Antibiotic Prescribing for Children in United States Emergency Departments: 2009–2014

We performed a cross-sectional retrospective study using the National Hospital Ambulatory Medical Care Survey (NHAMCS) ED public-use data files from 2009 to 2014 (the most recent available data at study onset).¹⁸ NHAMCS is conducted annually by the National Center for Health Statistics from a sample of nonfederal, hospital-based EDs, including representation from freestanding pediatric facilities.¹⁹ In the survey, researchers used a multistage probability sampling design in selecting participating hospitals and patient visits across the United States. Data for sampled visits are collected during a 4-week reporting period, weighted to produce national estimates, and classified by US census region.²⁰ Participating centers had an unweighted ED survey response rate of 75.5% to 92.0% over the study period.

Survey variables used included demographics (age, sex, and race), insurance status (private versus nonprivate), provider type (physician versus advanced practice provider), US census region, ED metropolitan statistical area (MSA) status, discharge diagnosis, and medications administered, prescribed, or continued at ED visits. MSA is defined by the Office of Management and Budget as geographical regions with high population densities. Diagnoses were determined by International Classification of Diseases, Ninth Revision codes. Duration of therapy and route of medication administration were unavailable; medications available only in topical formulations were omitted. For consistency across all included survey years, the first 3 diagnosis fields and 8 medications were considered.

Visits for all patients aged 0 to 17 years discharged from the ED were included. Antibiotic visits were defined as ED visits in which systemic (oral or parenteral) antibiotics were mentioned as continued, administered, or prescribed. Civilian noninstitutionalized population data from the US Census Bureau²¹ were used to determine antibiotic visits per 1000 in the US population. We estimated each ED’s proportion of pediatric visits by calculating the weighted proportion of visits by children out of total ED visits. The ED type was then defined as pediatric if >75% of all visits to that ED were by patients aged 0 to 17 years, or nonpediatric otherwise, a definition used for classifying pediatric EDs previously.^2,22,23 In a histogram of the proportion of weighted pediatric visits by ED (Supplemental Fig 3), it is indicated that most EDs defined as pediatric had >90% of visits by children; thus, our definition appears to accurately differentiate pediatric EDs from others.

The NHAMCS data set does not link medications with diagnoses. On the basis of a previously published classification system informed by national guideline recommendations, we assigned the top 3 listed diagnoses to 1 of 3 tiers: diagnoses for which antibiotics were (1) almost always indicated (eg, bacterial pneumonia), (2) may be indicated (eg, pharyngitis), or (3) generally not indicated (eg, bronchiolitis or bronchitis).⁵ Each visit was assigned a single diagnosis, with priority given hierarchically to tier 1 diagnoses, then tier 2 diagnoses, then tier 3 diagnoses. If a visit had multiple diagnoses from a single tier, the first-listed diagnosis was assigned.

Because >70% of ambulatory pediatric antibiotic prescribing is for acute respiratory tract infection (ARTI), we evaluated prescribing for ARTIs.⁴ We defined ARTI to include pneumonia, acute otitis media (AOM), nonsuppurative otitis media, sinusitis, pharyngitis (including viral and streptococcal), nonspecific upper respiratory tract infection, bronchitis or bronchiolitis, and influenza.⁵ 

Antibiotic types were classified by their generic components and therapeutic classes using the Multum Lexicon Drug Database. Antibiotics included the following: penicillins, cephalosporins, macrolides, quinolones, lincomycin derivatives, tetracyclines, sulfonamides, urinary anti-infectives (eg, nitrofurantoin), aminoglycosides, carbapenems, linezolid, and vancomycin. The general class of antibiotics (eg, penicillins) and subclasses (eg, first-generation cephalosporins) were determined using Multum Level 2 and Level 3 categories, respectively. The following antibiotics and antibiotic classes were considered narrow spectrum: penicillin, amoxicillin, first-generation cephalosporins, sulfonamides, and nitrofurantoin. All other antibiotics, including macrolides and combination β-lactam and β-lactamase inhibitors such as amoxicillin-clavulanate, were considered broad spectrum.

The diagnoses of AOM, pharyngitis, and sinusitis are common respiratory illnesses, and have evidence-based clinical practice guidelines in which specific and primarily narrow-spectrum antibiotic options are recommended as first-line therapy.^24,–26 On the basis of national guidelines, we defined first-line, guideline-concordant antibiotic use as amoxicillin or amoxicillin-clavulanate for AOM and sinusitis and as penicillin or amoxicillin for pharyngitis. Amoxicillin-clavulanate was included as first line for AOM because of recommendations to use it as first line for selected patients, for example, for children with concurrent conjunctivitis.

We studied 3 main outcomes. First, using NHAMCS-provided patient visit weights, we estimated the proportion and mean annual number of ED visits resulting in an antibiotic prescription, called antibiotic visits. We stratified antibiotic visits by patient demographic characteristics, patient insurance type, provider type, US census region, ED MSA status, and ED type. Six years of survey data were combined to increase sample size for analyses stratified by age. Second, we compared the proportion of antibiotic visits and antibiotic classes for ARTI, urinary tract infection (UTI), skin and soft tissue infection (SSTI), and diagnostic tier by pediatric and nonpediatric EDs. All pediatric age groups were combined for the analysis of antibiotic use by diagnosis to increase the sample size and produce reliable estimates, as suggested in the NHAMCS use files.²⁷ We assessed trends in proportion of antibiotic classes over time using logistic regression in 2-year increments (2009–2010, 2011–2012, and 2013–2014) as the predictor variable and percentage of antibiotics by class as the outcome. Finally, on the basis of previous work in ambulatory settings demonstrating factors associated with antibiotic prescribing,^4,28,–30 we used multivariable logistic regression, adjusting for patient demographic characteristics, patient insurance type, provider type, US census region, ED type, and diagnoses, to determine factors independently associated with the outcome of first-line, guideline-concordant prescribing for AOM, pharyngitis, and sinusitis. Odds ratio estimates for first-line, guideline-concordant prescribing by MSA were also determined. Data on MSA were not publicly available in 2012; therefore, odds ratio estimates for MSA were determined using a separate model that adjusted for all other variables but excluded observations from 2012.

We performed statistical analyses using Stata 14 (Stata Corp, College Station, TX), accounting for the complex survey design, including patient visit weights, strata, and primary sampling unit design variables. Estimates were not calculated if they were based on <30 visits or if the relative SE was <0.3, as recommended by NHAMCS standards for reliability and precision.²⁷ Antibiotic prescribing proportions were compared by using χ² tests for heterogeneity. Ninety-five percent confidence intervals (CIs) were calculated for all estimates. Significance was considered at a 2-sided P value ≤.05. NHAMCS data sets used in this analysis are deidentified and publicly available and thus were determined nonhuman subjects by the human subjects advisor at the National Center for Emerging and Zoonotic Infectious Diseases.

In 2009–2014, an average of 29 million (95% CI: 27–32 million) ED visits by children occurred annually, of which 23% (95% CI: 22%–24%) were antibiotic visits, accounting for an estimated 6.7 million (95% CI: 6.1–7.5 million) antibiotic visits annually (Table 1). Fourteen percent (95% CI: 10%–20%) of ED visits occurred in pediatric EDs. The percentage of ED visits resulting in an antibiotic prescription was significantly higher in nonpediatric EDs (24%; 95% CI: 23%–25%) compared with pediatric EDs (20%; 95% CI: 17%–22%, P < .01).

ARTIs accounted for the majority of diagnoses among antibiotic visits (55%; 95% CI: 54%–57%), followed by SSTI (9%; 95% CI: 8%–10%) and UTI (7%; 95% CI: 6%–8%). AOM and pharyngitis accounted for 45% (95% CI: 42%–47%) and 25% (95% CI: 23%–27%) of all ARTI antibiotic visits, respectively. Pediatric and nonpediatric EDs had similar proportions of antibiotic visits for ARTI, SSTI, and UTI (Table 2).

Of antibiotic visits for children in EDs nationally, 32% (95% CI: 30%–34%; 2.1 million per year) were for conditions for which antibiotics are generally not indicated. There were no significant differences in the proportion of visits in each diagnostic tier (antibiotics almost always, may be, or generally not indicated) for children in pediatric versus nonpediatric EDs (Table 2). However, nonpediatric EDs had significantly more antibiotic visits for diagnoses for which antibiotics are generally not indicated (33%; 95% CI: 31%–34%) compared with pediatric EDs (26%; 95% CI: 21%–31%, P = .02).

Narrow-spectrum penicillins were the most commonly prescribed antibiotic class in visits to both pediatric and nonpediatric EDs, followed by cephalosporins, macrolides, sulfonamides, and lincomycin derivatives (clindamycin) (Fig 1). Fluoroquinolones comprised 2% (95% CI: 2%–3%), and tetracyclines, urinary anti-infectives, and aminoglycosides together comprised 3% (95% CI: 3%–3%) of antibiotic prescriptions in all EDs; prescribing estimates for these agents in pediatric EDs did not meet NHAMCS standards for reliability or precision because of sample size. There was no significant change over time in the percentage of any 1 antibiotic class of all antibiotics prescribed (Fig 2).

Macrolides were prescribed in 17% (95% CI: 16%–19%; 1.1 million) of all visits to EDs in which antibiotics were prescribed. Among antibiotic visits, the diagnoses most frequently associated with macrolides included AOM (17%; 95% CI: 14%–20%), nonviral pneumonia (14%; 95% CI: 12%–17%), pharyngitis or tonsillitis (12%; 95% CI: 10%–14%), bronchitis or bronchiolitis (11%; 95% CI: 9%–13%), and upper respiratory tract infection (10%; 95% CI: 8%–13%). Of all ED macrolides prescribing to children, 44% (95% CI: 41%–47%; 500 000 annually) were for diagnoses for which antibiotics are generally not indicated. Macrolides were prescribed at a significantly higher frequency in nonpediatric EDs (18%; 95% CI: 17%–20%) compared with pediatric EDs (8%; 95% CI: 6%–10%, P < .0001; Fig 1).

Broad-spectrum antibiotics were prescribed in 44% (95% CI: 42%–45%) of antibiotic visits by children in EDs nationally, with no difference between nonpediatric (44%; 95% CI: 42%–45%) versus pediatric (44%; 95% CI: 40%–48%, P = .91) EDs.

Among children with AOM, sinusitis, or pharyngitis, the proportion of antibiotic visits during which patients received first-line, guideline-concordant therapy was 78% (95% CI: 76%–80%) and was higher in pediatric EDs (87%; 95% CI: 82%–90%) than nonpediatric EDs (77%; 95% CI: 75%–79%, P < .001; Table 3). This difference persisted in the multivariable model for the percentage of AOM, pharyngitis, or sinusitis visits prescribed first-line, guideline-concordant therapy, with an adjusted odds ratio (aOR) of 2.01 (95% CI: 1.38–2.92) for pediatric EDs compared with nonpediatric EDs. First-line, guideline-concordant antibiotic prescribing was also significantly more likely in visits by African American children (aOR: 1.56; 95% CI: 1.15–2.11) compared with visits by white children. First-line, guideline-concordant antibiotic prescribing was significantly lower in the Midwest (aOR: 0.51; 95% CI: 0.34–0.77), South (aOR: 0.46; 95% CI: 0.32–0.67), and West (aOR: 0.55; 95% CI: 0.35–0.87) when compared with the Northeast region (Table 3). Patients with sinusitis and pharyngitis had lower adjusted odds of receiving first-line antibiotic treatment (aOR: 0.51; 95% CI: 0.32–0.82 and aOR: 0.72; 95% CI: 0.53–0.96, respectively) compared with patients with AOM. The percentage of AOM, sinusitis, and pharyngitis visits with first-line, guideline-concordant prescribing did not change over the study period (Supplemental Fig 4A).

In this study, we provide a comprehensive overview of antibiotic use among visits involving children in US EDs. Nonpediatric EDs care for >85% of pediatric ED patients (>25 million visits annually) and consequently prescribe antibiotics in >6 million annual pediatric visits, nearly 10% of all pediatric outpatient antibiotic prescriptions.⁹ Compared with visits in pediatric EDs, clinicians in nonpediatric EDs prescribe antibiotics more frequently for conditions for which antibiotics are likely unnecessary, are more likely to select macrolides, and are less likely to select first-line, guideline-recommended antibiotics for children with common respiratory illnesses.

Although the proportion of children with viral infections receiving unnecessary antibiotics has been as low as 2.5% in selected, large, academic pediatric EDs,³¹ we evaluated all ED visits nationally and found 26% of antibiotic visits at pediatric EDs were for conditions for which antibiotics were generally not indicated. Our finding mirrors previous work on antibiotic prescribing for children with ARTI in the ambulatory and ED setting.^4,32 In comparison, we found that 33% of antibiotic visits at nonpediatric EDs were for conditions for which antibiotics were generally not indicated, accounting for an estimated 2 million unnecessary pediatric antibiotic prescriptions annually. Antibiotic prescribing for these conditions provides no benefit to patients, puts them at unnecessary risk for adverse events, and should be a target for quality improvement in EDs nationally, particularly in nonpediatric EDs.

More frequent unnecessary and guideline-discordant antibiotic prescribing by clinicians in nonpediatric EDs may reflect differences in clinical experiences during training. The American Academy of Pediatrics has published multiple evidence-based clinical practice guidelines for common pediatric conditions, created with multidisciplinary expert consensus, including representation from pediatric emergency medicine. ARTIs constitute the majority of antibiotic visits for children in EDs nationally, as is the case among outpatients seen by pediatric clinicians.^4,33 Pediatric-trained clinicians are more likely to work in pediatric EDs compared with nonpediatric EDs^34,35 and might thereby have increased familiarity and compliance with pediatric-specific guidelines, which could contribute to differences in quality of antibiotic prescribing between pediatric and nonpediatric EDs. Previous work reveals children are more likely to receive unnecessary antibiotics for upper respiratory tract infections in EDs or family practitioner offices compared with pediatrician offices.³⁶ Increased exposure of emergency medicine clinicians to pediatric-focused guidelines and antibiotic stewardship initiatives is an important strategy to improve antibiotic prescribing quality in nonpediatric EDs.

We found that broad-spectrum antibiotics constitute almost half (47%) of all antibiotics prescribed to children in the ED. Previous work has revealed similar rates of broad-spectrum antibiotic prescribing for children in other ambulatory settings and in the ED with respiratory tract infections, specifically.^4,32 In our study, macrolides accounted for 17% of all antibiotics prescribed to children in EDs, and nonpediatric EDs prescribed macrolides significantly more often than pediatric EDs (18% vs 8%, respectively). With our work, we add to a growing body of literature in which it is demonstrated that nonpediatric health care centers and clinicians prescribe macrolides more frequently than pediatric health care centers and clinicians.^37,–40 In our study, macrolides were most commonly prescribed for AOM; however, macrolides are not recommended for AOM treatment.²⁴ Streptococcus pneumoniae is the most common cause of AOM and sinusitis,⁴¹ yet ∼30% of US invasive S pneumoniae isolates are resistant to macrolides,⁴² so macrolides represent a suboptimal treatment choice for these conditions. Bronchitis or bronchiolitis and upper respiratory tract infections, diagnoses for which antibiotics are not recommended,^43,–46 constituted 21% of all macrolide use in our study. Inappropriate macrolide use should be a target for improvement in antibiotic prescribing for children in EDs and in nonpediatric EDs, in particular.

The Centers for Disease Control and Prevention’s Core Elements of Outpatient Antibiotic Stewardship⁴⁷ offers a framework for implementing antibiotic stewardship in outpatient settings, including EDs. Additionally, EDs can improve their antibiotic prescribing through standardized clinical pathways for guideline-concordant antibiotic prescribing and quality of care, clinical decision support tools, and clinical justification for nonrecommended antibiotic use.^47,48 

We acknowledge several possible limitations to this study. First, in our study period from 2009 to 2014, we captured the most recent survey data available at the time of the analysis. Although prescribing patterns could have changed in the interim, the trend in antibiotic prescribing did not change significantly during the study period (Supplemental Fig 4B). Second, indications for antibiotics were based on the 3 diagnosis codes available in the NHAMCS data set, which were not directly linked to medications mentioned during the visit. The hierarchical classification system used was designed to categorize the most likely diagnosis for which the antibiotic was prescribed. However, it is possible that a small proportion of antibiotic visits grouped as “generally not indicated” might still warrant antibiotic prescribing, such as visits among patients with sickle cell disease and fever. Third, patient allergies to medications are not captured in the NHAMCS data set, which could influence the antibiotic class prescribed. However, patient antibiotic allergies would be expected to occur at similar rates in pediatric and nonpediatric EDs. Fourth, longitudinal patient data were not available, nor were data on provider specialty. Finally, although our definition of pediatric EDs has been used previously^2,22,23 and is likely specific (Supplemental Fig 3), some large nonpediatric EDs might have a distinct pediatric subsection that constitutes the minority of visits but is nonetheless staffed by pediatric specialists.

In this nationally representative study of all ED visits by children, we found the vast majority of pediatric visits and antibiotics ordered or prescribed to children occur in nonpediatric EDs, highlighting the important role of nonpediatric emergency medicine practitioners in improving the quality of antibiotic prescribing for children nationally. Pediatric antibiotic stewardship efforts in nonpediatric EDs have the potential to eliminate >2 million unnecessary antibiotic prescriptions annually and increase the rates of guideline-concordant antibiotic use among children seen in this setting. Macrolides are frequently and inappropriately prescribed for children in EDs and in nonpediatric EDs, in particular. Antibiotic stewardship efforts should expand to EDs nationwide, particularly regarding avoidance of antibiotic prescribing for treatment of conditions for which antibiotics are not indicated, reducing macrolide use, and increasing first-line, guideline-concordant prescribing for children, especially in nonpediatric EDs.

AOM

acute otitis media

aOR

adjusted odds ratio

ARTI

acute respiratory tract infection

CI

confidence interval

ED

emergency department

MSA

metropolitan statistical area

NHAMCS

National Hospital Ambulatory Medical Care Survey

SSTI

skin and soft tissue infection

UTI

urinary tract infection