Objective. Although overuse of antibiotics in children has been well documented, relatively little information is known about provider and facility characteristics associated with this prescribing practice. This study was done to evaluate the differences in overuse of antibiotics among staff physicians and resident/interns (housestaff [HS]) who work in hospital-based outpatient clinics.
Methods. This cross-sectional study involved patient encounters in outpatient departments that were included in the US National Hospital Ambulatory Medical Care Survey database from 1995 to 2000. Encounters with patients who were aged <18 years and had a primary diagnosis suggestive of viral respiratory tract infection were evaluated. Patients with comorbid conditions that might justify antibiotic use were excluded.
Results. This study included 1952 patient encounters with a primary diagnosis suggestive of a viral infection and 33.2% of these patients receiving antibiotics. Overall, antibiotic use was significantly less among HS (19.5%) than staff physicians (36.4%; odds ratio [OR]: 0.44; 95% confidence interval [CI]: 0.33–0.59). This difference between HS (19.5%) and staff physicians (32.5%) persisted even within teaching hospitals (OR: 0.5; 95% CI: 0.4-0.7). Among staff physicians, antibiotic use was greater among those who work in nonteaching (39.6%) compared with teaching hospitals (32.5%; OR: 1.51; 95%: CI 1.15-1.98). Controlling for other patient and provider variables, antibiotic use occurred less among HS than among staff physicians in teaching hospitals (OR: 0.53; 95% CI: 0.38-0.75).
Conclusions. Antibiotic prescribing in the context of an outpatient visit for a diagnosis suggestive of a viral respiratory tract illness occurs more commonly among staff physicians than trainees and among staff physicians more commonly in nonteaching compared with teaching institutions.
Antibiotic resistance has become a global problem, with resistance among Streptococcus pneumoniae, among others, a major focus of concern.1,2 The association of antibiotic use and overuse with development of resistance3–6 and reversal of resistance patterns with decreased use has been previously demonstrated.7–10 Respiratory tract infections including otitis media are clearly the leading indication for outpatient antibiotic prescriptions.11 Numerous studies have shown that a substantial number of prescriptions for antibiotics are provided in cases in which antibiotic use has unproven benefit,12 such as for upper respiratory tract infections (URTIs), the common cold, bronchitis, and bronchiolitis.13–16
Although much has been written on patient characteristics and provider variables such as physician specialty and practice location, few have examined the influence of training status and teaching environment on antibiotic overuse.14,17–20 This study was done to evaluate the differences in potential overuse of antibiotics in hospital-based outpatient clinics among residents (housestaff [HS]) and staff physicians as well as among staff physicians who work in teaching versus nonteaching hospitals.
Study Design and Sample
We performed a cross-sectional study of patient encounters from outpatient departments included in the National Hospital Ambulatory Medical Care Survey (NHAMCS) from 1995 to 2000. The NHAMCS is a national probability sample survey conducted by National Center for Health Statistics. NHAMCS has 2 components: hospital outpatient departments and hospital emergency departments. To avoid the potential variability in patient and provider characteristics between an emergency department and an outpatient department setting, only the outpatient department data were used in the current study. Public use of National Ambulatory Medical Care Survey/NHAMCS data files continues to form the basis for numerous publications, including those related to prescribing practices.11–15,21,22 Details regarding the survey method used in the NHAMCS, including information on the survey instrument, data collection, processing, estimation, and reliability, are provided on the National Center for Health Statistics web site.23 The basic sampling unit for the NHAMCS is the patient visit or encounter. The survey uses a 4-stage probability design with samples of geographically defined areas, hospitals within these areas, clinics within hospitals, and patient visits within clinics. Statistics from the NHAMCS are derived by a multistage estimation procedure that has 3 basic components: (1) inflation by reciprocals of the sampling selection probabilities, (2) adjustment for nonresponse, and (3) a population weighting ratio adjustment. Hospital staff are instructed to complete patient record forms for a systematic random sample of patient visits during a randomly assigned 4-week reporting period. Information collected includes the reason(s) for visit, physicians' diagnosis, and medication(s) prescribed. The US Bureau of the Census acts as the field data collection agent for the NHAMCS. All medical and drug coding and keying operations are performed centrally by Constella Group and subject to quality control procedures. The keying error rate for nonmedical items is typically <1%. For items that require medical coding, discrepancy rates are generally <2%.
Inclusion criteria for this study were outpatient visits for patients who were <18 years of age and for whom the primary diagnosis was 1 of 4 common viral diagnoses: acute nasopharyngitis (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] code 460); acute URTI of multiple or unspecified sites (ICD-9-CM code 465); acute bronchitis and bronchiolitis (ICD-9-CM code 466); or bronchitis, not otherwise specified (ICD-9-CM code 490). Only visits by patients who were seen by a staff physician (autonomous visit not seen in conjunction with a resident/intern/nurse practitioners or physician assistants) or resident/intern were analyzed. We excluded visits in which a secondary diagnosis could have justified antibiotic use, such as streptococcal sore throat and scarlet fever (ICD-9-CM code 034), suppurative and unspecified otitis media (ICD-9-CM code 382), mastoiditis and related conditions (ICD-9-CM code 383), acute sinusitis (ICD-9-CM code 461), acute pharyngitis (ICD-9-CM code 462), acute tonsillitis (ICD-9-CM code 463), chronic sinusitis (ICD-9-CM code 473), pneumococcal or other bacterial pneumonia (ICD-9-CM code 481 and 482), chronic bronchitis (ICD-9-CM code 491), cystitis (ICD-9-CM code 595), and skin infections (ICD-9-CM codes 680-684, 686). In addition, we excluded visits in which a comorbidity such as asthma (ICD-9-CM code 493), cystic fibrosis (ICD-9-CM code 277), chronic airway obstruction (ICD-9-CM code 496), chronic cardiopulmonary disease (ICD-9-CM code 416.9), symptomatic HIV (ICD-9-CM code 042), or organ or tissue replaced by transplant (ICD-9-CM code V42) could have prompted an antibiotic prescription.
Oral or parenteral antibiotic use in a clinical setting suggestive of a viral respiratory tract infection as mentioned above was noted. Antibiotics were identified by searching for the National Drug Code Directory Drug Classes codes for penicillins, cephalosporins, macrolides, sulfonamides and trimethoprim, quinolones, tetracyclines, and other miscellaneous antibacterial agents.
Training Status and Hospital Type
We designated hospitals as “teaching” when any visit in the NHAMCS database from that hospital was to a trainee (intern or resident [HS]), whereas hospitals in which there were no patient visits to HS were designated “nonteaching.” This assignment as teaching or nonteaching was done before applying any inclusion or exclusion criteria by using data from all visits in the NHAMCS database from 1995 to 2000.
All analysis was done on unweighted data. The primary endpoint of the study was the prescription of an oral antibiotic for an eligible visit, and our primary exposures were trainee status and hospital teaching status, as defined above. Potential covariates included patient demographics and insurance information, hospital location, and clinic type. Temporal variables included year as well as whether the visit occurred after the publication of the guidelines on appropriate use of antibiotics in children in 1998 by the Centers for Disease Control and Prevention (CDC) and the American Academy of Pediatrics (AAP).24 Unadjusted odds ratios (ORs) and adjusted ORs were calculated using logistic regression analysis. To prevent the possibility that the prescribing practices in a hospital that contributed multiple patients to our study population could unduly influence our results, we adjusted all analysis for “clustering” of nonindependent observations within hospital.25 Because the NHAMCS provides a unique identification of individual hospitals, we used general estimating equations to account for nonindependence of observations within the same hospital. To explore further differences between antibiotic prescribing among staff physicians and HS, we conducted stratified analyses by selected characteristics. Data analysis was done using the statistical software SAS, version 8.1 (SAS, Cary, NC).
There were 174720 patient records in the NHAMCS database from 1995 to 2000, 45743 of which were for patients under the age of 18 years. The primary diagnosis of interest was recorded in 2839 visits; 513 of these were excluded because they had a secondary diagnosis that could have prompted antibiotic use. A total of 1952 visits of patients who were seen by a resident or staff physician met the study inclusion and exclusion criteria. Of these, 717 (36.7%) were seen by a staff physician in a teaching hospital, 866 (44.4%) by a staff physician in a nonteaching hospital, and 369 (18.9%) by residents or interns (HS). The most common primary diagnosis in the study population was acute URTI of multiple or unspecified sites (77.2%), followed by bronchitis, not otherwise specified (12.5%), acute bronchitis and bronchiolitis (7.1%), and acute nasopharyngitis (3.2%). Of 1952 visits included in this study, 1396 (71.5%) were for 1 of 5 symptoms: fever, nasal congestion, cough, head cold/URTI, and throat soreness. These 5 presenting symptoms combined as well as fever alone were similar across study visits to all 3 groups of physicians (P = .18 and 0.33, respectively).
Of the 1952 patients with a diagnosis suggestive of a viral infection, 648 (33.2%) had antibiotics prescribed. Penicillins (includes aminopenicillins), macrolides, and cephalosporins constituted 57%, 24%, and 12% of all oral antibiotic prescriptions, respectively. On the basis of the primary diagnosis, antibiotic use was seen most in patients who had a diagnosis of bronchitis, not otherwise specified (69.5%), followed by acute bronchitis and bronchiolitis (47.8%), acute nasopharyngitis (39.7%), and acute URTI of multiple or unspecified sites (25.7%). Given the disproportionately higher antibiotic use for patients who had a diagnosis of bronchitis, not otherwise specified, for the purpose of further analysis, we elected to compare this category with the other 3 diagnostic categories.
Antibiotic use in children and adolescents with a diagnosis suggestive of a viral respiratory tract infection was significantly less among HS (19.5%) than staff physicians (36.4%; OR: 0.44; 95% confidence interval [CI]: 0.33–0.59; P < .0001). This difference between HS (19.5%) and staff physicians (32.5%) persisted even within teaching hospitals (OR: 0.5; 95% CI: 0.4–0.7; P < .0001; Fig 1). Among staff physicians, antibiotic use was greater among those who worked in nonteaching (39.6%) compared with teaching hospitals (32.5%; OR: 1.51; 95% CI: 1.15–1.98; P = .003).
Other factors associated with antibiotic use were first examined using univariate analysis (unadjusted OR; Table 1). Those associated with an increased or decreased odds of antibiotic use relative to the referent in each category are shown in Table 1. The effect of time (secular trends) on antibiotic prescribing behavior was analyzed in 2 ways. The difference in antibiotic use before and after the publication of the CDC/AAP guidelines in 199824 regarding rationale antibiotic use was examined (Table 1). In addition, the secular trends in antibiotic use among the 3 groups of clinical providers were analyzed. Although a linear association was significant for antibiotic use and year of visit between 1995 and 2000 among staff physicians in nonteaching (P = .002) and teaching (P = .009) hospitals, a similar association was not seen for the HS (P = .6; Fig 2). This figure also demonstrates that the difference in prescribing practice between HS and staff physicians predates the publication of the CDC/AAP guidelines.
The difference in antibiotic use between the staff physicians in teaching hospitals and staff physicians in nonteaching hospitals and HS was then examined in the context of the other variables that were potentially associated with inappropriate antibiotic use (adjusted OR; Table 1). Controlling for other factors associated with antibiotic use, the odds of a child or adolescent being prescribed antibiotics for a diagnosis suggestive of a viral infection in a teaching hospital by HS was significantly less than by a staff physician (OR: 0.53; 95% CI: 0.38–0.75) in the same setting. Conversely, even in the multivariable analysis, the odds of being prescribed an antibiotic were more by a staff physician in a nonteaching hospital compared with those in a teaching hospital (OR: 1.40; 95% CI: 1.05–1.87). Other factors that were significantly associated with increased antibiotic use included receiving a diagnosis of bronchitis, otherwise unspecified; being seen in a nonpediatric clinic (predominantly general medicine); and being seen before the publication of the CDC/AAP guidelines. Practice location in the South or West was associated with less antibiotic use when compared with those seen in the Northeast. Note that the proportion of teaching hospitals that contributed data to this study from the Northeast, Midwest, South, and West was 57.2%, 69.9%, 41.5%, and 52.9%, respectively.
The odds of antibiotic use when seen by HS when compared with any staff physician (teaching or nonteaching) was also examined within each strata of the other variables associated with antibiotic use while including the remaining covariates in the model (Fig 3). Overall, the difference in antibiotic use between the 2 training cohorts remained significant and generally consistent within all strata (including before and after publication of CDC/AAP guidelines) except for being seen in a nonmetropolitan statistical area or being seen in the West.
This study demonstrates that potential overuse of antibiotics in children is related to both provider and facility characteristics. Antibiotic prescribing in the context of an outpatient visit for a diagnosis suggestive of a viral respiratory tract illness occurs more commonly among staff physicians than trainees and among staff physicians more commonly in nonteaching compared with teaching institutions. These differences persisted when controlling for potential confounders, including patient and visit demographics, and the difference between staff physicians and trainees was consistent across most patient- and visit-related strata.
There are several possible explanations for the lower use of antibiotics among trainees compared with staff physicians. First, although the reasons for a visit are similar, patients who see staff physicians might be sicker than those who are seen by trainees. However, the distribution of symptoms, specifically fever, as the major reason for visit was similar among all groups of physicians. Second, although we controlled for insurance status, there may be differing expectations among parents with staff physicians compared with trainees. Third, one can speculate that trainees may feel sheltered in an academic environment and perceive less medicolegal risk when withholding antibiotics in specific clinical situations compared with practicing clinicians in hospital-based outpatient departments. Finally, our findings may represent a “cohort effect”; trainees may be more familiar with recently administered guidelines and may be more comfortable with antibiotic restraint than providers who trained and practiced before the dissemination of these guidelines. We have observed a similar cohort effect in the choice of hypnotic agents for older adults between interns and attending physicians.26
Our findings are similar to those of Mincey et al,19 who noted that within a teaching facility, contrary to their expectations, junior residents were more likely to prescribe an appropriate antibiotic for sinusitis than senior residents and staff physicians and that with increasing level of training, residents' prescribing practices more closely resembled those of the supervising physicians. A study of antibiotic prescribing practices for patients with colds, URTIs, and bronchitis using emergency department records in the 1996 NHAMCS database found a similar difference in antibiotic use between HS and staff physicians as seen in the current study.14 Mainous et al,18 using information from the Kentucky Medicaid database, reported that the “high prescribers” of antibiotics for children with URTIs were significantly more years from medical school graduation (27 vs 19 years) than “low prescribers.”
The use of antibiotics was higher among staff physicians who practice in nonteaching compared with teaching institutions. A number of studies have compared the differences in quality of care between teaching and nonteaching hospitals, with some of the more rigorous studies indicating better care in major teaching hospitals for conditions such as myocardial infarction.27 In a study reported by Steinke et al,20 nontraining practices in Tayside, UK, were in general found to prescribe significantly more antibiotics as well as a higher proportion of broad-spectrum penicillins, a higher proportion of newer antibiotics, and a greater number of different antibiotics per doctor compared with training practices.
In addition, there may be differential access to recent literature among trainees (HS) who function in an academic environment versus staff physicians who are not engaged in formal teaching and may have variable access and commitment to continuing medical education. This may be one of the explanations for the much wider difference in antibiotic use among nonteaching staff physicians and HS. It was shown previously that a physician's performance on certification examinations deteriorates over time.28 Also, Tamblyn et al29 showed that scores achieved on certification and licensure examinations that are taken at the end of medical school in Quebec show a sustained relationship over 4 to 7 years with indices of preventive care and acute and chronic disease management in primary care practice. Furthermore, over the past decade, numerous national organizations have actively campaigned toward reducing antibiotic overuse, and the greatest impact of the message may have been on the most impressionable part of the clinical workforce, namely the HS.24,30 A number of factors have been identified as barriers that physicians face in adhering to clinical practice guidelines.31
As has been reported by others,21,22,32 the encouraging overall trend in decreased antimicrobial use in the outpatient setting was also seen in this study. Despite using the CDC/AAP guidelines regarding appropriate use of antibiotics in pediatric URTIs as the time point around which secular trends in antibiotic prescribing were examined in this study, we acknowledge that these guidelines are one among many measures taken over the past decade to decrease antibiotic overuse.33 Hence, the causal relationship between publication of these guidelines and decreased antibiotic use, although implied, is not exclusive. Nambiar et al17 reported the results of a survey that assesses awareness and compliance of pediatric residents with the CDC and AAP guidelines regarding “principles of judicious use of antibiotics in URI.” On the basis of responses to the survey questionnaire, including clinical vignettes, they found a trend toward improving awareness of the guidelines and decreased antibiotic use with increasing years of training.
Several limitations of this study warrant additional discussion. First, it should be noted that because many of the HS visits may have been partially or completely supervised by staff physicians, the antibiotic prescribing behavior of the HS seen in this study reflects in part the prescribing style of the supervisors. Despite accounting for the teaching status of the hospital where the clinic was located, we cannot totally rule out the contribution of direct or indirect supervision in the prescribing style of the HS. Second, with the NHAMCS database, we cannot confirm the accuracy of the diagnosis reported by the participating providers to identify visits for presumed viral respiratory tract infection. However, similar to previous investigators who have used encounter data in the National Ambulatory Medical Care Survey database to evaluate inappropriate antibiotic use,12,13 we restricted our study to include ICD-9 codes (listed for the primary diagnosis) for conditions for which antibiotic use is not recommended.24 Furthermore, we excluded visits for related chronic conditions (eg, chronic bronchitis) because appropriate use of antibiotics in these cases has not been established. That said, in the absence of medical history, we cannot account for differences in decision making influenced by the medical history of a patient such as a child with multiple previous episodes of otitis media or the premature infant with multiple postnatal problems. We also cannot be sure how many excluded visits for conditions for which antibiotic use is considered acceptable (eg, otitis media, sinusitis) were cases of overdiagnosis or miscoding to “justify” the antibiotic prescription.34 Aside from the possibility that such a practice was far more prevalent among trainees than staff physicians, such coding practices would underestimate rather than overestimate the antibiotic use and difference in antibiotic use observed in this study. One way, albeit not perfect, of addressing this possibility is to examine among the different clinical providers the proportion of clinic visits of patients who were <18 years of age and for whom a diagnosis of otitis media or sinusitis was made. This was seen in 8%, 5.2%, and 4.3% of overall pediatric visits to staff physicians (nonteaching), staff physicians (teaching), and HS, respectively, and if anything was less among HS than staff physicians. Third, it is also possible that despite the stringent exclusion criteria, some comorbid conditions that justify use of an antibiotic were missed. Fourth, differences in antibiotic prescribing may have been related to differences in severity of illness. Although we could not assess the same, we did look for differences in presenting symptoms, such as fever, in the visits to the 3 groups of providers and found no difference. Finally, we used the record of any HS visit from a hospital in the entire NHAMCS database as a surrogate measure to assign teaching status to that hospital. Although this definitely helped in teasing out the role of the teaching environment, it is an indirect measure with the potential of having a less-than-ideal sensitivity, ie, there may have been HS in the hospital but they did not contribute to the sample used for NHAMCS survey resulting in that hospital being designated “nonteaching.”
In conclusion, this study suggests that prescribing practices of trainees during patient visits suggestive of a viral respiratory tract infection are more consistent with the existing antibiotic prescribing guidelines than those of staff physicians in hospital-based outpatient departments. Although we are hopeful that this cohort of trainees will continue this prescribing behavior once they enter practice, it is also conceivable that practice patterns will deteriorate as they exit the teaching environment. This study provides additional evidence that additional, focused measures to educate and remind the staff physicians in hospital-based outpatient clinics, including continuing medical education,35–37 educational outreach intervention,38 and point-of-care interventions,39 would further complement the gains made in decreased antibiotic use over the past decade.
This study was supported by National Cancer Institute Cancer Center Support CORE Grant P30 CA 21765 and by the American Lebanese Syrian Associated Charities. Dr Hare and Dr Shorr are supported by the 21st Century Scholars Program, funded by LHS Inc.
We thank Grant W. Somes, PhD, Department of Preventive Medicine, University of Tennessee Health Science Center (Memphis, TN), for assisting in the analysis and interpretation of data in this study.
- Accepted July 27, 2004.
- Reprint requests to (A.H.G.) Department of Infectious Diseases, St Jude Children's Research Hospital, 332 N Lauderdale St, Memphis, TN 38105-2794. E-mail:
No conflict of interest declared.
- ↵Reichler MR, Allphin AA, Breiman RF, et al. The spread of multiply resistant Streptococcus pneumoniae at a day care center in Ohio. J Infect Dis.1992;166 :1346– 1353
- ↵Austin DJ, Kristinsson KG, Anderson RM. The relationship between the volume of antimicrobial consumption in human communities and the frequency of resistance. Proc Natl Acad Sci U S A.1999;96 :1152– 1156
- ↵Brook I, Gober AE. Prophylaxis with amoxicillin or sulfisoxazole for otitis media: effect on the recovery of penicillin-resistant bacteria from children. Clin Infect Dis.1996;22 :143– 145
- Nowak R. Hungary sees an improvement in penicillin resistance. Science.1994;264 :364
- ↵Gonzales R, Malone DC, Maselli JH, Sande MA. Excessive antibiotic use for acute respiratory infections in the United States. Clin Infect Dis.2001;33 :757– 762
- ↵Steinke DT, Bain DJ, MacDonald TM, Davey PG. Practice factors that influence antibiotic prescribing in general practice in Tayside. J Antimicrob Chemother.2000;46 :509– 512
- ↵National Center For Health Statistics—Ambulatory Health Care Data. Available at: www.cdc.gov/nchs/about/major/ahcd/ahcd1.htm. Accessed July 12, 2004
- ↵Dowell SF, Marcy SM, Phillips WR, Gerber MA, Schwartz B. Principles of judicious use of antimicrobial agents for pediatric upper respiratory tract infections. Pediatrics.1998;101(suppl) :163– 165
- ↵Zwar N, Henderson J, Britt H, McGeechan K, Yeo G. Influencing antibiotic prescribing by prescriber feedback and management guidelines: a 5-year follow-up. Fam Pract.2002;19 :12– 17
- ↵Finkelstein JA, Davis RL, Dowell SF, et al. Reducing antibiotic use in children: a randomized trial in 12 practices. Pediatrics.2001;108 :1– 7
- ↵Christakis DA, Zimmerman FJ, Wright JA, Garrison MM, Rivara FP, Davis RL. A randomized controlled trial of point-of-care evidence to improve the antibiotic prescribing practices for otitis media in children. Pediatrics.2001;107(2) . Available at: www.pediatrics.org/cgi/content/full/107/2/e15
- Copyright © 2005 by the American Academy of Pediatrics