In 2019, the American Academy of Pediatrics (AAP) acknowledged racism as a social determinant of health. Reference Trent, Dooley and Douge1 Although racial and ethnic inequities in health care have been described for decades, Reference Williams and Rucker2,Reference Flores and Committee On Pediatric3 the 2019 coronavirus (COVID-19) pandemic has brought these inequities into the spotlight over the past 3 years. In 2021, the Centers for Disease Control and Prevention (CDC) declared systemic racism a public health threat, raising awareness about the health inequity faced by people of color in the United States. 4 Multiple studies have reported differences in rates of antibiotic prescribing in relation to race and ethnicity. Black non-Hispanic and Hispanic children evaluated in emergency departments are less likely to receive antibiotics for viral acute respiratory tract infections than are White non-Hispanic children. Reference Goyal, Johnson and Chamberlain5 Additionally, a study of a large pediatric network found that Black children are less likely to receive diagnoses warranting antibiotics, less likely to receive antibiotic prescriptions, and more likely to receive guideline-concordant narrow-spectrum antibiotics than non-Black children. Reference Gerber, Prasad and Localio6 Although these results appear to demonstrate that children in disadvantaged groups receive more guideline-concordant care, this may be a paradoxical consequence of unconscious bias. This pattern is documented in other common pediatric conditions in which resource overuse is a problem, such as asthma and bronchiolitis. Reference Tackett, Farrow, Kopel, Coutinho, Koinis-Mitchell and McQuaid7–Reference Hester, Nickel, Watson and Bergmann10
Antimicrobial stewardship programs (ASPs) have been successful in reducing antibiotic overuse while improving patient outcomes. Reference Nathwani, Varghese, Stephens, Ansari, Martin and Charbonneau11 Acknowledging outpatient antibiotic overuse and the impact of ASPs, the CDC, the AAP, and the Pediatric Infectious Diseases Society have identified high-quality antibiotic prescribing practices as an important target for outpatient ASPs. Reference Gerber, Jackson and Tamma12,Reference Sanchez, Fleming-Dutra, Roberts and Hicks13 ASPs have the potential to reduce health inequities by developing guidelines that standardize care and reduce implicit bias, allowing all patients to receive appropriate care. However, previous studies have demonstrated that, unless specifically targeted, improvement efforts have a variable effect on health equity and may even result in worsening inequities. Reference Schechter, Pantell and Parikh9,Reference Corl, Levy, Phillips, Terry, Friedrich and Trivedi14,Reference Trivedi, Grebla, Wright and Washington15
When we analyzed the impact of our outpatient ASP, which we implemented in our pediatric urgent care (PUC) clinics starting in August 2018, we detected overall improvement in the rate of choosing firstline antibiotics for common pediatric infections. Reference Nedved, Lee, Hamner, Wirtz, Burns and El Feghaly16 However, we sought to understand the rates of firstline antibiotic agents prescribed for common pediatric infections in relation to race, ethnicity, preferred language, and insurance type prior to implementation of our outpatient ASP and to determine the effects the outpatient ASP had on these differences in the PUC setting. We hypothesized that following the implementation of the outpatient ASP, differences in the rates of firstline antibiotic choice within socioeconomic subgroups would be reduced.
Methods
Setting
We performed a quasi-experimental analysis of antibiotics prescribed by clinicians at 3 freestanding PUC centers located throughout a Midwestern metropolitan area. The PUCs are part of a large academic-affiliated pediatric healthcare institution and are staffed by general pediatricians and advanced-practice registered nurses. These core clinicians work primarily in the PUC setting, and additional “moonlighting” clinicians from other pediatric subspecialties help fill in shifts when needed. The 3 PUCs care for ∼90,000 patients every year.
Our outpatient ASP team includes pediatric infectious diseases physicians, pharmacists, and a biostatistician. We have developed strong collaborations with frontline clinicians from the PUCs since its inception in 2018. The implemented outpatient ASP initiatives align with the CDC’s core elements Reference Sanchez, Fleming-Dutra, Roberts and Hicks13 and include the following: intermittent feedback and education to frontline clinicians; electronic health record (EHR) changes; clinical resources such as an outpatient antibiotic handbook that gives recommendations on firstline antibiotics, dosing, duration, and alternatives for specific common pediatric infections; 17 integration of evidence-based practice with development of clinical practice guidelines; and multiple quality improvement (QI) projects. We described the timeline of our interventions and the impact of our efforts on antibiotic prescribing in a recent manuscript. Reference Nedved, Lee, Hamner, Wirtz, Burns and El Feghaly16
Data collection
We used International Classification of Diseases, 10th Revision (ICD-10) codes to develop an EHR report that identified all PUC encounters of patients >60 days to 18 years of age with the following discharge diagnoses: acute otitis media (AOM), community-acquired pneumonia (CAP), group A streptococcal (GAS) pharyngitis, urinary tract infection (UTI), and certain skin and soft-tissue infections (SSTIs) (nonfacial cellulitis, abscess, and animal bites). Reference Nedved, Lee, Hamner, Wirtz, Burns and El Feghaly16 We included only patients who were prescribed systemic antibiotics. We excluded patients who were transferred or admitted, and those with a concomitant diagnosis who may have required systemic antibiotics thus possibly biasing the antibiotic prescribed for the condition of interest. Reference El Feghaly, Burns, Goldman, Myers, Purandare and Lee18 Our baseline period extended from July 2017 through July 2018, and the postimplementation period extended from August 2018 through December 2020. For each patient, we collected demographic information (ie, age, sex, race, ethnicity, language, insurance) and clinical information (ie, all primary and secondary diagnoses and any oral antibiotics prescribed).
In our institution, demographic information is collected by self-identification through an electronic form filled by the family prior to the appointment.
Data Analysis
We used established national guidelines 19–Reference Stevens, Bisno and Chambers21 and our local antibiogram 22 to determine appropriateness of antibiotic choice for each diagnosis. We considered penicillin G benzathine, penicillin V potassium, and amoxicillin firstline agents for GAS pharyngitis. For AOM and CAP, we considered amoxicillin the firstline agent. Based on the local susceptibilities, we considered cephalexin the firstline agent for UTI and nonfacial cellulitis; clindamycin, or trimethoprim–sulfamethoxazole for abscess; and amoxicillin-clavulanate for animal bites.
We compared the frequency distributions of demographic factors in each period to demonstrate similarity in patient cohorts. We also compared the proportion of patients receiving firstline antibiotic therapy between demographic factors, within each period. We identified trends in rates of firstline antibiotic use over time by race category and diagnosis group trends within race category. We performed unadjusted and multivariable logistic regression models to calculate the odds ratios (ORs) of appropriate antibiotic agent by age, sex, race and ethnicity, language, and insurance type as documented in the EHR. Difference-in-difference effect estimates for each period were derived using interaction terms in the regression models (eg, age group by before and after the ASP was implemented). All analyses were completed using R software (R Core Team, Vienna, Austria).
This study was given exempt status by our institutional review board.
Results
The study included 12,757 encounters during the baseline period (July 2017–July 2018) and 21,846 encounters after the implementation of our ASP (August 2018–December 2020). Socioeconomic variables were, in general, similar between the baseline and study periods (Table 1). Overall, AOM accounted for 55.4% of diagnoses for which antibiotics were prescribed during this time-period. When stratifying the data by diagnosis for the entire study period (July 2017–December 2020), we detected differences in the rates of prescribing firstline antibiotics for all socioeconomic variables for all included diagnoses (Table 2).
Table 1. Demographics of the Study Population
a Includes 43 unique languages.
b Includes commercial multiple peril (CMP), other/government, and unknown.
Table 2. Proportion of Patients of Each Socioeconomic Group That Received Firstline Antibiotics by Diagnosis
a Includes 43 unique languages and unknown.
b Includes commercial multiple peril (CMP), other/government, and unknown.
Note. AOM, acute otitis media; CAP, community-acquired pneumonia; GAS, group A streptococcal; UTI, urinary tract infection.
Evaluation of demographic variables demonstrated a higher proportion of patients receiving firstline antibiotic for the diagnoses evaluated throughout both periods for children aged 2–12 years, those whose preferred language was not English, and those on Medicaid or who self-paid. The proportion of firstline antibiotics was higher for patients who were Asian, Black, Hispanic/Latino, or Native Hawaiian/Pacific Islander compared to children who were White non-Hispanic (Table 3).
Table 3. Proportion of Patients Receiving Firstline Antibiotics
a Includes 43 unique languages.
b Includes commercial multiple peril (CMP), other/government, and unknown.
Our logistic regression model showed that prior to implementation of our ASP, Asian, Black non-Hispanic, Hispanic, Native Hawaiian or Pacific Islander, and patients with unknown or unspecified race and ethnicity were more likely to receive firstline antibiotics for the infections evaluated compared to White non-Hispanic patients with the following ORs: 1.48 for Asian; 1.65 for Black non-Hispanic; 1.32 for Hispanic; 3.88 for Native Hawaiian or Pacific Islander; and 1.3 for unspecified or unknown. Following implementation of the ASP, these differences persisted for all races except Asian and American Indian patients, but we also detected an increase in the OR of multiracial patients to receive firstline antibiotics compared to White non-Hispanic patients (Table 4).
Table 4. Logistic Regression Model Results: Unadjusted Odds of Receiving FirstLine Antibiotic
a Largest category within each demographic subset was chosen as reference group.
b Includes 43 unique languages.
c Includes commercial multiple peril (CMP), and other/government.
Similarly, compared to patients with English reported as a preferred language, patients with Spanish as a preferred language and those with another preferred language were more likely to receive a firstline antibiotic prior to ASP implementation (OR, 1.38; 95% CI, 1.03–1.89; P = .037 and OR, 1.64; 95% CI, 1.08–2.59; P = .026), respectively. This difference persisted following implementation of the ASP (OR, 1.42; 95% CI, 1.12–1.82; P = .005; and OR, 1.59; 95% CI, 1.17–2.21; P = .004) for Spanish and other languages, respectively (Table 4).
Regarding insurance type, patients with government insurance were more likely than those with commercial insurance to receive a firstline antibiotic prior to ASP implementation (OR, 1.30; 95% CI, 1.19–1.43; P < .001). This difference persisted following implementation of our outpatient ASP (OR, 1.21; 95% CI, 1.13–1.3; P < .001). Similarly, patients who self-paid were more likely to receive a firstline antibiotic compared to those with commercial insurance (OR, 1.36; 95% CI, 1.09–1.71, P = .007). Following implementation of our outpatient ASP, the difference between commercial insurance and self-pay increased (OR, 2.05; 95% CI, 1.75–2.42; P < .001) (Table 4).
Patients <2 years of age had lower odds of receiving firstline antibiotics compared to those 2–12 years of age before and following our ASP interventions (OR, 0.65; 95% CI, 0.59–0.71; P < .001; OR, 0.69; 95% CI, 0.64–0.74; P < .001, respectively). Similarly, patients ≥13 years of age were less likely to receive firstline antibiotics prior to the ASP (OR, 0.78; 95% CI, 0.66–0.92; P = .003) and after the implementation of the ASP (OR, 0.7; 95% CI, 0.61–0.79; P < .001). Patients identified as male in the EHR had lower odds of receiving firstline agents compared to female patients (OR, 0.86 before and after ASP implementation) (Table 4).
Trends in firstline antibiotic use improved over time for all diagnoses, but the slopes differed by race and ethnic categories (Fig. 1). Supplemental Figure 1 shows the firstline antibiotic use trend by diagnosis group and race and ethnic categories. We evaluated the difference-in-difference after versus before the change across race and ethnicity, language, and insurance groups. We did not see an overall significant change in the odds of receiving a firstline antibiotic following implementation of our outpatient ASP (Table 4).
Figure 1. Trends in firstline antibiotic use over time for all diagnoses combined, by race category.
Discussion
We identified differences in the rate of prescribing firstline antibiotics for common pediatric conditions relating to race and ethnicity, language, insurance type, age, and sex in the PUC setting. Although overall antibiotic prescribing practices improved for many common pediatric infections following initiation of our outpatient ASP in August 2018, the differences based on socioeconomic status persisted, and in some instances (eg, insurance type) increased. These findings are consistent with previous studies suggesting that quality improvement interventions may not influence differences within socioeconomic subsets unless a conscious effort is made to identify specific drivers of health equity and to consider the impact that interventions will have on different socioeconomic subsets. Reference Schechter, Pantell and Parikh9,Reference Corl, Levy, Phillips, Terry, Friedrich and Trivedi14,Reference Trivedi, Grebla, Wright and Washington15,Reference Penner, Blair, Albrecht and Dovidio23,24
White non-Hispanic children, children with commercial insurance, and children with English as their preferred language were less likely to receive firstline antibiotics compared to their socioeconomic counterparts. Rather than receive the recommended narrow-spectrum antibiotics, these groups appear to receive unnecessary broad-spectrum antibiotics more frequently. Although we did not evaluate the rate of penicillin or amoxicillin allergy in our study, this allergy is more commonly reported in White non-Hispanic children, Reference Taylor, Joerger and Li25,Reference Hersh, Shapiro, Zhang and Madaras-Kelly26 which may contribute to this group not receiving amoxicillin as the firstline antibiotic for respiratory diagnoses. However, data show that >95% of children labeled as “amoxicillin-allergic” do not in fact have an IgE-mediated reaction after cutaneous or systemic exposure to amoxicillin. Reference Norton, Konvinse, Phillips and Broyles27,Reference Vyles, Adams, Chiu, Simpson, Nimmer and Brousseau28
Language in antibiotic prescribing has not yet been well studied. Families with English as their preferred language may feel more comfortable suggesting alternatives to recommended firstline treatments. Although our data suggest that families with English as their preferred language were less likely to receive guideline-concordant care, many patient-facing quality improvement interventions target families whose preferred language is English. More research into the influence preferred language has on prescribing behaviors is needed.
In our study, children identified as male were less likely to receive firstline antibiotics compared to those identified as female for all diagnoses except CAP. More studies are needed to understand differences in prescribing based on sex. However, studies on inequity in pediatric pain assessment reported that male patients are perceived to experience more discomfort compared to female patients exhibiting similar behaviors and clinical presentations; thus, receiving better pain control. Reference Earp, Monrad, LaFrance, Bargh, Cohen and Richeson29,Reference Cohen, Cobb and Martin30
Our data suggest that children aged <2 years were less likely to receive firstline antibiotics. However, this finding may be due to diagnostic distribution, with AOM accounting for >50% of all diagnoses. In this age group, the increased frequency of respiratory illnesses may result in re-evaluations for perceived treatment failures.
Barriers that contribute to differences in antibiotic prescribing behaviors between socioeconomic subsets, and countermeasures that promote antibiotic stewardship while reducing these differences, are not well studied. Implicit bias may have a higher impact when clinicians experience time pressure, high cognitive demand, and fatigue, Reference Penner, Blair, Albrecht and Dovidio23 all of which are commonly experienced in PUC settings. Developing interventions that reduce implicit bias, such as EHR decision tools, has been shown to reduce health inequity in recognition of sepsis. Reference Raman, Johnson, Hayes and Balamuth31 Our ASP interventions included EHR changes but did not appear to affect the socioeconomic differences noted. Additional EHR decision support tools may need to be included. Mangione-Smith et al Reference Mangione-Smith, Elliott, Stivers, McDonald, Heritage and McGlynn32 noted that parental beliefs and clinicians’ perception of parental expectation regarding antibiotics differed based on race and ethnicity and were primary drivers of unnecessary antibiotics in the primary care setting. Future studies should evaluate the influence of implicit bias on antibiotic prescribing and what interventions encourage high-quality, equitable antibiotic prescribing in the PUC setting.
Although specific drivers to reduce differences in antibiotic prescribing in relation to socioeconomic status have not been identified, the CDC developed the CORE Commitment to Health Equity, a strategy for integrating health equity into quality improvement and research initiatives. 33 Additionally, the Institute for Healthcare Improvement defined strategies to eliminate racism and address social determinants of health that may help narrow the gap in health equity for antibiotic prescribing. These strategies include understanding how structural racism and bias influence the relationship between the clinician and family, acknowledging the impact racism and implicit bias have on health equity through culture change and transparent communication, screening for social determinants of health regarding access to care and medications, and developing resources that all families can use and understand. 34 We believe that all ASPs should interrogate their antimicrobial prescribing data relative to socioeconomic status and design interventions with a health-equity perspective.
Our study had several limitations. It had a retrospective design, and we relied on the accuracy of the EHR in determining socioeconomic variables for each encounter. During registration, patients self-reported their race and ethnicity and preferred language. However, the options available to classify each person into a social construct were limited and may have failed to accurately capture the self-identity of each individual. Additionally, we did not evaluate differences in gender identity or immigration status, both of which are important variables to include but are difficult to measure. We did not evaluate documented penicillin allergy, the use of delayed antibiotics, or no antibiotics prescribed for eligible encounters based on socioeconomic variables. Additionally, the recommended treatment options are derived from guidelines. In some instances, guidelines may not be followed, such as in cases of severe illness, antibiotic allergy, or treatment failure. However, outside the reported penicillin allergy differences, we would not expect to see a difference in prevalence of these exceptions between socioeconomic subsets. Finally, the generalizability of these findings may be limited because we evaluated only 3 practices in a Midwestern academic-affiliated PUC setting.
In conclusion, differences in firstline antibiotic prescribing in relation to socioeconomic variables did not improve following implementation of an outpatient ASP, despite overall improvement in firstline, guideline-concordant antibiotics for common pediatric infections. Future studies are needed to identify additional drivers, to develop countermeasures, and to evaluate their impact on health equity in antibiotic prescribing. The findings of this study will contribute to future projects seeking to identify barriers of health equity in antibiotic prescribing behaviors and to develop interventions that promote equitable care to all children.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2023.107
Acknowledgments
We acknowledge the Children’s Mercy Antimicrobial Stewardship Advisory Board for their support, the Children’s Mercy Medical Writing Center for editing the manuscript, and the Children’s Mercy Division of Urgent Care for their willingness to participate in our antimicrobial stewardship efforts.
Financial support
No financial support was provided relevant to this article.
Competing interests
Unrelated to this article, Dr Nedved is the primary investigator for an antibiotic stewardship collaborative through the American Academy of Pediatrics and sponsored by Pfizer. Drs El Feghaly and Nedved are the recipients of an investigator-initiated grant from Merck to address health inequity in antimicrobial prescribing in urgent care (received in April 2023 after this work was completed). All other authors also report no conflicts of interest relevant to this article.
