Changes in Antibiotic Prescribing for Children After a Community-wide Campaign FREE

Increasing resistance to antibiotics is compromising treatment of Streptococcus pneumoniae and other serious bacterial infections.^1- 2 "Antibiotic pressure," the use and misuse of large quantities of antibiotics, is the driving force behind the worldwide resistance phenomenon.^3- 4 Concern regarding this problem has motivated a review of antibiotic prescribing patterns and treatment guidelines.^5- 10

Children represent a population of particular concern because they have the highest rates of antibiotic use and infection with antibiotic-resistant pathogens.^2,11- 12 By the early 1990s, US children younger than 15 years were receiving an average of almost 1 oral antibiotic prescription per year, primarily for the treatment of respiratory infections.¹¹ A substantial proportion of this prescribing is thought to be unnecessary.^13- 14 Responsible antimicrobial use as a means to curtail drug resistance has been broadly promoted by the Centers for Disease Control and Prevention (CDC), the American Academy of Pediatrics, and other groups and institutions.^5,15 In the United States, reductions in antibiotic prescribing have been demonstrated in targeted primary care practices¹⁶ and in 2 relatively isolated rural populations.^17- 18 However, little is known about the effectiveness of broader community- or state-level campaigns.^5,19

Tennessee had the highest rate of prescription utilization in the United States during 1998, based on capitated retail sales data.²⁰ Similarly, utilization rates for penicillins, cephalosporins, and trimethoprim-sulfamethoxazole in Tennessee were at least 20% higher than national averages.²⁰ Within Tennessee, strains of S pneumoniae isolated from invasive infections have demonstrated a high prevalence of penicillin resistance (35%)²; the highest resistance levels (>50%) were reported for Knox County (1998 population, 375 000²¹; county seat: Knoxville).²² In response to this problem, and stimulated by a recent multidrug-resistant pneumococcal meningitis outbreak in a nearby county,²³ the Knox County Health Department (KCHD) conducted a multifaceted educational intervention. In this article, we evaluate the intervention's effect on pediatric antibiotic prescribing.

The KCHD led a year-long community-wide intervention (May 1997 through April 1998) to decrease unnecessary antibiotic use among children. In 1997, KCHD created the East Tennessee Drug Resistance Task Force, with representatives of state, regional, and county health departments, a university medical center, a commercial laboratory, a managed care organization, and the CDC, to develop and implement the intervention. The campaign was directed at 3 main audiences: (1) 250 key health care providers (eg, pediatricians and family physicians) who provided most routine health care services to Knox County children; (2) parents of young children; and (3) the general public.

The educational campaign stressed that (1) antibiotics should be used only for bacterial infections; (2) colds and most coughs and sore throats are caused by viruses and should not be treated with antibiotics; and (3) when antibiotics are prescribed, they should be narrow spectrum. Major components of the campaign were as follows.

Provider education consisted of (1) lectures by a CDC physician to 150 of the key providers; (2) presentations at hospital staff meetings, grand rounds, continuing medical education seminars, resident conferences, and targeted primary care clinics; (3) distribution of new prescribing guidelines for pediatric respiratory infections⁶ to the 250 key providers; and (4) related articles published in the KCHD quarterly newsletter mailed to all 1500 Knox County physicians.

Parent education consisted of (1) distribution of 40 000 pamphlets to parents of all children in day care and grades K-3; (2) distribution of patient education materials to the 250 key providers; and (3) distribution of pamphlets to the parents of every newborn (incorporated with informational mailings regarding childhood immunizations).

Public education consisted of (1) distribution of 30 000 pamphlets to hospitals, clinics, and dental offices; (2) communication of information through television, radio, and newspaper coverage and public service announcements; (3) distribution of 38 000 pamphlets to families receiving influenza vaccine; and (4) distribution of 53 000 pamphlets to pharmacy clients. One of the KCHD staff members, a medical epidemiologist, managed the campaign and performed the majority of the educational presentations. The majority of the educational materials (eg, pamphlets entitled "Antibiotics and Your Child") distributed in the campaign were developed by and are available from the CDC.^24- 25

Because there have been both general and localized efforts to promote the appropriate use of antibiotics,²⁵ the impact of the Knox County intervention was evaluated by comparing outcomes with suitable control counties. Tennessee's 3 other urbanized counties were selected; these counties and their respective county seats were Shelby (Memphis), Davidson (Nashville), and Hamilton (Chattanooga). The control counties are geographically distant from Knox County and did not conduct similar community-wide interventions. Prescription of oral antibiotics to children was compared between the intervention county and control counties during the 12-month periods before, during, and after the intervention. The intervention-attributable effect was defined as the excess percentage reduction in prescription rates in Knox County relative to the control counties. The study also examined rates of outpatient visits for respiratory illnesses, ratios of prescriptions to respiratory illness visits, and antibiotic resistance among pneumococcal isolates from invasive infections. Because the absolute rate of antibiotic use was of interest, prescriptions were not linked to specific office visits and diagnoses, thus avoiding problems or omissions associated with mistakes or shifts in coding,²⁶ and prescriptions resulting from telephone consultations or emergency department visits.

The study population included children younger than 15 years who were residents of 4 counties (Knox, Davidson, Hamilton, and Shelby) and who were enrolled in the Tennessee Medicaid Managed Care Program (TennCare). On January 1, 1994, Medicaid was supplanted in Tennessee by TennCare, a managed care system that extended health insurance coverage to approximately 50% more persons than were eligible for Medicaid and shifted the care of enrollees to physicians in private practice.²⁷ The TennCare enrollment file identifies persons who are eligible to receive benefits; the specific dates of coverage; and the sex, race, date of birth, and county of residence of the enrollees. Children entered the study on the first day after May 1, 1996, on which they met the inclusion criteria, which were enrollment in TennCare within 30 days of birth or for 1 year, age younger than 15 years, and residence in 1 of the 4 study counties. Children were followed until loss of coverage by TennCare, age 15 years, death, or April 30, 1999. Person-time as a hospital inpatient was excluded. Children not designated as either white or black represented a small proportion (4%) of the eligible population and were also excluded. Person-time contributed by children included in the study was expressed in person-years and grouped according to age (<1 year, 1 to <5 years, and 5 to <15 years), county of residence, and race.

The 12-month periods before (May 1996 through April 1997), during (May 1997 through April 1998), and after (May 1998 through April 1999) the educational intervention were categorized as years 1, 2, and 3 for this study. Computerized TennCare files for 1996 through 1999 were used to determine the primary study outcome, oral antibiotic prescription use, and a secondary outcome, outpatient visits to a physician for a diagnosed respiratory illness, for enrolled children during the 3 study years. These files contained information on medical services reimbursed, including dates of medical services received on an outpatient basis, associated diagnoses and procedures, and prescriptions filled. Prescriptions were not linked with visits, and no more than 1 visit per patient per day was counted.

Prescriptions included were those filled for antimicrobial drugs administered orally and typically used for treatment of respiratory infections in pediatric outpatients. Antibiotic categories included were penicillins (eg, amoxicillin and ampicillin), cephalosporins, trimethoprim-sulfamethoxazole, and macrolides/other (eg, erythromycin, erythromycin-sulfisoxazole, azithromycin, and clindamycin). For the purpose of this study, respiratory illness was defined as an outpatient diagnosis of otitis media (International Classification of Diseases, Ninth Revision [ICD-9]²⁸ codes 381-382), common cold (ICD-9 code 460), sinusitis (ICD-9 codes 461, 473), pharyngitis (ICD-9 code 462), tonsillitis (ICD-9 code 463), laryngitis/tracheitis (ICD-9 code 464), bronchitis (ICD-9 codes 466, 490, 491); pneumonia and influenza (ICD-9 codes 480-487); or unspecified acute respiratory illness (ICD-9 code 465).

Data on cases of invasive S pneumoniae infections occurring in Knox County children younger than 15 years were provided by an ongoing, active, population-based surveillance system sponsored by the CDC (the Active Bacterial Core Surveillance program²⁹). A case of invasive pneumococcal disease was defined as isolation of S pneumoniae from blood, cerebrospinal fluid, or other usually sterile site. Isolates were tested for antimicrobial susceptibility using the broth microdilution method by a designated reference laboratory. Susceptibilities to penicillin, cefotaxime, trimethoprim-sulfamethoxazole, and erythromycin were examined. Isolates were defined as resistant in this study if they exhibited intermediate- or high-level resistance to a particular antibiotic.

Prescription rates were calculated by dividing the number of antibiotic prescriptions administered in ambulatory care by the person-years observed. Prescription rates were stratified by study county, study year, race, age group, and antibiotic category and were expressed per 100 person-years. When stratified by race, baseline prescription rates were similar among the 3 control counties (Table 1). For this reason, the control counties were combined in the subsequent analyses.

Negative Binomial Regression models³⁰ were used to estimate adjusted prescription rates, test whether a significant intervention-attributable effect existed, and examine potential interactions among the intervention-attributable effect with age and race. Each model included variables representing county (Knox vs control), age group (<1 year, 1 to <5 years, and 5 to <15 years), race (white and black), and study year (year 1, 2, and 3), as well as all relevant interaction terms. In addition, each model included random effects to account for unobserved heterogeneity within the counties. The intervention-attributable effect was represented by the excess percentage change in prescription rates from year 1 to year 3 for Knox County over and above any temporal reductions observed in the control counties. The intervention-attributable change in the crude rate was determined by multiplying the intervention-attributable excess percentage change by the baseline rate in Knox county and represents the estimated reduction in the number of prescriptions in Knox county in year 3 attributable to the intervention.

Visit rates, calculated by dividing the number of outpatient respiratory illness visits by the person-years observed, and ratios of antibiotic prescriptions to the number of outpatient visits (ie, prescription to visit ratios) were examined as secondary outcome variables. Both were analyzed in a similar manner as described above for antibiotic prescription rates.

The prevalence of resistance among S pneumoniae isolates from invasive infections occurring in Knox County was calculated using the number of resistant isolates divided by the total number of isolates tested; these rates were expressed as percentages and stratified by study year and antibiotic category. P = .05 was the level of significance; basic analyses were performed using SAS Version 8.2 (SAS Institute Inc, Cary, NC).

The 3-year study encompassed 464 200 person-years of observation representing 36% of all children younger than 15 years residing in the 4 study counties (Table 1). The average annual study population ranged from approximately 18 000 to 84 000 children per county per year. While the study populations were similar with regard to age and sex, the racial makeup differed substantially, consistent with underlying demographic differences among the counties; the Knox County study population had the highest proportion of white children (73%), whereas Shelby County had the highest proportion of black children (90%).

The total baseline (year 1) rates of antibiotic prescriptions and respiratory illness visits reflected the racial composition of each county's study population, with higher rates corresponding to higher proportions of white children (Table 1). For each of the 4 counties, both prescription and outpatient visit rates among white children were approximately twice as high as among black children. However, even when white and black children were examined separately, Knox County's prescription and outpatient visit rates remained elevated relative to the control sites.

Overall, antibiotic prescription rates declined 19% and 8% from year 1 to year 3 among Knox County and control county children, respectively, yielding an 11% (95% confidence interval [CI], 8%-14%) intervention-attributable decline (P<.001) in prescriptions (Table 2). In Knox County, prescription rates declined by 16% and 21% for white and black children respectively (Figure 1), with most of the decline occurring during year 2 when the intervention was conducted. Among controls, the decrease in prescription rates from year 1 to year 3 was more evident among white children (12%) than black children (3%). The changes in Knox County's prescription rates exceeded these secular trends for black children (P<.001) but not white children (P = .18).

Antibiotic prescription rates in Knox County and in the control counties decreased among all age and race groups examined (Table 2). These age-specific decreases were similar in magnitude among white and black children in Knox County, but not in the control counties where the decrease was more pronounced among white children. The intervention's effect on prescription rates was significant (P<.001) for black children in all 3 age groups (attributable decreases ranging from 16%-20%) and for white children in the 1- to less than 5-year age group (8% attributable decrease). The intervention-attributable effects, expressed in terms of their impact on baseline prescribing rates, were equivalent to decreases of 18 and 25 prescriptions per 100 child-years among white and black children aged 1 to less than 5 years, respectively.

The observed changes in prescription rates did not appear to result from changes in respiratory illness visit rates; these rates were relatively stable during the 3-year study period (Figure 1), and there was no evidence of an unusual respiratory illness season among the study sites. The ratios of antibiotic prescriptions to respiratory illness visits were also examined (Figure 1). At baseline, the prescription to visit ratios were similar when comparing Knox County and the control sites. During the intervention year, the prescription to visit ratio for both black and white children declined substantially in Knox County, but not among the controls. By year 3 (postintervention), the prescription to visit ratio in Knox County had decreased approximately 20% among both race groups; for black children the excess decline relative to the controls was 13% (P<.001), while for white children it was 8% (P = .06). In the 1- to 4-year age group, the prescription to visit ratios declined significantly among both white children (11%, P = .02) and black children (17%, P<.001).

Use patterns for specific antibiotics were generally similar between Knox County and the control counties and for both white and black children (Figure 2). Overall, the majority of antibiotic prescriptions were for penicillins, and this proportion increased significantly (P<.001) during the study period among black and white children in the intervention and control sites. Prescription rates among black and white children in Knox County declined for all of the antibiotic categories examined. These declines were consistently greater among Knox County children than among controls except for cephalosporins, which had a significantly greater decline in white children in the control counties.

Resistance rates among isolates of invasive S pneumoniae occurring in Knox County children remained high (≥50%) from year 1 through year 3 for all 4 antibiotic categories examined. Reductions in S pneumoniae resistance rates were not evident (Table 3).

Knox County's public health community recognized the problem of antibiotic resistance and responded by conducting an educational intervention. Our evaluation used Medicaid data to demonstrate that antibiotic use declined substantially among Knox County children in concert with the campaign. When control counties were examined, a secular downward trend in antibiotic prescription rates was documented; this finding is in agreement with other recent reports^31- 32 and may reflect the impact of both national and local efforts to promote appropriate antibiotic use. In Knox County, the reduced antibiotic prescribing was beyond that observed in the control counties, as demonstrated by 2 measures: population-based prescription rates and prescription to visit ratios. Since this study was not a controlled clinical trial, it is possible that the observed reduction in prescribing in Knox County was due to other unmeasured factors or to the greater influence of temporal trends in a region with high baseline prescribing. However, the strong temporal association of the "excess reduction" in antibiotic prescribing that occurred coincident with the intervention confirmed our a priori hypothesis and suggests that this was indeed an effect of the intervention.

To our knowledge, this is the first report to demonstrate reductions in pediatric antibiotic prescribing as a consequence of a community-wide educational intervention in a large urbanized setting. Of note, the magnitude of the intervention-attributable reductions in antibiotic prescription rates reported herein was similar to decreases reported in other recent interventions targeting pediatric antibiotic use.^17- 18,32

Certain aspects of the intervention were noteworthy. The intervention was spearheaded by the local county health department and conducted on a limited budget using existing personnel. A coalition was developed that included key members of the local medical community. The educational efforts engaged both medical providers and parents, while also reaching the general public. Reductions in prescribing were sustained for at least 1 year after Knox County's campaign ended, though the greatest effect occurred during the intervention year. Although encouraging, this suggests that interventions might need to be maintained over longer periods.

The evaluation was limited to TennCare enrollees, which to some degree controlled for access to care and socioeconomic factors. The effect of the intervention on children of higher socioeconomic status, is not known. On the other hand, the study population was large and represented a substantial portion of the total pediatric population. Also, the exact mechanism by which prescription rates were lowered in Knox County could not be ascertained. Although prescriptions were not linked to actual visits, the prescription to visit ratio provided a measure of the likelihood that a respiratory illness encounter resulted in an antibiotic prescription. The decreased ratios of prescriptions to visits suggest that, as a result of the educational intervention, Knox County physicians adopted more stringent criteria for writing antibiotic prescriptions and/or that parents' expectations for receiving prescriptions was reduced.

The intervention effect was greater among black children, even though their baseline antibiotic prescription rates were lower. White children received care for respiratory illnesses approximately twice as often as black children. Higher rates of ambulatory health care among white children for otitis media and other respiratory conditions have been described previously.^33- 34 The reasons why white children received physician care for respiratory illnesses more often than black children in our study are unknown and could not be examined directly. Because this study included only children who were covered by TennCare, differences in access to care should have been minimized. Antibiotic prescription rates also were approximately twice as high among white children as black children. Consequently, much of the observed excess in antibiotic prescribing at the intervention site appeared to be caused by the high proportion of white children in Knox County. Within Knox County, and among the control sites, prescribing levels among black and white children were similar when differences in respiratory illness visit rates were accounted for. These data indicate a need to examine the appropriateness of different levels of antibiotic and ambulatory care use.

Younger children have the highest rates of antibiotic use.¹² For example, prescription rates among children aged 1 to less than 5 years were approximately twice as high as among children aged 5 to less than 15 years in our study. Therefore, educational interventions should target the parents of preschoolers, as was done in Knox County. In fact, the largest intervention effects in Knox County were observed among children aged 1 to less than 5 years.

The correlation between levels of ambulatory care visits and antibiotic prescriptions identified in our study suggests that the provider visit itself could be a driving force behind antibiotic prescribing. In other words, unnecessary office visits might precipitate unnecessary prescriptions.^35- 37 Therefore, more attention should be given to the education of parents and child care center personnel regarding the appropriate indications for seeking physician care and antibiotic treatment for children with acute respiratory illnesses.

Among this study population, a general decline occurred across antibiotic categories. A primary message of this intervention was that narrow-spectrum antibiotics should be used preferentially. Our data indicated that even before the intervention, narrow-spectrum penicillins dominated antibiotic use and that, during the study period, the proportion of antibiotic use accounted for by penicillins increased. Limited data were available regarding resistance among invasive S pneumoniae isolates among Knox County children. There was no indication of any decrease in pneumococcal resistance 1 year after the Knox County intervention; however, the sample size limited our ability to detect modest changes in resistance. It is likely that further sustained reductions in antibiotic use will be required to lower the prevalence of antibiotic resistance among S pneumoniae and other community-acquired bacterial pathogens. Clonal spread of resistant strains is thought to contribute to their persistence after introduction to a community,¹ and substantial time lags are expected to accompany their disappearance.³⁸

A national action plan to address the problem of antibiotic-resistant infections in the United States was recently developed by a governmental task force.^5,39 Educational interventions to counter the development and spread of microbial resistance through promotion of responsible antibiotic use were identified as essential to this plan. The decline in antibiotic prescribing documented after the Knox County intervention should encourage these broader efforts.