Durations of Antibiotic Treatment for Acute Otitis Media and Variability in Prescribed Durations Across Two Large Academic Health Systems

Abstract

BACKGROUND

Antibiotic resistance is a global health threat, fueled by overuse and misuse of antibiotics. Unnecessary antibiotic use contributes to adverse drug events such as rash, diarrhea, and abdominal pain, and contributes to almost 70 000 emergency department visits in children annually [1, 2]. Each excess day of antibiotic prescribed has been associated with a 5% increase in the odds of antibiotic-associated adverse events [3].

Acute otitis media (AOM) affects approximately 5 million children in the United States annually and accounts for roughly 25% of all antibiotics prescribed to children each year [4, 5]. Therefore, understanding prescribing habits for AOM is essential to developing effective strategies to reduce overall pediatric antibiotic use. Studies suggest that for most children with uncomplicated AOM (defined as AOM without otorrhea) [6], the risk of treatment failure does not differ between those who receive 5 days of therapy and those who receive 7 or more days of therapy [7, 8]. Additionally, 5-day regimens provide similar rates of clinical cure and bacterial eradication for children ages 2 years and older compared to traditional 10-day regimens [9, 10]. The American Academy of Pediatrics (AAP) AOM guidelines published in 2013 recommend short (5–7 day) durations for children ages 2 years and older with nonsevere and uncomplicated AOM [6]. Despite these recommendations, most children with AOM receive longer (10 day) antibiotic durations [8, 11, 12]. These prior studies were either performed in single centers and may not be generalizable [11] or using large insurance claims databases [8, 12] and unable to provide a granular level of detail to identify potential interventions to promote more widespread use of short treatment durations and first-line antibiotics.

The aims of this study were to evaluate durations of therapy prescribed for children 2 years and older with uncomplicated AOM across the ambulatory care systems of two, large pediatric academic health systems and to assess the variability in prescribed durations between and within the systems, from the individual to the institutional level. We additionally evaluated proportions of nonfirst-line antibiotic use and clinical outcomes such as treatment failure, AOM recurrence, and mastoiditis.

METHODS

This retrospective analysis of electronic medical record (EMR) data was performed among 135 care locations across two, large academic centers (Vanderbilt University Medical Center [VUMC] and Washington University). Both centers use Epic (Verona, WI) for their EMR. Both centers have robust inpatient antimicrobial stewardship programs and VUMC has a robust outpatient antimicrobial stewardship program. At VUMC, medical directors of outpatient clinics received quarterly, clinic-level feedback on antibiotic choice for AOM during this period. Feedback on antibiotic duration for AOM was added in Q3 2021. Washington University did not have a formal outpatient antimicrobial stewardship program, but there were informal educational sessions held ad hoc at outpatient sites.

We included outpatient (emergency department [ED], urgent care, primary care clinic, or retail clinic) in-person or telehealth encounters for children ages 2–17 years (inclusive) in 135 sites of care that occurred from January 1, 2019, through December 31, 2022, with an International Statistical Classification of Diseases and Related Health Problems-10 (ICD-10) diagnosis of AOM. ICD-10 codes for AOM were defined as H65.X (any “nonsuppurative otitis media”), H66.X (any “suppurative otitis media and unspecified otitis media”), H67.X (any “otitis media in diseases classified elsewhere”), or H72.X (any “perforation of tympanic membrane”). We included encounters with an ICD-10 code of H65.X (any “nonsuppurative otitis media”) to ensure that we did not miss true AOM cases given the risk for miscoding or misclassification by clinicians [13–15]. Any visit with an ICD-10 diagnosis of AOM occurring more than 30 days after the initial visit was abstracted and analyzed as an independent event (ie, a new episode of AOM). We excluded patients with alternate or competing bacterial diagnoses using the Centers for Disease Control and Prevention’s (CDC) Tiers 1 and 2 diagnoses, for which antibiotics are always or sometimes indicated [16]. We additionally excluded encounters for patients who had received an antibiotic within 30 days prior to the index visit and patients with history of tympanostomy tubes prior to the encounter date. We performed a subanalysis of antibiotic choice and duration by diagnosis code and patient age group (2 to <6 years, 6 to <12 years, and 12 to <18 years) to evaluate for differences in treatment choice and duration.

Patient demographic and clinical variables were abstracted electronically from the EMR. Study personnel (S. K. and L. S.) independently validated 30 randomly selected cases from each organization to evaluate concordance between the medical record and electronic data extracted for the following key variables: (1) the diagnosis of AOM; (2) whether an antibiotic was prescribed at the visit; and (3) if an antibiotic was prescribed, the agent prescribed and duration of therapy. Data were de-identified and transferred electronically from VUMC and Washington University to Denver Health and Hospital Authority (DHHA) for analysis. The study was approved by the Colorado Multiple Institutional Review Board (COMIRB).

The primary endpoint was the percent of prescriptions for a 5-day duration. Secondary endpoints included the proportion of prescriptions for a 7-day duration, the proportion of prescriptions for a 10-day duration, the proportion of prescriptions for nonfirst-line antibiotics (antibiotics other than amoxicillin) [6], the proportion of cases associated with treatment failure (new antibiotic associated with an AOM encounter within 3–14 days of the initial encounter) [8, 17], AOM recurrence (new antibiotic associated with an AOM encounter within 15–30 days of the initial encounter) [8, 17], or an adverse drug event. Adverse drug events were medically attended visits within 14 days of the index visit with an ICD-10 diagnosis code for nausea, vomiting, diarrhea, candidiasis, noncandidal skin rash, or allergic reaction [17].

Descriptive statistics for patient and clinical characteristics were performed using R version 4.1.3 (R Foundation for Statistical Computing; Vienna, Austria). Categorical characteristics were summarized with count and percentage overall and by year and site. The duration of antibiotics prescribed by the provider was depicted visually with mean and range.

RESULTS

The final study cohort included 73 198 encounters for uncomplicated AOM (24 815 [34%] from VUMC and 48 383 [66%] from Washington University) among 54 473 unique children 2 years and older (Figure 1). The number of encounters varied by year but was markedly lower during 2020—the first year of the SARS-CoV-2 pandemic. During the validation by manual review of 60 medical records, clinical diagnosis of AOM, whether an antibiotic was prescribed, and the agent prescribed were concordant with electronically abstracted data in 59 (98%) of cases. In 58 (97%) cases, the prescribed duration was concordant. Patient demographics are shown in Table 1. Most encounters were for white (n = 54 548; 75%), non-Hispanic/Latino (n = 53 750; 73%), English-speaking (69 103; 94%), and commercially insured (n = 70 706; 97%) patients. Visits were primarily with physicians (n = 45 854, 63%) and the most common settings were primary care clinics (n = 34 046; 47%) or urgent-care clinics (n = 20 031; 27%; Table 1). Additional clinical characteristics are described in Supplemental Table 1.

Figure 1.

Study cohort description.

Open in new tabDownload slide

A total of 61 612 (84%) encounters resulted in an antibiotic prescription (Table 2). The percentage of prescriptions for 7 days increased over time (11% in 2019 vs 26% in 2022), while 10-day prescriptions decreased over time (84% in 2019 vs 67% in 2022; Table 2). A higher proportion of encounters at VUMC (n = 22 268; 90%) resulted in an antibiotic prescription than at Washington University (n = 39 344; 81%; Table 3). Antibiotic selection was similar across both health systems. The most common antibiotic prescribed was amoxicillin (n = 43 367; 70%); nonfirst-line antibiotics were prescribed in 30% of cases, most commonly cefdinir (n = 9606; 16%; Table 2 and Supplemental Table 3). A beta-lactam allergy was listed in 8052 (11%) encounters.

Overall, most antibiotic prescriptions were for 10 days (n = 45 689; 75%), 20% were for 7 days (n = 12 060), and only 5% were for 5 days (n = 3144; Table 2 and Supplemental Table 3). Variability was observed between the two health systems: 10-day durations were more common at Washington University than VUMC (78% vs 70%) while 7-day prescriptions were more common at VUMC than Washington University (25% vs 17%; Table 3). However, the use of 5-day durations was infrequent at both (5% overall). There was also variability in prescribed durations within the systems. For example, at Washington University 10-day durations were more commonly prescribed in primary care clinics as compared with emergency departments, urgent care, and medical specialty clinics (Figure 2). In both health systems, 5-day durations were infrequent across all types of care locations. There was limited variability in prescribed durations across individual care locations and clinician type (Figure 3A and C); the mean duration at most sites was between 9 and 10 days. No sites had a mean duration of 7 days or less. In contrast, there was more variability among individual clinicians (Figure 3B). The mean prescribed duration was between 5 and 7 days for only 5.6% of clinicians (40/719 total; 21/329 [6.4%] at VUMC and 19/390 [4.9%] at Washington University).

Figure 2.

Proportion of antibiotics prescribed for 5, 7, 10, or 14+ days, by care location and health system. Proportion of encounters with 5-, 7-, 10-, or 14+ duration of antibiotics at each health system and by care location (convenient/walk-in/after hours, emergency department [ED], medical specialty clinic [otolaryngology, allergy/immunology, and pulmonary], primary care clinic, or retail health clinic). Washington University does not have any sites that qualify as retail health clinics thus the absence of data in that panel.

Open in new tabDownload slide

Open in new tabDownload slide

Open in new tabDownload slide

Open in new tabDownload slide

Figure 3.

(a) Variability in antibiotic duration by care location. Each line represents a separate care location. The point represents the mean duration prescribed and the line ranges from the minimum to the maximum duration prescribed across all encounters. Care locations include primary care clinics, convenient/walk-in/after-hours clinics, emergency departments (ED), retail health clinics, and medical specialty clinics. (b) Variability in antibiotic duration by clinician. Each line represents a separate clinician. The point represents the mean duration prescribed and the line ranges from the minimum to the maximum duration prescribed across all encounters. Clinicans are grouped by 10-100 encounters, 100-500 encounters and greater than 500 encounters. (c) Variability in antibiotic duration by clinician type. Each line represents a separate clinician. The point represents the mean duration prescribed and the line ranges from the minimum to the maximum duration prescribed across all encounters. Clinician types include physicians, nurse practitioners, physician assistants, and “other” clinicians.

Across both health systems, treatment failure, AOM recurrence, hospitalization, mastoiditis, and office, ED or urgent-care visits for AOM within 30 days after the index visit were rare (Table 4). No patients went on to be diagnosed with C. difficile infection. Medically attended visits for adverse drug events were uncommon (n = 1888, 3%; Table 4).

In the subanalysis of encounters by diagnosis code, there were fewer encounters for nonsuppurative otitis media with an antibiotic prescription written (7380/13 421; 55%) compared to the rate for patients with acute suppurative otitis media (54 223/58 642; 92%) (Supplemental Table 2). Mean duration of antibiotic prescriptions was similar for both suppurative (9.1 days) and nonsuppurative otitis media (9.2 days).

DISCUSSION

In this evaluation of antibiotic prescribing practices for uncomplicated AOM in children ages 2 years and older across 135 ambulatory care locations in two large academic health systems, an antibiotic was prescribed in 84% of cases. When an antibiotic was prescribed, three-quarters were for 10-day durations while only 5% were for 5 days. The pattern of prescribed durations was similar across all types of care locations. One-quarter of antibiotic prescriptions were for nonfirst-line antibiotics.

Despite existing national guideline recommendations for 5–7-day antibiotic durations for children 2 years and older with uncomplicated AOM, we found that most (75%) antibiotic prescriptions were for 10 days [6]. These findings are similar to previously published studies using large insurance claims databases, which found 88–93% of children were prescribed 10 days of antibiotics [8, 12]. Over the present 4-year study period, there were 45 689 prescriptions for 10 days, equating to 114 223 antibiotic days per year. If these prescriptions were instead written for 7 days or 5 days, 34 267 or 57 111 antibiotic days per year, respectively, would have been avoided. Such a reduction in antibiotic days could have broad cost-saving opportunities both in terms of costs associated with the drug itself, antibiotic resistance, adverse effects, and potential school/daycare and workdays missed. Such changes in prescribing practices are feasible. A multifaceted intervention to promote shorter durations of antibiotics for uncomplicated AOM through individualized prescribing feedback to clinicians in general pediatrics clinics with peer comparison, EMR prescription prompts, and patient and clinician education increased the proportion of prescriptions for 5-day durations from 11 to 85% and did not result in increased episodes of treatment failure or recurrence [18]. Scaling such an intervention could, therefore, have a dramatic impact on reducing unnecessary antibiotic exposure for children with AOM. A large, multicenter, cluster-randomized trial to further evaluate the generalizability, effectiveness, acceptability, and sustainability of this approach has been funded and is being developed [19].

The observed infrequent use of 5-day durations was remarkably similar between the two health systems and across all types of care locations. Although there was variability in prescribed durations among individual clinicians, most tended to prescribe longer durations. It was noteworthy that a small subset of clinicians tended to prescribe shorter durations (mean 5–7 days). Such clinicians may be particularly useful for antibiotic stewardship programs to engage, both to help champion interventions to promote shorter durations of therapy and to demonstrate the feasibility of changing prescribing behavior. This finding also highlights the potential utility of feeding back data to clinicians on their prescribed durations of therapy in comparison to their peers as a method of driving change in prescribing behavior [18].

Shortening the duration of antibiotic prescriptions for AOM could significantly impact rates of antibiotic resistance and adverse drug events. Studies of community-acquired pneumonia in children have shown that 5-day durations of amoxicillin are associated with less frequent colonization with antibiotic-resistant bacteria as compared to 10-day durations [20, 21]. An observational study in Israeli children with AOM demonstrated that middle ear fluid samples had lower rates of pneumococcal resistance in seasons with lower antibiotic prescription rates as compared to seasons with high antibiotic prescription rates [22]. Parent-reported adverse drug events are common among children who receive antibiotics (25–36%) and are more frequent among children who receive longer versus shorter durations [3, 7, 17]. These findings highlight the potential for future outpatient antimicrobial stewardship efforts to reduce unintended consequences of antibiotics through shortening treatment durations.

Amoxicillin is the recommended antibiotic for most children who have not had AOM in the prior 30 days [6], but we observed only 70% of antibiotic prescriptions at index encounters for AOM were for amoxicillin. This is in the range of prior studies using commercial insurance claims databases where 57–71% of patients were prescribed amoxicillin [8, 12]. Notably, 16% of antibiotic prescriptions in our cohort were for cefdinir. This may in part be a response to reported beta-lactam allergies, present in 11% of encounters. However, it is known that over 90% of children with a reported allergy are able to tolerate the antibiotic [23, 24]. A study among children enrolled in Kentucky Medicaid found similarly high rates of cefdinir prescribing for AOM, with 27–30% of all cefdinir prescriptions having an indication of AOM [25]. Cefdinir is a broader-spectrum antibiotic than amoxicillin, and based on pharmacokinetic and pharmacodynamic properties, oral cephalosporins are unlikely to achieve sufficiently high levels in middle ear fluid to effectively treat most pneumococcal ear infections [26]. Rising rates of cefdinir use for common outpatient infections are concerning and, therefore, represent a target for future stewardship interventions.

This study has several strengths. Our dataset included a diverse array of ambulatory care locations across two large health systems consisting of 135 different sites of care and over 700 clinicians. We were able to use patient-level EMR data to capture diagnoses of AOM and associated antibiotic prescriptions. To our knowledge, this is the first study describing variation in prescribed durations of therapy for AOM at the site of care and clinician levels. There are several important limitations to this study. First, identification of cases using ICD-10 codes and electronic abstraction of data from the EMR invariably leads to misclassification. Reassuringly, however, we demonstrated high concordance between manual chart review and the electronic data abstraction methods used for key variables including the diagnosis of AOM, antibiotics prescribed, and duration of therapy, although we only validated a small subset of encounters (0.08% of charts). Additionally, we could not verify the accuracy of clinicians’ diagnoses of AOM. Prior studies have shown that around 50% of children diagnosed with AOM do not actually meet objective diagnostic criteria for AOM [15, 27]. However, since children are routinely misdiagnosed with AOM and treated with antibiotics in clinical practice, it is important to capture and describe antibiotic use irrespective of whether the diagnosis is correct. In fact, a potential benefit of promoting the broad use of 5-day durations of therapy is that the short duration minimizes harm when AOM is incorrectly diagnosed and treated. Second, we were unable to evaluate prescriptions written for watchful waiting or safety-net antibiotic prescriptions. Watchful waiting is an important antibiotic stewardship intervention shown to reduce antibiotic exposure in children with AOM [28]. Finally, our study population was mostly white, English-speaking patients with commercial insurance from two geographically disparate academic health systems in the United States, thus our findings may not be generalizable to other populations.

Antibiotic prescribing for AOM is an important opportunity to improve antibiotic stewardship in children. Across two large health systems, we demonstrated that three-quarters of antibiotic prescriptions were longer than recommended. The use of 5-day durations was uncommon across all types of care locations, and there was limited variability among individual clinicians. Shortening durations of therapy for AOM have the potential to markedly reduce antibiotic exposure among children; this may lead to important reductions in the development of antibiotic resistance, adverse events, and other unintended consequences of antibiotics. Scaling of interventions to promote more widespread use of short treatment durations and first-line antibiotics therefore represents an urgent need in outpatient pediatric antibiotic stewardship. Our data highlight that such interventions must include the spectrum of care locations across ambulatory care.

Supplementary Data

Supplementary materials are available at the Journal of The Pediatric Infectious Diseases Society online (http://jpids.oxfordjournals.org).

Acknowledgments

Author contribution. Dr Sophie E. Katz was involved with conceptualization, methodology, validation, writing—original draft, and writing—review and editing; Dr Timothy C. Jenkins was involved with conceptualization, methodology, and writing—review and editing; Dr Amy B. Stein was involved with methodology, formal analysis, and writing—review and editing; Ms Gale Thomas was involved with methodology, formal analysis, and writing—review and editing; Ms Nancy Koenig was involved with methodology, formal analysis, and writing—review and editing; Mr Luke Starnes was involved with methodology, formal analysis, and writing—review and editing; Dr Jason G. Newland was involved with writing—review and editing; Dr Ritu Banerjee was involved with writing—review and editing; Dr Holly M. Frost was involved with conceptualization, methodology, and writing—review and editing.

Financial support. This work was supported by the Agency for Healthcare Research and Quality (AHRQ), U.S. Department of Health and Human Services (HHS) (grant number 1R01HS029153-01). The authors are solely responsible for this document’s contents, findings, and conclusions, which do not necessarily represent the views of AHRQ. Readers should not interpret any statement in this report as an official position of AHRQ or of HHS.

Potential conflicts of interest. None of the authors has any affiliation or financial involvement that conflicts with the material presented in this report.

Data availability statement. The data that support the findings of this study are available from the corresponding author (S. E. K.) upon reasonable request.

References