OBJECTIVES: Previous analyses of data from 3 large health plans suggested that the substantial downward trend in antibiotic use among children appeared to have attenuated by 2010. Now, data through 2014 from these same plans allow us to assess whether antibiotic use has declined further or remained stable.
METHODS: Population-based antibiotic-dispensing rates were calculated from the same health plans for each study year between 2000 and 2014. For each health plan and age group, we fit Poisson regression models allowing 2 inflection points. We calculated the change in dispensing rates (and 95% confidence intervals) in the periods before the first inflection point, between the first and second inflection points, and after the second inflection point. We also examined whether the relative contribution to overall dispensing rates of common diagnoses for which antibiotics were prescribed changed over the study period.
RESULTS: We observed dramatic decreases in antibiotic dispensing over the 14 study years. Despite previous evidence of a plateau in rates, there were substantial additional decreases between 2010 and 2014. Whereas antibiotic use rates decreased overall, the fraction of prescribing associated with individual diagnoses was relatively stable. Prescribing for diagnoses for which antibiotics are clearly not indicated appears to have decreased.
CONCLUSIONS: These data revealed another period of marked decline from 2010 to 2014 after a relative plateau for several years for most age groups. Efforts to decrease unnecessary prescribing continue to have an impact on antibiotic use in ambulatory practice.
- AIC —
- Akaike information criterion
- CDC —
- Centers for Disease Control and Prevention
- UTI —
- urinary tract infection
What’s Known on This Subject:
Outpatient antibiotic use for children has decreased substantially since the 1990s. Previous evidence from 3 geographically diverse health plans suggested a slowing of this decline by 2010. Additional concerns included increasing use of broad-spectrum macrolides and third-generation cephalosporins.
What This Study Adds:
Analysis of data from 3 health plans revealed new, substantial declines in outpatient antibiotic use between 2010 and 2014, despite previous evidence that a plateau had been reached. Prescribing has decreased for diagnoses for which antibiotics are clearly not indicated.
The Centers for Disease Control and Prevention (CDC) estimates that 2 million individuals in the United States acquire antibiotic-resistant infections each year.1 These include organisms that are found primarily in hospitalized patients but also those commonly carried in the community (eg, Streptococcus pneumoniae).2,3 Although many factors promote resistance, experts agree that the high frequency of use of antibiotics in outpatient settings contributes directly to resistance in the community.4–6
Public health officials, professional societies, and researchers have been on a decades-long quest to decrease unnecessary antibiotic use in both children and adults as a way to slow the spread of resistant pathogens.7–9 Guidelines have been developed and implemented by professional societies for key conditions,10–13 large-scale clinician and patient education interventions have been employed,14–16 and other clinician behavior change interventions have been tested.17,18 The CDC has recently summarized the action steps that can result in more judicious prescribing in outpatient settings that complement recommendations for in-hospital antibiotic stewardship.19
Although antibiotic resistance continues to be a major public health concern, patterns of antibiotic use have changed substantially for children in outpatient settings from the mid-1990s. For example, data from the National Ambulatory Medical Care Survey documented decreasing rates of prescribing to children in this time frame.20,21 We, and others, have measured dispensing rates in the defined populations of health plans and have also documented substantial declines in young children since 1996.22,23 A previous analysis by this research team, on the basis of data from 3 large health plans, suggested that this downward trend in antibiotic use appeared to be over or at least had significantly attenuated by 2010.24 Now, data have become available through 2014 from these same plans. Our purpose with this study was to assess whether antibiotic use had declined further, rebounded, or continued at the levels seen in 2010.
For this study, we analyzed data through 2014 from 3 commercial health plans located in New England (A), the Mountain West (B), and the Midwest (C) regions of the United States. Data from these 3 plans have been monitored since 2000. Over this period, we have used consistent methods for determining rates and trends of antibiotic use (overall and by class) and assigning diagnoses to dispensing events.23,24 Plan A is a commercial health insurer that reimburses care provided by a large network of affiliated practices; Plan B is an integrated health care delivery system including a multispecialty group practice; and Plan C is an insurer with both an integrated health care delivery system and affiliated network practices.
The study included children 3 months to <18 years of age enrolled for a minimum of 7 consecutive days with concurrent pharmacy benefits. The data presented here include the period from September 2000 through August 2010 that were previously published24 and extends data through August 2014. For readability, we have, in places, abbreviated the label for the 2000–2001 year as “2000” and the 2013–2014 year as “2014.” This study was approved by the Institutional Review Board of Harvard Pilgrim Health Care; institutional review board oversight was ceded to Harvard Pilgrim Health Care by the 2 other health plans.
Data collection methods have previously been described in detail.23,24 Briefly, each study year was defined as the 12 months from September 1 to August 31 so that only 1 respiratory illness season was included. Using enrollment and disenrollment files, we calculated the number of days a subject contributed in each study year to each of the age categories of 3 to <24 months, 2 to <4 years, 4 to <6 years, 6 to <12 years, and 12 to <18 years. An individual could contribute data to multiple age groups as they aged, and some contributed to 2 age groups in a single year.
Oral antibiotics dispensed were identified in pharmacy claims data by using their National Drug Codes (updated as new products became available). Antibiotics were grouped into first-line penicillins (ie, penicillin v potassium, amoxicillin), amoxicillin-clavulanate, erythromycin, second-generation macrolides (azithromycin, clarithromycin), and first-generation (eg, cephalexin), second-generation (eg, cefuroxime), and third-generation (eg, cefdinir, cefixime) cephalosporins. Other oral antibiotics were grouped as “Other,” except for the tetracycline class for adolescents, which was analyzed as a separate group in this age stratum only. Antitubercular, antihelminthic, parenteral preparations, and topical antibiotics were excluded.
Insurance claims for in-person ambulatory, urgent care, and emergency department encounters were linked with antibiotic dispensings. The diagnosis assigned at the most recent visit was assigned to the antibiotic dispensing. A previously described algorithm was used to identify a primary International Classification of Diseases diagnosis for each visit in cases for which 2 or more diagnoses were listed. This algorithm, previously published,22 was used to prioritize respiratory tract diagnoses that would likely be treated with an antibiotic (eg, pneumonia, otitis media, etc) over less common urinary tract and skin and/or soft tissue infections and over infections that would not be treated with an antibiotic (eg, viral illness). If no visit occurred in the 3 days before an antibiotic dispensing, the dispensing was classified as “unlinked.”
Within each age group, we calculated population-based antibiotic dispensing rates (number of antibiotic dispensings divided by the number of person-years aggregated across individuals) by health plan for each study year. Overall percent declines in total antibiotic rates (with 95% confidence intervals) between the first and last study years were calculated for each age group and for each health plan.
With our previous analysis,24 we used the Akaike information criterion (AIC) to select a single inflection point in the decline in antibiotic use rates between 2000–2001 and 2009–2010 for each health system and age category. In that analysis, we first fit Poisson regression models that accounted for person-years enrolled. Each model allowed separate linear slopes in the years preceding and the years after a particular year in the interval between 2000–2001 and 2009–2010. The models differed by the year set as the inflection point with all years in this interval considered. The AIC was then used to select among these possible 2 slope models. The model with the smallest AIC provides the best balance between goodness of fit and overfitting. In that analysis, the postinflection periods identified each revealed a slower rate of decline in antibiotic use (leveling off) compared with the preinflection period.24
In the current analysis, we used similar methods to determine whether the antibiotic use trends continued across health plans and age groups through 2014. Using all data available between 2000–2001 and 2013–2014 for each health plan and age group, we fit multiple Poisson regression models but now allowed up to 2 inflection points. We considered the first inflection point that had been identified in our previous analysis (at which the rate of decline generally slowed). Each model differed by the location of the second inflection point, with all years between the first inflection point and 2013–2014 considered. The 3-slope model with the smallest AIC was selected as the final 3-slope model. In each case, we conducted a likelihood ratio (LR) test comparing the final 3-slope model to the 2-slope model that included only the inflection point from our previous analysis. Our null hypothesis was that the 3-slope model would not fit the data significantly better than the 2-slope model. We rejected this null hypothesis for LR test P values <.05. On the basis of the final models for each plan and age group, we calculated the per-year percent change in dispensing rates in each of the 3 periods (along with 95% confidence intervals), that is the period before the first inflection point, the period between the first and second inflection points, and the period between the second inflection point and 2013–2014.
In addition to the examination of overall rate changes over time, we also examined whether the relative contribution to overall dispensing rates of common diagnoses for which antibiotics were prescribed changed over this period. These diagnoses included otitis media; pharyngitis; sinusitis; viral respiratory tract infections; pneumonia; acne; and a composite of urinary tract infection (UTI), skin and/or soft tissue infection, and other bacterial infections. All analyses were conducted by using SAS 9.3 (SAS Institute, Inc, Cary, NC).
The period analyzed from 2000 to 2014 included 6 232 060 person-years of observation among children 0 to 18 years of age. Table 1 and Figs 1 and 2 show the observed decrease in antibiotic dispensing over this 14-year period in all 3 health plans. As expected, the highest prescribing rates were in the youngest age group. Consistent with findings in the previous studies in these health plans,24 antibiotic dispensing rates continued to be markedly lower in the integrated delivery system health plan in the Mountain West (Plan B) through 2013–2014.
Although it previously appeared that the decline in antibiotic use in all 3 plans had plateaued by 2009–2010, there were clear and substantial additional decreases between 2009–2010 and 2013–2014 (Figs 1 and 2). In all age groups and health systems, LR P values comparing the final 3-slope model to a 2-slope model were <.05. Thus, results reported in all cases are based on a 3-slope model. With Table 2, we report, by age group and health plan, the estimated percent decline in antibiotic use rate per year along with 95% confidence intervals on the basis of the 3-slope Poisson regression model in the years before the first inflection point, the years between the first and second inflection points, and years after the second inflection point. In almost every age group and health system, there was an identifiable period of flattening, followed by steeper decline. The second inflection point was selected between 2009 and 2012 in all but 1 age group in health plan B. For example, in children 3 to <24 months of age, before inflection 1, the estimated decline ranged from 5.0% to 9.3%; in the plateau period, estimated declines ranged from −0.15% to 2.3%; and after the second inflection, the estimated declines were observed to be 6.9% to 7.9% in the 3 health plans. Similar patterns are observed in other age groups. The lower bound of the 95% confidence interval for percent decline per year in years before the first inflection point and after the second inflection point excluded 0 across all age groups and health systems.
The number of dispensings per person-year for the common diagnoses is shown in Fig 3. Among children <2 years of age, we found that, although the overall rate of antibiotic prescribing for otitis has decreased dramatically (by 47% between 2000–2001 and 2013–2014), the fraction of all prescribing associated with otitis media was stable (63.8% in 2000–2001 and 64.5% in 2013–2014). In all children >2 years of age, the fraction of antibiotics prescribed for otitis media decreased modestly over the study period. Other bacterial infection diagnoses (pneumonia, UTI, or soft tissue infection) increased as a fraction of antibiotics prescribed, as did pharyngitis (except in children 12–18 years old). The fraction of antibiotics associated with diagnoses of viral illnesses decreased in all age groups. As has been seen previously,23,24 we were not able to link all antibiotic dispensings to a visit. The fraction of unlinked encounters decreased in every age group since 2000–2001 except for the 12- to 18-year-old group. Understanding the reasons behind this decrease would require more detailed information than is available through health plan claims data but may include less prescribing over the phone (without an in-person visit) as providers have become generally more judicious in their antibiotic prescribing. All changes in the relative frequency of diagnoses must be interpreted in the context of the dramatic decreases in antibiotic use presented above.
We also analyzed dispensings according to antibiotic class (Supplemental Figs 4–8). While penicillins (eg, amoxicillin) account for a large fraction of antibiotics dispensed, their use has decreased along with overall antibiotic dispensing rates. Concern has been focused more on broad-spectrum antibiotics whose use increased between 2000–2001 and 2009–2010. In this analysis, we observed that second-generation macrolide (eg, azithromycin) use has dropped markedly since 2009–2010 (between ∼35% and 49%) in the 2 plans that had relatively high use. The health plan that has historically had the most judicious antibiotic prescribing had slight increases in absolute prescribing of second-generation macrolides but still had the lowest rates of dispensing of these agents among the 3 plans. Third-generation cephalosporin use increased markedly from 2000–2001 to 2009–2010 (depending on the particular health plan). Since 2009–2010, we observed small relative changes (increases and decreases) that varied by age group and health plan, but overall did not see further substantial increases in their use.
Previous research, using multiple data sources, documented declines in antibiotic use among children beginning in the 1990s.20–23 Data from the 3 health plans studied here previously revealed that the rate of decrease in antibiotic use had slowed in the several years before 2009–2010.24 To our surprise, inclusion of more recent data (2010–2011 to 2013–2014) from the same health plans showed a period of marked decline after this relative plateau for most age groups. In general, the second period of substantial decline started between 2008–2009 and 2011–2012, depending on age and health plan. In aggregate, this has resulted in overall declines across all 3 health plans from 2000–2001 to 2013–2014 of 40% to 50% among children <6 years old, who have the highest rates of antibiotic use of any age group. Declines in the range of 20% to 40% were observed in older children and adolescents.
This is good news from the perspective of all those concerned about antibiotic resistance. It is reasonable to assume that attention by public health officials, the lay press, and professional organizations have all contributed to declines in outpatient antibiotic use by children. Specific interventions included national campaigns (led by the US CDC)25 and statewide interventions conducted as research or public health interventions.14 The largest contribution to decreasing antibiotic use among children before 2010, however, was likely the changing thresholds for diagnosis of otitis media, as well as studies suggesting observation without antibiotics for selected cases. Standards for diagnosis and the option for observation without antibiotics for some patients were both included in American Academy of Pediatrics guidelines for the diagnosis and treatment of acute otitis media in 2004.26 The proximate cause of the second wave of declines after 2010 observed here is less clear. Updated American Academy of Pediatrics guidelines for acute otitis media were released in 2013 that continued to allow observation without antibiotics for selected patients with nonsevere illness, depending on age, but with greater focus on the need for diagnostic certainty.10 On the basis of the data analyzed here we cannot comment on the specific contribution to decreasing rates of otitis media from the serotypes included in the 13-valent pneumococcal conjugate vaccine released in 2010. However, the decreased risk of serious bacterial infection among febrile infants and toddlers, because of pneumococcal (and other) vaccines, is now ingrained in the risk assessment of clinicians and has had, we believe, contributed to the decrease in antibiotic prescribing over time. In addition, antibiotic-resistant organisms continue to be the subject of frequent media reports that may drive changes in perceived parental requests for antibiotics. New antibiotic stewardship programs that are focused primarily on resistant organisms in hospital settings may have also played a role. Given the range of factors driving inappropriate prescribing27 and the myriad forces promoting more judicious prescribing, it will likely be impossible, in fact, to isolate a particular cause of the decline after 2010. What we observe in these and other data may merely represent the nonlinear nature of change in parental expectations and physician practice, as both become more accustomed to new approaches to treating minor infections.
From 2009–2010 to 2013–2014, we also saw good news related to the use of broad-spectrum macrolides and cephalosporins, after marked increases in the decade before. Neither of these is considered first-line treatment of most common infections. Although recent work on antibiotic selection for pneumonia continues to highlight this problem,28 their use appears to be decreasing in the health plans studied here.
The data source used in this study has both strengths and limitations. The data reported here extend only through 2014 and may not reflect more recent changes in antibiotic use particularly in light of continuing attention to antibiotic stewardship in both inpatient and community settings. Health plan dispensing data reliably includes actual dispensing events (which is more proximal to actual ingestion than prescribing but not as closely linked to prescriber behavior). Although our methods exclude plan members without pharmacy coverage as part of their insurance benefit, dispensing data will not include medicines purchased solely out of pocket or that are not covered by insurance for other reasons. Furthermore, typical pharmacy insurance claims do not include the diagnosis for which the agent was prescribed. Our assignment of diagnoses to dispensing events based on the most recent ambulatory visit claim (which does have a diagnosis code) may misclassify a small fraction of the diagnoses. Finally, these 3 health plans may not be representative of practice across the United States. In particular, the south region of the United States is not represented by these health plans; this region has been shown to have high rates of antibiotic prescribing for several age groups.29 However, the large amount of data from the same plans over the course of the entire 14-year study period allows important observations about changes over time in antibiotic use in 3 geographically dispersed US health plans.
We believe that child health professionals can be justifiably proud of the major change in practice that has occurred in outpatient prescribing of antibiotics. We do not, however, mean to imply that the job is done. Antibiotic resistance in both hospital and community settings remains a serious threat to public health. A substantial fraction of antibiotic prescribing in all age groups remains unnecessary.29 Recognizing that more work to promote judicious prescribing was warranted, the federal government launched a new action plan to combat antibiotic resistance, led by the US CDC.9,25 The data from one of our health plans, with rates consistently lower than the others, suggest that further reductions in antibiotic use are still possible while maintaining the benefits of this life-saving class of medications. However, we appear to be getting closer to the elusive balance between eliminating unnecessary prescribing, while treating illnesses for which there is evidence of benefit.
- Accepted October 1, 2018.
- Address correspondence to Jonathan A. Finkelstein, MD, MPH, Department of Pediatrics, Boston Children’s Hospital, 300 Longwood Ave, Boston, MA 02115. E-mail:
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.
FUNDING: Supported by a grant from the National Institute of Child Health and Human Development (K24HD060786). Funded by the National Institutes of Health (NIH).
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
- Centers for Disease Control and Prevention
- Butler JC,
- Hofmann J,
- Cetron MS,
- Elliott JA,
- Facklam RR,
- Breiman RF
- Paschke AA,
- Zaoutis T,
- Conway PH,
- Xie D,
- Keren R
- Costelloe C,
- Metcalfe C,
- Lovering A,
- Mant D,
- Hay AD
- Centers for Disease Control and Prevention
- Interagency Task Force
- The White House
- Hersh AL,
- Jackson MA,
- Hicks LA; American Academy of Pediatrics Committee on Infectious Diseases
- Shulman ST,
- Bisno AL,
- Clegg HW, et al
- Hersh AL,
- Shapiro DJ,
- Pavia AT,
- Shah SS
- Greene SK,
- Kleinman KP,
- Lakoma MD, et al
- Finkelstein JA,
- Stille C,
- Nordin J, et al
- Vaz LE,
- Kleinman KP,
- Raebel MA, et al
- American Academy of Pediatrics Subcommittee on Management of Acute Otitis Media
- Hersh AL,
- Kronman MP
- Handy LK,
- Bryan M,
- Gerber JS,
- Zaoutis T,
- Feemster KA
- Copyright © 2019 by the American Academy of Pediatrics