Association of Sociodemographic Factors With Reported Penicillin Allergy in Pediatric Inpatients Free

We conducted a retrospective cohort study of pediatric patients admitted to the general pediatric units of an academic medical center between January 1, 2019, and December 31, 2019. Our hospital is a tertiary-care academic medical center serving patients across much of the Southeastern United States. Within the medical system is our children’s hospital, which includes 190 pediatric inpatient beds and accounts for ∼6000 total annual admissions across all units. All study data were extracted from the EMR. The study was approved by the university’s institutional review board with a waiver of informed consent.

We included patients aged 0 to 18 years with a reported penicillin allergy admitted to the general pediatric inpatient units. Patients admitted to ICUs or other subspecialty units were excluded. Penicillin allergy was defined as a reported allergy to penicillin, amoxicillin, amoxicillin-clavulanate, ampicillin, and ampicillin-sulbactam documented in the EMR.

Allergies can be entered by clinicians, nurses, and allied health professionals, and are widely visible across encounters. The typical process of entering allergies in the EMR in the inpatient setting occurs during the admission intake process. Nursing staff asks patients and families to self-identify their gender, race, ethnicity, and preferred language, and the response is documented in the EMR. Notably, in the EMR available at our institution, the ethnicity options are Hispanic/Latino, non-Hispanic/Latino, or not reported/declined and are not linked to the race options. Separately, patient registration staff enter insurance payer information in the EMR upon admission.

Sociodemographic factors evaluated included gender, age, race, ethnicity, preferred language, and insurance payer. For patients with a reported penicillin allergy, we also reviewed the allergy labels in the medical record to identify reaction history documentation. For patients whose reaction history was documented as unknown or other, sociodemographic factors were examined.

We used χ² tests of association to compare distribution of demographic variables between patients with and without a penicillin allergy. Univariable logistic regression was performed to evaluate associations between demographic variables and penicillin allergy. Odds ratios (ORs) and 95% confidence intervals (CIs) were reported. Additionally, we conducted multivariable logistic regression analyses to evaluate adjusted associations between predictor variables significant in univariate analysis and penicillin allergy. All analyses were conducted using Stata SE 16.1 (StataCorp, College Station, TX). We considered a P < .05 statistically significant with and without adjustment for multiple comparisons.

We identified a total of 3890 pediatric patients admitted to our hospital’s general inpatient units in 2019. There was a total of 299 (7.7%) individual patients who met the inclusion criteria for a reported penicillin allergy, which accounted for 407 patient encounters because multiple patients were readmitted through the course of the year.

As seen in Table 1 and Fig 1, the odds of reporting a penicillin allergy are lower for patients who identify as Black or African American (OR 0.42, 95% CI 0.30–0.57) compared with the White reference group and who prefer a language other than English (OR 0.55, 95% CI 0.33–0.94) compared with the English reference group. The odds of reporting a penicillin allergy are higher for patients who identify as non-Hispanic/Latino ethnicity (OR 1.55, 95% CI 1.02–2.37) compared with the Hispanic/Latino ethnicity group and have private insurance (OR 1.35, 95% CI 1.04–1.74) compared with the Medicaid reference group. The data demonstrate increasing odds of having a penicillin allergy with increasing age, with those aged 16 to 18 years being almost 6 times more likely to have a penicillin allergy compared with those aged <1 year (OR 5.92, 95% CI 3.40–10.31). No significant difference was found in reported gender between the 2 groups.

A multivariable logistic regression model included gender, age, race, ethnicity, language, and insurance. Black race (adjusted OR [aOR] 0.42, 95% CI 0.30–0.59) remained associated with lower odds of penicillin allergy, whereas age >1 year remained associated with increased odds of penicillin allergy (aOR 3.23, 95% CI 1.90–5.49), (aOR 5.17 95% CI 3.04–8.77), (aOR 5.27, 95% CI 3.14–8.85), (aOR 5.65, 95% CI 3.22–9.90) for children aged 1 to 5 years, 6 to 10 years, 11 to 15 years, and 16 to 18 years, respectively, relative to those aged <1 year. Ethnicity, language preference, and insurance payer were no longer significant after adjustment.

A total of 407 inpatient encounters with a documented penicillin allergy occurred during the 1-year review period. The top 3 reactions comprising 83% (337 of 407) were hives, rash, and unknown/other (Table 2). The subgroup of 36 unique patients, representing 46 patient encounters, whose allergic reaction was unknown/other demonstrated a cohort of mostly children of White race (72%), non-Hispanic/Latino ethnicity (94%), and English as the primary language (100%). Notably, 13% (53 of 407) of patients had reactions documented as isolated diarrhea, itching, vomiting, abdominal pain, and nausea, which are most likely an intolerance or anticipated side effects.

In this study, we characterize the sociodemographic characteristics of pediatric inpatients with reported penicillin allergy and reaction history documentation. Our data demonstrate significantly lower odds of penicillin allergy in children who identify as Black and prefer a language other than English. We identified that patients with non-Hispanic/Latino ethnicity, private insurance, and adolescents had increased odds of having penicillin allergy label. Of note, gender was not associated with increased odds of allergy in our patient population. In multivariable analysis, Black race remained associated with decreased odds of penicillin allergy and increased age remained associated with increased odds. After controlling for other variables, ethnicity, language preference, and insurance payer were no longer a barrier to allergy reporting. Previous known bias in the predictor variables makes constructing and interpreting the multivariable model challenging because it may perpetuate bias.

Many of the larger studies involving pediatric patients with reported drug allergy have also found distinctions in demographic characteristics. One study of >1800 children with reported β-lactam allergy who then underwent skin and challenge testing found that there was ultimately no gender or age association with β-lactam allergy.²¹  Another study of >400 000 patients, which included almost 5000 children aged <20 years with a reported penicillin allergy, found increased age and female gender was associated with increased rate of reported antibiotic allergy, although exclusively considering the pediatric population, female gender did not appear to have an association with penicillin allergy.²²  Two other studies of adults and children suggest drug allergies are reported more frequently in females, White race, adults, and inpatients.^20,23  The difference in demographic characteristics among studies and as compared with this study may be in part because of varying patient populations and length of review time periods.

An important contributory consideration for the observed sociodemographic factors associated with penicillin allergy are racial and ethnic differences in the prescribing of antimicrobial agents. Racial and ethnic differences in antimicrobial prescribing have been reported nationally, as well as in pediatric outpatient and emergency department settings. One study of a United States national survey found that respondents who identified as White reported twice as many antimicrobial medication fills compared with other identified races.²⁴  In pediatric outpatient prescribing, Black children receive fewer antibiotics than non-Black children.²⁵  In a pediatric emergency department study using the Pediatric Emergency Care Applied Research Network registry including nearly 40 000 emergency department visits, non-Hispanic Black and Hispanic patients were less likely to receive antibiotics compared with White patients for acute viral respiratory tract infections.²⁶  Our finding that penicillin-allergic labels were less likely to be found in Black and Hispanic patients may reflect lower overall exposure to antibiotics in these groups compared with non-Hispanic and White individuals.

Differences in penicillin allergy labeling by language, race, and ethnicity may also reflect the quality of clinician–family communication and difficulty accessing care.^27–30  Allergy documentation reflects multiple health care system encounters. Consider that, to have a documented penicillin allergy in the EMR, one must have previously received a prescription for penicillin, taken the medication, developed a reaction, and subsequently sought care at a point in time after the reaction. The differences in sociodemographic factors raise the question that a reported penicillin allergy may actually be an indicator of overall access to health care. When interpreting these results, it is important to consider that, although health disparities are systematic, traditionally disadvantaged groups may fare better on some indicators of health, such as in the case of decreased odds of having a penicillin allergy label. The difference may not itself indicate a health disparity but rather a distinction that warrants public health consideration given the potential underlying reasons for sociodemographic difference, which include differences in antibiotic prescribing, access to care, and communication.³¹ 

In addition to demographic findings in our study, the majority of patients had a reaction history of hives, unspecified rash, or unknown consistent with previous studies describing the frequency of rash and inadequate allergy documentation.^19,32–34  Patients with isolated mild cutaneous symptoms are considered at relatively low risk of true allergy and should be referred to allergy testing.³⁵  In the inpatient setting, we have an obligation to our patients to elicit drug allergy reaction history with the support of medical interpreters when language barriers exist, and to adequately document and consider referral for further allergy testing when feasible. A recent publication outlines recommendations for hospitalists to obtain drug allergy history because many allergy labels can be removed on the basis of reaction history, and emphasized the importance of adequate documentation, particularly because many patients have tolerated penicillin since initial reaction.³⁶ 

We acknowledge the limitation of a single-center study at an academic center, which may not be generalizable across pediatric inpatient populations. The study is also limited to retrospective data of patients admitted to general pediatrics units during the course of 12 months and did not include ICUs or other subspecialty units. Notably, sociodemographic and allergy history data from the EMR are patient- or caregiver-reported, entered in the record by health care staff, and limited by the options available in the EMR. Although we did conduct multivariable logistic regression modeling, sample size limitations may have precluded us from identifying all potential associations and any interactions between predictor variables.

This study adds to the limited literature regarding the characteristics of pediatric inpatients who carry the label of penicillin allergy and may also be an indicator of persistent racial and ethnic difference in the prescribing of antimicrobial agents and access to health care. A future opportunity could investigate differences in clinician prescribing for particular diagnoses or indications (ie, pneumonia or surgical site prophylaxis) by patient sociodemographic factors as suggested by a recent study of differential management in inpatient antibiotics for skin and soft tissue by race after controlling for the presence of a penicillin allergy label.³⁷  The further evaluation of the relationship of sociodemographics to reported drug allergy may elucidate trends which could then be used for targeted patient and clinician education and delabeling initiatives. Establishing hospital and clinic partnerships within health systems for targeted outpatient allergy testing could promote more accurate allergy labels and contribute to antibiotic stewardship efforts. Additionally, the potential for inpatient pediatric allergy testing may become more routine as we define further the penicillin allergic sociodemographics and impact. The results of this study will direct future efforts on ensuring completeness of penicillin allergy documentation and risk stratification of reactions to guide delabeling initiatives.