Study design and setting

This retrospective cross-sectional analysis of national Veterans’ Affairs (VA) data included adult patients (aged ≥18 years) with an outpatient dental clinic visit between 2015 and 2018. From this cohort, patients who received an antibiotic prescribed by a dentist within 7 days before or after the dental visit and with a penicillin allergy prior to the antibiotic dispense date were identified. A penicillin allergy was defined as clinician chart documentation of an “allergic reaction” to penicillin in the Corporate Data Warehouse (CDW) medication allergy domain, or an International Classification of Disease, Ninth or Tenth Revision (ICD-9/10) combined with E codes indicating a penicillin allergic reaction (Supplementary Table 1 online). Allergic reactions were identified from 1992 (oldest data available) until the prescription dispense date. Patients with a history of anaphylaxis (ie, had an ICD-9/10 code for anaphylaxis) to penicillin were excluded because it is considered a high-risk reaction, and these patients should only have their allergy evaluated by an allergy or immunology specialist. ^{Reference Shenoy, Macy, Rowe and Blumenthal10} Patients with a documented oral infection using ICD-9/10 codes 7 days prior to the dental visit were excluded, consistent with prior research. ^{Reference Carlsen, Durkin and Gibson24} The remaining penicillin allergic patients receiving a dental antibiotic were categorized into 2 patient cohorts: those who received a noncephalosporin antibiotic and those who received a cephalosporin antibiotic. Each cohort then underwent stratified random sampling by geographic region, so that each region had the same number of patients. Cohort 1 comprised patients with a penicillin allergy who received a noncephalosporin antibiotic (n = 100; 25 per region). Cohort 2 comprised patients with a penicillin allergy who received a cephalosporin (n = 200; 50 per region) (Supplementary Fig. 1 online). The group of patients with a documented penicillin allergy who received a cephalosporin were intentionally oversampled due to the infrequent number of these cases and to capture postcephalosporin allergic reactions.

Table 1. Penicillin Allergy Histories for Penicillin-Allergic Patients Who Did Not Receive Cephalosporin

Note. Some patients had multiple reactions. Significant results (P < .0007, Bonferroni corrected) are shown in bold.

^a 19 patients did not have a specific allergic reaction documented in their electronic medical record and are not shown in the table, as cell frequencies would be zero.

^b Fisher exact test.

^c These penicillin intolerance histories had frequencies of 0: chills (rigors), headache, fatigue.

^d These low-risk allergy histories had frequencies of 0: family history, patient denies allergy but is on record.

^e These moderate- to high-risk allergy histories had frequencies of 0: cough, wheezing, nasal symptoms.

^f These high-risk histories had frequencies of 0: Stevens-Johnson, dystonia, drug reaction eosinophilia, serum sickness, organ injury (liver/kidney), anemia, thrombocytopenia, erythema multiforme, acute generalized exantematous.

Data were extracted from the VA CDW, a national repository that includes clinical and administrative data from the Veterans’ Health Administration (VHA). These data are updated on a continual basis and were used to obtain patient demographics, comorbidities, facility, and dental visit characteristics. Patient demographics included sex, race, age, smoking status, and antibiotic allergy status. Dental visit information was also extracted, including procedure type, and antibiotic associated with the visit. Visits could have >1 procedure type and >1 antibiotic prescription associated with that visit. Antibiotic hypersensitivity 14 and 30 days after the antibiotic prescription date using ICD-9/10 code and E codes were extracted, as well as anaphylaxis 14 and 30 days after the antibiotic prescription date. Antibiotic hypersensitivity was defined as ICD-9/10 codes for dermatitis due to drug, allergic urticaria, angioneurotic edema, and anaphylaxis. Anaphylaxis was defined as ICD-9/10 codes for anaphylaxis. ^{Reference Hohl, Karpov, Reddekopp, Doyle-Waters and Stausberg25,Reference Harduar-Morano, Simon, Watkins and Blackmore26} The definition of invasive procedures that involved manipulation of the gingiva was consistent with past work and American Heart Association/American Dental Association (ADA/AHA) guidelines. ^{Reference Suda, Calip, Zhou and Rowan27} Facility characteristics included rurality, facility complexity, and geographic region. Facility complexity levels were based on patient characteristics, clinical programs, and teaching programs and categorized as high complexity (1a–c), moderate complexity (2), and low complexity (3). Geographic region was based on US Census Bureau categories. Comorbidities were identified using ICD-9/10 codes in the 365 days prior to the dental visit and were used to calculate the Charlson comorbidity index, Gagne combined comorbidity score, and Elixhauser comorbidity index. The definitions of cardiac conditions and prosthetic orthopedic implant were consistent with past work and AHA/ADA guidelines. ^{Reference Suda, Calip, Zhou and Rowan27}

Manual review of the electronic health record identified the dental provider specialty and documented signs and symptoms of the penicillin allergic reaction (if available). More than 1 dental provider and allergic reaction could be associated with each visit or patient. Dental residents may have been involved in the visit, but the extent of supervision was not documented in the visit note. Each allergy reaction was then categorized into intolerance, low risk, moderate-to-high risk, and high risk based on published penicillin allergy evaluation algorithms by Shenoy et al ^{Reference Shenoy, Macy, Rowe and Blumenthal10} (Supplementary Table 2 online). These intolerances, low-risk, moderate-risk, and high-risk categories were then used to quantify the number of patients who were eligible for oral penicillin test challenge, those who would be eligible for penicillin skin testing, and those for whom penicillin allergy testing would be contraindicated. A high-risk reaction would indicate that the patient is contraindicated for allergy testing. A moderate-to-high-risk reaction would indicate that the patient is eligible for skin testing. Penicillin intolerance or low-risk reaction indicates that the patient is eligible for skin or oral testing. Patients with multiple allergic reactions to penicillin were categorized by the most severe allergy (eg, a patient with both a low-risk and high-risk allergy to penicillin were considered high risk). Based on the findings of Shenoy et al, ^{Reference Shenoy, Macy, Rowe and Blumenthal10} patients were then placed into the following allergy categories: true penicillin allergy, pseudo penicillin allergy (defined as intolerance and low risk by Shenoy et al), and missing allergic reaction documentation (Supplementary Table 2 online). A true allergy was defined as moderate-to-high-risk or high-risk allergic reaction to penicillin. Pseudo penicillin allergy was defined as penicillin intolerances or low-risk allergic reactions to penicillin. Missing allergic reaction documentation was defined as “allergic reactant” to penicillin using CDW medication allergy domain. In this case, a clinician entered a reaction to penicillin into their medical record, but ICD-9/10 codes were missing for dermatitis due to drug, allergic urticaria, angioneurotic edema, and anaphylaxis as were E codes. The institutional review board at the Edward Hines, Jr, VA Hospital and VA Pittsburgh Healthcare System approved this study.

Table 2. Patient Demographics and Facility Characteristics for Penicillin-Allergic Patients who Did Not Receive a Cephalosporin Antibiotic

Note. Statistical significance = P < .0007, Bonferroni corrected.

^a Fisher exact test.

^b Kruskal-Wallis test.

Statistical analyses

Descriptive and bivariate statistics, including global χ², the Fisher exact test, and the Kruskal-Wallis test, were used to summarize and compare facility and patient demographics between those with a pseudo penicillin allergy, those with a true penicillin allergy, and those who were missing documentation of the type of allergic reaction in cohort 1. A Bonferroni correction was applied, and the level of significance was set at P < .0007. Unadjusted odds ratios (ORs) and 95% confidence intervals (CIs) were used to calculate the odds of a specific characteristic being associated with a cephalosporin antibiotic prescription. SAS version 9.4 software (SAS Institute, Cary, NC) was used for data and statistical analyses.