Dynamics of Antibiotic Prescribing for Children

The dynamics that influence antibiotic prescribing by office-based physicians in the United States are complex. Physicians have been receiving repeated messages to curtail antibiotic use from the biomedical literature, medical and public media, health insurance companies, key opinion leaders, alternative medicine advocates, and some patients. The message has been consistent, frequent, and loud, so it cannot have been missed. The crescendo of research during the past decade has established that (1) antibiotic overuse is a major public health problem¹ ; (2) approximately 50% of prescriptions for children written by community-based practitioners are unnecessary,² and (3) the single most important factor in the emergence of antibiotic resistance among respiratory bacterial pathogens is selection pressure from antimicrobial agents.³ Thus, as a public health policy, there seems little doubt that overzealous prescribing habits and inappropriate use of antibiotics should be reduced on a community-wide basis.

This all sounds good, but what about the individual patient seeking care? Faced with an ill-appearing, febrile child or adolescent, anxious parents, a busy office schedule, a potential insurance company audit of medical records, and the omnipresent concern for malpractice litigation, the physician must consider a different dynamic. Considerations when making treatment decisions for the individual patient include diagnostic uncertainty; sociocultural and economic pressures; meeting parent/patient and insurance company expectations; and taking defensive action in case of litigation.⁴ Occasionally, some bacterial infections that are preceded by viral illness have a rapid course, and patients may not always return to see the physician if the illness worsens or persists. Thus, the inclination is to err on the side of prescribing antibiotics even if the chance of a bacterial infection is low. The perceived risk for the individual patient to have an illness evolve with significant consequences often seems more important than the risks of dealing with an antibiotic-resistant organism later.⁴

In this issue of THE JOURNAL, McCaig and colleagues⁵ assess trends in antibiotic prescribing from 1989-1990 through 1999-2000 for US children and adolescents younger than 15 years. The authors used data from standardized office visit record forms completed by nationally representative samples of office-based physicians, who participated in annual National Ambulatory Medical Care Surveys (NAMCSs), to calculate population- and visit-based antibiotic prescribing rates. This study is a follow-up to a prior analysis of NAMCS data,⁶ which demonstrated increasing annual population-based rates of antibiotic prescribing in outpatient settings for pediatric patients from 1980 through 1992. The take-home message of the current study is that antibiotic prescribing appears to have decreased overall and for common respiratory tract infections among children. That is good news, if true.

Several factors other than appropriate use, however, might explain the observed decline in antibiotic prescribing. Some of the decline might be accounted for by a decreasing secular trend in physician visits over the study period. Perhaps the worried well and patients with milder illness are simply not visiting physicians as often. The NAMCS only captures drugs dispensed at a physician visit, so if physician visits decline, antibiotic prescriptions will decline at a population level. Visit-based rates, however, which do not depend on the number of office visits, also declined overall and for respiratory tract infections. Increased telephone dispensing of antibiotics might be another variable to explain the observed trends. McCaig et al used urinary tract infections as an indicator condition for changes in telephone prescribing. However, physicians might regard urinary tract infections quite differently than respiratory tract infections in terms of the need for in-office evaluations. The pattern of care and use of telephone triage for respiratory tract infections may not be the same as for urinary tract infections. Patient self-administration of leftover antibiotic prescriptions might also be occurring. One study suggests that up to 66% of children arriving for outpatient care have already been self-administering antibiotics.⁷

For otitis media and bronchitis, McCaig et al noted decreases in the population-based antibiotic prescribing rate, but not in the visit-based prescribing rate. The authors speculate that only patients with more serious infections came to physician offices; that diagnosis may have been more accurate; or that the incidence of otitis media and bronchitis may have decreased with no change in prescribing practices. However, there is no evidence to suggest that parents are waiting until their child appears to have a more serious illness before seeking care. The notion that diagnostic accuracy has improved also could be questioned. A recent study suggests that pediatricians have a poor knowledge about pneumatic otoscopy and misdiagnose otitis media about 50% of the time.⁸ With interactive continuing medical education instruction, sustained changes in otitis media diagnosis and an associated reduction in antibiotic use may occur.⁹ The possibility that the incidence of bronchitis may have decreased seems plausible because more bronchitis may be recognized as asthma.

Because of the considerable attention focused on antibiotic overprescribing, physicians may be changing their diagnostic labeling depending on whether they intend to prescribe antibiotics. For example, if a physician intends to treat a patient with a cough illness with antibiotics, the diagnosis might be recorded as bronchopneumonia or Mycoplasma pneumoniae, whereas if antibiotics are not prescribed, the diagnosis of bronchitis might be used. For a child with an upper respiratory tract infection, fever, loss of appetite, ear tugging, and a poor night of sleep, the physician might record a diagnosis of otitis media when antibiotics are prescribed, whereas the diagnosis of upper respiratory tract infection or otalgia might be used if antibiotics are not prescribed. To assess the possibility of diagnosis shifting, McCaig and colleagues analyzed prescribing for a combination of respiratory tract infection visit diagnoses. It is reassuring that the authors still found a significant overall decline in antibiotic prescriptions.

Another methodologic concern relates to the veracity of reporting antibiotic prescriptions. The numerous publications and public information campaigns by the Centers for Disease Control and Prevention, the American Academy of Pediatricians, American Academy of Family Physicians, and other professional organizations might have heightened awareness of criticism regarding excessive antibiotic use. As a consequence, physicians may have changed their behavior or they simply may have changed their responses to the NAMCS. In a recent Centers for Disease Control and Prevention study of pediatricians and family physicians, wide differences between survey responses regarding adherence to diagnostic criteria and treatment guidelines and actual antibiotic prescribing practice as determined by chart review were found for pediatric respiratory tract infections.¹⁰ For this concern, there are no remedies for NAMCS data because no reabstraction studies have been performed to determine the validity of NAMCS responses.

In this issue of THE JOURNAL, an accompanying article by Perz at al¹¹ allays some of the concerns inherent in interpretation of survey data and corroborates the secular trend for declining use of antibiotics. The investigators assessed the impact of a community-wide education campaign in Knox County, Tennessee, designed to reduce unnecessary antibiotic use. They compared trends in antibiotic prescribing rates among Medicaid recipients younger than 15 years in Knox County with those in 3 other (control) counties in Tennessee before, during, and after the education intervention. This commendable effort was associated with an overall 11% intervention-attributable decline in antibiotic prescription rates, although in a much smaller population than that studied by McCaig et al. This decline in antibiotic prescribing is within the expected range based on other intervention studies.^{12 - 15} Also, because Perz et al used the excess percentage change in antimicrobial prescription rates in Knox County before and after the intervention relative to control counties as the primary outcome of their trial, prescriptions were not linked to specific office visits or diagnoses, avoiding issues of diagnosis shifting and telephone prescribing.

Several limitations of the study by Perz et al should be considered. Only 4 counties in Tennessee were studied, and their nonrandom assignment increases the potential for bias. Some sources of bias could be identified (such as racial distribution) and controlled for in the analysis, but other unrecognized sources for bias may exist (eg, differences in media exposure regarding antibiotic use). The study used available data for health care services provided to a low-income population that may not be applicable to the general population. Because baseline prescribing rates were substantially higher in Knox County compared with control counties, even after stratifying for race, regression to the mean could have contributed to the observed reduction in prescription rates. Lastly, the International Classification of Diseases, Ninth Revision, Clinical Modification codes used for respiratory tract infections did not include the code for streptococcal sore throat (034.0), and cases of pharyngitis that might have been given this code would not have been included in the analysis.

Despite the apparent decline in antibiotic prescribing for children and adolescents described by McCaig et al and Perz et al, resistance among bacteria that cause respiratory tract infections continues to increase.¹⁶ Mathematical models of the epidemiology and population genetics of antibiotic treatment and resistance in open and closed communities have explored (1) the relationship between antibiotic consumption and the frequency of antibiotic resistance in bacterial populations, (2) methods of controlling increases in the development of bacterial resistance, (3) dissemination and persistence of antibiotic resistance in these settings, (4) the extent to which development of resistance can be controlled, and (5) the speed with which the effects of control measures might be realized.¹⁷ These models predict that in open communities, it would take years or even decades to see substantial reductions in the frequency of antibiotic resistance solely as a result of more prudent use of antibiotics.

To achieve more selective, appropriate use of antibiotics for pediatric patients, several key clinical points should be emphasized. First, physicians need to learn to differentiate by otoscopic examination acute suppurative otitis media (for which antibiotics are appropriate) and otitis media with effusion (for which antibiotics may be deferred).^{18 - 19} Second, for patients with pharyngitis, laboratory documentation by rapid antigen detection test or throat culture for group A streptococcus should be a standard of care before antibiotics are prescribed.^{20 - 21} Third, upper respiratory tract infections are viral, and unless associated fever lasts more than 5 days²² or purulent rhinitis (which can be caused by a viral, allergic, or bacterial etiology) lasts more than 10 days,²² it would be prudent and preferable to postpone initiation of antibiotic therapy. Fourth, cough is usually caused by viral illness, and unless associated fever lasts more than 5 days²³ or purulent sputum (which can be caused by a viral, allergic, or bacterial etiology) lasts more than 10 days²³ or there are physical examination and/or chest radiograph findings providing evidence of pulmonary consolidation, it would be prudent and preferable to postpone initiation of antibiotics. Fifth, rhinosinusitis does not occur in all patients who complain of purulent nasal discharge and cough. The diagnosis of rhinosinusitis generally should not be made until such symptoms persist beyond 10 days.²⁴ Finally, low dosages of antibiotics (as prescribed by physicians or taken by patients as self-medication from leftover supplies) promote bacterial resistance.²⁵ Higher adequate doses²⁶ for shorter time intervals²⁷ are preferred for most patients with bacterial respiratory tract infections treated with antibiotics in the ambulatory care setting.

In summary, the studies by McCaig et al and Perz et al provide hope that inappropriate and excessive use of antibiotics to treat children with respiratory tract infections is declining. To continue this trend, physicians need to be convinced that presumptive antibiotic use for viral respiratory tract infections does not minimize or prevent the development of secondary bacterial infections⁴ ; that patients receiving presumptive antibiotics have no difference in rate of return visits²⁸ ; that overuse of antibiotics promotes the development of bacterial resistance; that adverse effects of antibiotics can occur and may be serious; and that costs for unnecessary antibiotics should be avoided.⁴ Physicians want to do the right thing for their patients. Faced with an ill child or adolescent and worried parents, most often the right thing is reassurance, symptomatic therapy, and availability for follow-up—not antibiotics.