics associated with being seen at routine measurement versus occasional measurement clinics, and odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. We also calculated standardized mean differences (SMDs) to evaluate the magnitude of the differences in patient characteristics, using cutoffs previously defined by Cohen⁷ : no difference if SMD was <0.2, small if SMD was 0.2–0.49, moderate if SMD was 0.5–0.79, and large if SMD was 0.8–1.

To determine the impact of routine temperature measurement and incidental fever detection, we used χ² statistics and a generalized estimating equations regression to compare rates of interventions pursued at routine measurement versus occasional measurement clinics and at visits with fever versus without fever among visits in which temperature was measured at routine measurement clinics. Our multivariable model adjusted for patient characteristics determined to have moderate and large differences (SMD: ≥0.5) between routine measurement and occasional measurement clinics (age, ethnicity, race, and primary insurance) and accounted for clustering by clinic and patient. Because preliminary findings revealed that providers at each clinic did not deviate from overall clinic practice regarding temperature measurement, we did not account for clustering by provider.

Because incidental fever is a subjective attribution, we performed a sensitivity analysis to compare rates of interventions at visits with different categories of incidental fever to rates of interventions at visits lacking fever. Because of a low number of patients within each category, this sensitivity analysis was an unadjusted descriptive analysis. We performed chart review on well-child visits with documented fever for infants ≤90 days of age to better understand the impact of capturing incidental fever in this cohort, who are at higher risk for missing serious bacterial infection in the setting of fever.⁸  All analyses were conducted by using R 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria).

We analyzed a total of 274 351 well-child visits in 50 571 children. Temperature was measured at 161 639 (58.9%) well-child visits. A total of 16 of the 24 clinics (67%) measured temperature at >90% of well-child visits, and the remaining 8 clinics (33%) measured temperature at <20% of well-child visits (Fig 1).

Table 1 displays characteristics of patients at routine measurement versus occasional measurement clinics. The mean patient age for both groups was 5 years old. Routine measurement clinics had a lower proportion of visits for neonates and higher proportions of visits for Hispanic or Latinx, Black, and “other” race patients and patients with government insurance than occasional measurement clinics.

In multivariable analysis, Black patients and patients with government insurance were more commonly seen than White patients and patients with private insurance, respectively, at both routine measurement and occasional measurement clinics. The ORs for these characteristics were higher for routine measurement clinics.

A total of 8 of 9 (89%) clinic managers representing the 24 ambulatory clinics responded to our post hoc informal survey. The survey revealed that some clinics routinely measure temperature at all visit types (preventive and urgent care) and others routinely measure temperature on the basis of patient age, although no clear explanations for why some clinics chose to routinely measure temperature at all visit types or how age thresholds were chosen were provided. Detailed survey results are described in Supplemental Information.

In unadjusted analysis, antibiotic prescriptions were prescribed more frequently at routine measurement than occasional measurement clinics (1.7% vs 1.4%; P < .001), but ≥1 diagnostic tests were obtained marginally less commonly (1.3% vs 1.4%; P < .01). After adjustment and clustering, the odds of prescribing antibiotics at routine measurement clinics remained higher than at occasional measurement clinics (adjusted odds ratio [aOR]: 1.21; 95% CI 1.13–1.29), and the odds of obtaining diagnostic testing remained lower (aOR: 0.76; 95% CI 0.71–0.82).

At routine measurement clinics, 270 of 155 527 (0.2%) well-child visits had documented fever (≥100.4°F; 38.0°C). Of these 270 patients, 167 (61.9%) were classified as unlikely incidental fever, 56 (20.7%) were classified as possible incidental fever, 19 (7.0%) were classified as probable (noticed) incidental fever, and 28 (10.4%) were classified as probable (unnoticed) incidental fever. Table 2 shows rates of interventions by fever status. After adjustment and clustering, antibiotics and diagnostic testing had higher odds of being pursued at visits with a documented fever than visits with documented afebrile temperature (aOR: 10.2 [95% CI 7.4–14.1]; aOR: 17.6 [95% CI 12.8-24.3], respectively). Of visits with fever and associated interventions, 3 were for infants aged 0–90 days old. (The details of each case are described in Supplemental Information).

In visits in which fever was classified as probable (noticed or unnoticed) incidental fever, vaccines were deferred at 12 of 24 (50%) visits but were given by the next well-child visit in 9 of these 12 patients (75%) (Table 3). A total of 6 of 12 (50%) of vaccine deferrals were due to “parental preference,” 3 of 12 (25%) were due to “physician choice,” 2 of 12 (16.7%) had unclear progress note documentation (thus, the reason for vaccine deferral was “unable to be determined”), and 1 of 12 (8.3%) had a reason for deferral unrelated to the concurrent fever and illness (“other”). Examples of reasons for deferral are listed in Supplemental Table 6.

Within a large regional network of pediatric primary care clinics, we found that temperature was measured at slightly more than one-half of well-child visits, a proportion that aligns with our previous analysis using national data.³  In this current study, we provide more granular insight into the drivers of this practice and found that two-thirds of clinics measured temperature at nearly all visits and one-third at few. Routine temperature measurement may contribute to inappropriate antibiotic prescription, in part due to incidental fever detection. Although the overall rate of incidental fever detection is low, it may trigger overuse of health care resources and unnecessary vaccine deferral.

Why clinics appear to use an “all or none” approach to temperature measurement is unclear. We noted that patients with government insurance were more likely to be seen at routine measurement clinics, which aligns with our previous findings using national data.³  In some states, Medicaid requires temperature measurement at well-child visits.⁹  Although we are not aware of such a requirement in California (J. Elliot, Staff Services Manager of the Policy Development and Analytics Team and Benefits Division at the Department of Health Services, personal communication, November 19, 2020), Medicaid requirements in other states may have influenced decisions over temperature measurement. Additionally, our post hoc survey results noted that some clinics routinely measure temperature at all visit types. Routine measurement may be perceived to improve clinic efficiency by maintaining similar intake protocols across all visits.

We noted that routine measurement clinics prescribed antibiotics more frequently than occasional measurement clinics. Although causality for the higher rates of antibiotic prescription is challenging to prove, it is possible that the detection of fevers (or even borderline temperatures that did not meet our definition of fever) that would not otherwise have been noted influences decision-making around antibiotic prescriptions. However, we also noted that, after adjustment for possible confounders and clustering by clinic and patient, diagnostic testing was slightly less common at routine temperature measurement clinics. Although it is possible that providers are opting for antibiotic prescriptions in lieu of diagnostic testing, future research should be done to confirm these findings.

Incidentally detected fever may prompt inappropriate antibiotic prescription in the setting of viral symptoms or misdiagnosed bacterial infections. A cross-sectional study on febrile children seen at 2 ambulatory clinics determined that 64% of febrile patients were managed with oral antibiotics despite frequent diagnoses of viral infections, such as upper respiratory tract infections and gastroenteritis.¹⁰  Another study showed that pediatricians have low rates of compliance with the Centers for Disease Control and Prevention’s diagnostic criteria for acute otitis media (AOM) and that 99% of diagnosed AOM were treated with antibiotics, suggesting that overdiagnosis or misdiagnosis of AOM is contributing to excess antibiotic prescriptions.¹¹  In addition, fever may prompt antibiotic treatment of infections that are often self-resolving. In a study including in-person and telephone pediatric ambulatory visits, the most common diagnoses for visits associated with fever (5%) were upper respiratory infections (URIs), AOM, and tonsillitis,¹²  of which the latter 2 infections often trigger antibiotic prescriptions, despite often improving without treatment.^13–16  Although incidental fever detection through routine temperature measurement at well-child visits is rare, the practice may contribute to overuse of resources. Of the 3 visits for infants 0–60 days old with documented fever and associated interventions, 1 visit was categorized as probable (noticed) incidental fever and had normal infectious screen results. The 2 remaining visits were categorized as unlikely incidental fever and ultimately required hospital admission. Future large-scale research should evaluate the impact of incidental fever detection through routine temperature measurement in well-appearing infants on rates of infectious diagnostic testing, antibiotic prescription, and hospital admission as well as diagnostic outcomes.

Given the high number of well-child visits that occur in the United States, even a small impact of temperature measurement on subsequent interventions could have a considerable effect at the population level. For example, if the association between routine temperature measurement and antibiotic prescription is in fact causal, the absolute risk difference in antibiotic prescriptions of 0.3% signifies that 3 additional antibiotic prescriptions are triggered by every 1000 visits in which temperatures are routinely measured. Based on an estimate of 52 million annual well-child visits from our previous US study,³  routine temperature measurement practices at well-child visits could translate to >150 000 additional antibiotic prescriptions per year.

We also found that vaccines were deferred at more than one-half of well-child visits associated with incidental fever detection, with rates varying by the category of incidental fever. Previous literature suggests that underimmunization varies by age, community, and the type of vaccine, with up to 25% of children experiencing at least 1 missed vaccination opportunity by 2 years of age.^17,18  In these studies, the most common non–well-child visit diagnoses found to be associated with missed vaccine opportunities were AOM, upper respiratory tract infection, gastroenteritis, resolving illness, eczema, and impetigo.^17,18  Investigators for both studies also considered fever (temperature: ≥100.4°F) to be a valid contraindication to vaccination. Although the incidence of incidental fever was low in our sample, the overall population impact on vaccine deferral may be clinically significant, given the high number of well-child visits annually in the United States. Furthermore, the fact that vaccine deferral was common in the context of incidental fever has significant implications for “catch-up” vaccination strategies, such as vaccination at urgent care visits. The Centers for Disease Control and Prevention Advisory Committee on Immunization Practices states that mild illness, including “low-grade fever, an upper respiratory infection (URI), a cold, otitis media, or mild diarrhea,” is an “invalid contraindication to vaccination.”¹⁹  Future educational and quality improvement efforts emphasizing the Advisory Committee on Immunization Practices’ recommendations may help curtail vaccine deferral.

Our study has limitations. First, acute complaints may have prompted temperature measurement. This was not evident in limited chart review and is unlikely to explain the wide variation in temperature measurement across clinics (>90% versus <20% of the time). Second, our categorization of incidental fever was based on progress note review and relied on accurate provider documentation of patients’ infectious symptoms. Furthermore, we only performed chart review on visits associated with fever. Some patients may have had elevated temperatures that did not meet the definition of fever but still triggered interventions because of concern for infection. Third, rates of fever, antibiotic prescriptions, and diagnostic testing were determined by temperature value documentation and relevant medication and procedure codes, respectively. We did not assess previous antipyretic use, the reasons for pursuing interventions, or patient compliance with prescribed antibiotics or diagnostic testing. Finally, in this study, we only include a single network of primary care clinics and may limit generalizability to other northern California Bay Area or US clinic practices.

Despite the lack of supportive recommendations, routine temperature measurement occurred at half of well-child visits in this study and seems to be a clinic-driven practice. The harm-benefit profile of the practice warrants consideration, especially given the rise in measurement as a screening tool during the coronavirus disease 2019 pandemic.

ACIP

Advisory Committee on Immunization Practices

AOM

acute otitis media

aOR

adjusted odds ratio

CBC

complete blood cell count

CI

confidence interval

CRP

C-reactive protein

EMR

electronic medical record

ICD

International Classification of Diseases

OR

odds ratio

SMD

standardized mean difference

URI

upper respiratory infection