OBJECTIVES. The goals were to describe trends in acute otitis media, treatment failure, and relapse and in high-dose amoxicillin use and to determine whether treatment of acute otitis media with high-dose amoxicillin was associated with treatment failure or relapse.
METHODS. We conducted a retrospective study of acute otitis media visits made between 1996 and 2004 by children 2 months to 12 years of age in a large group practice, using computerized data. We defined acute otitis media as an otitis media visit with antibiotics dispensed (preceded by 30 days without otitis media visits), treatment failure as initiation of treatment with a second antibiotic before the first prescription was finished, and relapse as initiation of antibiotic treatment after the first prescription was finished but within 30 days after the index acute otitis media episode. The primary independent measure was high-dose amoxicillin (>70 mg/kg per day). We evaluated changes over time and determined whether high-dose amoxicillin use was associated with otitis media treatment failure or relapse.
RESULTS. We identified 111335 acute otitis media visits over a 9-year period. The incidence of acute otitis media decreased from 385.1 visits per 1000 enrollees in 1996 to 188.8 visits per 1000 enrollees in 2004. The proportion of acute otitis media visits treated with high-dose amoxicillin increased from 1.7% in 1996 to 41.9% in 2004. Both otitis media treatment failure and relapse rates decreased from 1996 to 2004 (from 3.9% to 2.6% and from 9.2% to 8.9%, respectively). The odds of treatment failure or relapse did not differ between acute otitis media episodes treated with high-dose and low-dose amoxicillin.
CONCLUSIONS. During the past decade, acute otitis media, treatment failure, and relapse became less common and high-dose amoxicillin use increased. However, high-dose amoxicillin treatment did not reduce the risk of individual infections resulting in adverse outcomes.
Acute otitis media (AOM) is the most common reason for antibiotic treatment in young children in the United States.1 Observational studies suggest that 5% to 21% of AOM cases require a second course of antibiotics to achieve resolution.2–5 National rates of antibiotic use for AOM have decreased in recent years.6,7 However, it is not known whether the likelihood of treatment failure or relapse after an initial course of antibiotic treatment for AOM has changed in the past decade.
The need for current information on AOM treatment failure and relapse is underscored by concerns about antibiotic resistance among Streptococcus pneumoniae and other bacteria that cause AOM.8 In 1999, the Centers for Disease Control and Prevention (CDC) recommended that AOM in children at high risk for antibiotic resistance be treated with high-dose amoxicillin or cefuroxime.9 In 2004, the American Academy of Pediatrics (AAP) broadened these recommendations to suggest that AOM in all children ≤12 years of age should be treated with high-dose amoxicillin.1 Scant data exist on how widely high-dose amoxicillin treatment has been adopted in actual practice and whether its use has affected rates of AOM treatment failure or relapse.
To address these gaps in knowledge, we conducted a study of AOM visits to a multispecialty provider group between 1996 and 2004. Our objectives were (1) to describe trends in AOM initial diagnosis, treatment failure, and relapse, (2) to describe rates of use of high-dose amoxicillin to treat uncomplicated AOM, and (3) to evaluate whether treatment of uncomplicated AOM with high-dose amoxicillin was associated with lower likelihood of treatment failure or relapse.
Study Design and Data Collection
We conducted a retrospective observational study of children who received care from Harvard Vanguard Medical Associates, a multispecialty provider group with 14 practice sites at which ∼80 pediatric clinicians care for ∼275000 patients in the greater Boston area. Computerized data on outpatient visits and prescription medications were collected from the claims data of Harvard Pilgrim Health Care, a nonprofit health plan. Only Harvard Vanguard Medical Associates patients who were insured by Harvard Pilgrim Health Care were included.
The study population included children 2 months to 12 years of age with ≥90 days of coverage by Harvard Pilgrim Health Care between January 1, 1996, and December 31, 2004. Uncomplicated AOM visits were defined as visits having (1) an International Classification of Diseases, Ninth Revision, Clinical Modification code for otitis media (OM) (381.x or 382.x), (2) an antibiotic prescription filled within 3 days after the visit, and (3) no other antibiotic-treated OM visits during the preceding 30 days. We excluded OM visits linked to prescriptions for sulfisoxazole or to antibiotic prescriptions with a supply of >14 days, because these patterns suggested that antibiotics were being administered for OM prophylaxis. We also excluded claims for treatment of OM in hospitalized children and those for outpatient visits made by children with sickle cell disease, cystic fibrosis, asplenia, HIV, or a malignancy.10
The 2 primary outcomes were OM treatment failure and OM relapse, as defined by Pichichero.11 Specifically, we defined treatment failure as occurring after an uncomplicated AOM visit if a second AOM visit was associated with a prescription for a different antibiotic and occurred before the initial prescription would have been completed. We defined relapse as occurring after an uncomplicated AOM visit if a second AOM visit associated with a different antibiotic prescription occurred after the index prescription was completed but within 30 days after the initial AOM visit. Only the first incidence of treatment failure or relapse after each uncomplicated AOM visit was included in this analysis.
We defined high-dose amoxicillin, the primary predictor in our analysis, as a dose of ≥70 mg/kg per day and usual-dose amoxicillin as a dose of <70 mg/kg per day. This threshold was similar to CDC and AAP guidelines, which define high-dose amoxicillin as 80 to 100 mg/kg per day. Other predictors included patient age, patient gender, calendar year of index OM diagnosis, and number of uncomplicated OM episodes the patient had experienced before the episode being analyzed. We categorized patients’ age on the day of the visit as 2 through 6 months, 7 through 11 months, 1 through 2 years, 3 through 4 years, or 5 through 12 years. We defined practice site as the clinical location where the AOM was diagnosed. To measure OM history, we measured the number of AOM visits the patient made between birth and each AOM visit, as well as the number made in the months preceding each AOM visit.
We calculated the annual incidence rates (incidence density)12 of uncomplicated AOM, treatment failure, and relapse by dividing the number of events by the person-time under observation during each year. The incidence rates of treatment failure and relapse after uncomplicated AOM were calculated for each year. After weighting the age categories according to month, we calculated weighted AOM rates stratified according to age category. To determine whether incidence rates of the outcomes differed according to year, age, or antibiotic prescribed, accounting for subjects with multiple, uncomplicated, AOM visits, we used bivariate, generalized, linear, mixed models accounting for nonindependence of AOM episodes (clustering) within subjects.13 We used Wilcoxon rank-sum tests to test whether our 2 measures of OM history were associated with either the use of high-dose amoxicillin to treat AOM or the occurrence of treatment failure or relapse.
We created multivariate, generalized, linear, mixed models13 to assess whether high-dose amoxicillin was associated independently with treatment failure or relapse after each subject's first-ever uncomplicated AOM. These models controlled for year, patient gender, and age category, as well as correlations within practice. To minimize the effects of confounding by treatment indication, we restricted the primary analysis to each subject's first uncomplicated AOM episode. In addition, we included only OM visits made between January 1, 2000, and December 31, 2004, in these analyses, because we presumed that clinical practice would not have changed in response to the CDC 1999 OM treatment guidelines before 2000.9 In secondary analyses performed to determine whether OM history confounded the relationships between amoxicillin dose and treatment failure or relapse, we included all OM visits (not only first-ever OM episodes) made between January 1, 2000, and December 31, 2004, in multivariate, generalized, linear, mixed models that controlled for OM history in addition to amoxicillin dose, year, age category, gender, and clustering according to practice site and patient.
During the 9-year study period, we identified 111335 visits for uncomplicated AOM. The incidence rates of uncomplicated AOM decreased significantly over the study period (P < .0001), from 385.1 visits per 1000 person-years in 1996 to 188.8 visits per 1000 person-years in 2004 (Fig 1). A greater proportion of AOM visits occurred in older infants (weighted average: 37.1%), followed by toddlers (24.4%), younger infants (18.9%), preschool-aged children (14.2%), and school-aged children (5.4%). Subjects made a mean of 1.57 AOM visits (SD: 0.89 visits) in the 6 months before an index OM visit, whereas the mean number of AOM visits from birth to an index OM episode was 3.36 (SD: 3.07 visits).
Antibiotics Used to Treat Uncomplicated AOM
The proportion of AOM visits treated with high-dose versus usual-dose amoxicillin changed over the 9-year study period (P < .0001), with an sharp increase occurring from 1998 to 2000 (Fig 2). Usual-dose amoxicillin was the most frequently used antibiotic during the first 7 years of the study, but high-dose amoxicillin supplanted it in 2003 and 2004. High-dose amoxicillin was more frequently used to treat preschool-aged children. In 2004, high-dose amoxicillin was used to treat 84.2% of AOM visits made by 7- to 12-month-old children, compared with 79.2% for 2- to 6-month-old children, 63.9% for 1- and 2-year-old children, 66.0% for 3- and 4-year-old children, and 26.0% for 5- to 12-year-old children (P < .0001).
The frequency with which other antibiotics were used in AOM treatment also changed during the study, although the effect sizes and rates of increased were modest, compared with those for amoxicillin (Fig 2). The frequency of azithromycin use increased significantly during the study, from 0% of AOM episodes in 1996 to 13.0% in 2004 (P < .0001) (the Food and Drug Administration approved azithromycin for use for children with AOM in 199514). In contrast, the use of trimethoprim/sulfamethoxazole decreased significantly over the study period, from 13.0% of AOM episodes in 1996 to 0.3% in 2004 (P = .0001).
Treatment Failure and Relapse
Overall, during the 9-year study period, 3.5% of uncomplicated AOM visits resulted in treatment failure and 9.3% in relapse. The proportion of uncomplicated AOM visits followed by treatment failure decreased from 3.9% in 1996 to 2.6% in 2004 (P < .0001). The proportion of uncomplicated AOM visits followed by relapse decreased from 9.2% in 1996 to 8.0% in 2004 (P < .0001) (Table 1). The annual incidence densities of treatment failure and relapse also differed significantly over the 9-year study period (P < .0001 for both) (Fig 3). The antibiotics most commonly used to treat cases of OM treatment failure and OM relapse were trimethoprim/sulfamethoxazole (26.0% of OM treatment failures and 26.6% of OM relapses), erythromycin/sulfisoxazole (21.3% of OM treatment failures and 19.7% of OM relapses), amoxicillin/clavulanic acid (21.1% of OM treatment failures and 19.9% of OM relapses), and cefixime (11.3% of OM treatment failures and 8.6% of OM relapses).
Association of Amoxicillin Dose With Treatment Failure and Relapse Rates
In unadjusted analyses, the proportions of AOM visits treated with high-dose amoxicillin that resulted in OM treatment failure or OM relapse initially seemed higher than the event rates for AOM visits treated with usual-dose amoxicillin (P < .0001 for both comparisons). Of AOM visits treated with high-dose amoxicillin, 3.4% resulted in OM treatment failure and 10.2% resulted in OM relapse, whereas 3.2% of AOM visits treated with usual-dose amoxicillin resulted in OM treatment failure and 8.2% resulted in OM relapse.
In primary multivariate analyses of first-ever AOM visits treated with amoxicillin, the use of high-dose amoxicillin was not an independent predictor of subsequent treatment failure (odds ratio [OR]: 0.99; 95% confidence interval [CI]: 0.77–1.28) or relapse (OR: 1.00; 95% CI: 0.85–1.18), after adjustment for variables associated with OM treatment failure or OM relapse (age, calendar year, and gender) and clustering according to practice site (Table 2). Although high-dose amoxicillin treatment did not predict treatment failure or relapse, subject age strongly predicted both outcomes. With children 5 to 12 years of age as the reference group, infants 7 to 12 months of age had the highest adjusted odds of OM treatment failure (OR: 4.07; 95% CI: 2.81–5.90) and OM relapse (OR: 5.09; 95% CI: 3.95–6.55).
In secondary analyses of all amoxicillin-treated AOM visits (not just first-ever AOM visits treated with amoxicillin), greater numbers of AOM visits since birth or in the previous 6 months were both associated with increased likelihoods of treatment failure or relapse (all P < .0001). Similarly, greater numbers of AOM visits since birth or in the previous 6 months were both associated with the use of high-dose amoxicillin to treat uncomplicated AOM visits (all P < .0001). However, in multivariate analyses of all AOM visits that controlled for the numbers of AOM visits since birth and in the previous 6 months, in addition to age, calendar year, and gender, the use of high-dose amoxicillin did not independently predict subsequent treatment failure (OR: 0.97; 95% CI: 0.81–1.16) or relapse (OR: 1.01; 95% CI: 0.91–1.12).
We found that the incidences of antibiotic-treated uncomplicated AOM, treatment failure, and relapse decreased from 1996 to 2004. In fact, antibiotic-treated AOM was twice as common in 1996 as it was in 2004. During this 9-year period, the use of high-dose amoxicillin to treat AOM increased and the use of usual-dose amoxicillin decreased. These changes in amoxicillin use began coincident with the publication of the CDC AOM treatment guidelines recommending the use of high-dose amoxicillin.9 Although the AAP subsequently recommended high-dose amoxicillin for AOM treatment,1 both the CDC and AAP recommendations were based on microbiologic data showing increasing penicillin nonsusceptibility among S pneumoniae and on expert opinion, rather than on studies of actual treatment outcomes. Observational studies in the setting of such changes in treatment strategy can provide important information about their effectiveness in actual practice.
Comparisons With Other Studies
Our results are consistent with past work and extend our epidemiological understanding of OM outcomes. Our findings that the incidence rates of antibiotic-treated AOM decreased from the middle 1990s to the middle 2000s are consistent with research that found decreasing use of antibiotics for a variety of conditions in the 1990s.6,7
Our study is unique in describing trends in OM treatment failure and relapse incidences over 9 years encompassing the publication of the CDC OM guidelines in 1999. Other studies of antibiotic use observed a decrease in the incidence of OM before 2003, but none also evaluated OM treatment failure or relapse. We found rates of treatment failure and relapse similar to those found in studies of AOM episodes in the middle to late 1990s using Colorado Medicaid2 and Medical Expenditure Panel Survey3 data. Our study found rates of treatment failure and relapse lower than those found in Quebec4 and higher than those found in St Louis,5 possibly because of differences in the definitions of these events. High-dose amoxicillin was used to treat AOM more often in our setting than in St Louis5 or Quebec4 during similar years, surpassing the use of usual-dose amoxicillin in 2003.
We did not find that the use of high-dose amoxicillin, compared with usual-dose amoxicillin, for uncomplicated AOM treatment independently predicted subsequent treatment failure or relapse. With the acknowledgment that clinicians might have chosen high-dose amoxicillin for patients with more-severe AOM, we restricted our primary analysis to the first-ever uncomplicated AOM episode, in an effort to minimize confounding by indication. Results of these analyses were consistent with our secondary analyses of all uncomplicated AOM episodes that adjusted for patients’ AOM history; neither showed an independent effect of high-dose amoxicillin on failure or relapse, compared with usual-dose amoxicillin.
The design of this study did not enable us to discern why the incidence of AOM decreased markedly between 1998 and 2004. It seems unlikely that the microbiologic flora or anatomic susceptibility of children to OM changed markedly during this period. A recent study of children in Tennessee and New York found that frequent OM and tympanostomy tube insertion rates decreased between 1998 and 2002.15 The investigators suggested that this decrease might be attributable to the introduction of pneumococcal conjugate vaccine in the United States in 2000, but they were not able to include pneumococcal vaccination status in the analysis. Our study of AOM visits over 9 years (1996–2004), including the years studied by Poehling et al15 (1998–2002), demonstrated similar decreases in the incidences of uncomplicated OM, treatment failure, and relapse. However, we do not think that the introduction of pneumococcal vaccine explains our findings, for 2 reasons. First, the incidences of AOM, OM treatment failure, and OM relapse in our study population decreased at least as much from 1996 to 2000 as they did from 2000 to 2004. Second, randomized trials demonstrated that pneumococcal conjugate vaccine has a modest effect on OM.16,17 We think that the explanation most consistent with our results is that clinicians have substantially changed their treatment threshold for AOM, likely in response to pressure to decrease their prescribing of antibiotics. Such a change in diagnostic threshold would explain decreases in relapse and failure rates as well, even over a period in which nonsusceptibility to commonly used antibiotics has increased.
These results represent AOM diagnosed and treated by ∼80 different clinicians in a community setting with no routine system in place to standardize diagnostic and treatment thresholds, serving a large population of relatively diverse race/ethnicity. However, the study population comes from 1 geographic area, and the patients’ socioeconomic status tended to be higher than that of the general US population. Almost all subjects had private insurance; therefore, these results may be less generalizable to uninsured patients.
Our finding that the use of high-dose amoxicillin did not reduce the rates of treatment failure or relapse should be interpreted with caution. We were able to adjust for many other variables in these analyses, but we did not have access to data on all possible confounders. We attempted to control for confounding by indication, but our computerized data did not contain measures of the clinical severity of AOM. Although we were unable to control for subjects’ immunization status, we assume that rates of receipt of the pneumococcal conjugate vaccine were uniformly high, because immunization coverage rates among Harvard Pilgrim Health Care enrollees are consistently >90%.18 We identified OM on the basis of coded diagnoses, and we did not have data on clinical severity or diagnostic accuracy. Therefore, if children with more-severe AOM were more likely to be treated with high-dose amoxicillin than with usual-dose amoxicillin, our findings that high-dose amoxicillin was not associated with lower rates of treatment failure or relapse might have missed finding a true benefit of this regimen. Our findings should not be interpreted as undermining recommendations to use high-dose amoxicillin, because use of this treatment regimen may be justified to prevent the spread of resistant bacteria on a population level, although this assertion should be tested.
Uncomplicated AOM, treatment failure, and relapse have all become less commonly diagnosed over the past decade. The use of high-dose amoxicillin became more frequent during this period but was not associated with reduced likelihood of treatment failure or relapse of individual infections. The trends we observed most likely reflect temporal changes in AOM treatment thresholds, suggesting that clinicians are applying stricter diagnostic criteria. These changes in practice are consistent with recommendations by the CDC9 and the AAP1 and probably represent several converging influences, including concerns regarding antibiotic overuse expressed in many different forums.
- Accepted August 23, 2007.
- Address correspondence to Colin M. Sox, MD, MS, Department of Pediatrics, Boston Medical Center, Dawling 3 South, 771 Albany St, Boston, MA 02118-2393. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
What's Known on This Subject
Little is known about trends in acute otitis media, treatment failure, and relapse in recent years or whether high-dose amoxicillin use is associated with acute otitis media outcomes.
What This Study Adds
This study clarifies that acute otitis media, treatment failure, and relapse became less common as high-dose amoxicillin use increased. However, high-dose amoxicillin treatment did not reduce the risk of individual infections resulting in adverse outcomes.
- ↵American Academy of Pediatrics, Subcommittee on Management of Acute Otitis Media. Diagnosis and treatment of acute otitis media. Pediatrics.2004;113 (5):1451– 1465
- ↵Berman S, Byrns PJ, Bondy J, et al. Otitis media-related antibiotic prescribing patterns, outcomes, and expenditures in a pediatric Medicaid population. Pediatrics.1997;100 (4):585– 592
- ↵Howard DH, McGowan JJ. Initial and follow-up costs by treatment outcome for children with respiratory infections. Pediatrics.2004;113 (5):1352– 1356
- ↵Garbutt J, Jeffe DB, Shackelford P. Diagnosis and treatment of acute otitis media: an assessment. Pediatrics.2003;112 (1):143– 149
- ↵Finkelstein JA, Stille C, Nordin J, et al. Reduction in antibiotic use among US children, 1996–2000. Pediatrics.2003;112 (3):620– 627
- ↵Lee GM, Fleisher GR, Harper MB. Management of febrile children in the age of the conjugate pneumococcal vaccine: a cost-effectiveness analysis. Pediatrics.2001;108 (4):835– 844
- ↵Koepsell TD, Weiss NS. Incidence rate. In: Koepsell TD, Weiss NS, eds. Epidemiologic Methods: Studying the Occurrence of Illness. Oxford, England: Oxford University Press; 2003:45–50
- ↵Pfizer. Pfizer's Zithromax receives FDA approval as a once-daily, three-day treatment for acute bacterial sinusitis. Pfizer News.2004. Available at: www.pfizer.com/pfizer/are/news_release/2004pr/mn_2004_0120.jsp. Accessed March 2, 2007
- ↵Poehling KA, Szilagyi PG, Grijalva CG, et al. Reduction of frequent otitis media and pressure-equalizing tube insertions in children after introduction of pneumococcal conjugate vaccine. Pediatrics.2007;119 (4):707– 715
- ↵Lin ND, Kleinman K, Chan KA, et al. Impact of the introduction of pneumococcal conjugate vaccine on immunization coverage among infants. BMC Pediatr.2005;28 (5):43
- Copyright © 2008 by the American Academy of Pediatrics