National Trends in the Use of Antibiotics by Primary Care Physicians for Adult Patients With Cough FREE

OVER THE past few years, substantial concern has accompanied the documented rise in antibiotic resistance among common respiratory pathogens.^1- 5 Initial work primarily focused on the prevalence of resistance among hospitalized patients but more recently has expanded to identify antibiotic-resistant pathogens in several ambulatory care settings.^6- 9 Additional evidence has linked prior antibiotic use with the identification of antibiotic-resistant pathogens.^6,9- 10 As a result, several authors^11- 16 have emphasized the importance of more appropriately regulating the use of antibiotics for common outpatient illnesses as a means to combat the problem of rising antimicrobial resistance. Whether these efforts at improving physician knowledge about the risks of excessive antibiotic use will ultimately lead to any reductions in antibiotic use is unknown. However, an understanding of current trends in antibiotic prescribing will be useful for interpreting any future trends in antibiotic use.

This study focuses on the use of antibiotics for patients presenting with cough rather than patients diagnosed as having specific acute respiratory diseases. In the primary care setting, a chief complaint of cough may indicate a wide array of infectious and noninfectious causes. Even among the infectious causes there are diagnoses for which antibiotics are appropriate (eg, pneumonia), diagnoses for which antibiotics are inappropriate (eg, common cold), and a diagnosis for which the efficacy of antibiotics has not been clearly established despite a number of randomized clinical trials (ie, acute bronchitis).¹⁷ However, the choice of diagnoses assigned to these patients is imprecise and may well be driven by the choice of therapies, an important source of confounding in the measurement rates of antibiotic use. We were interested in studying changes in the use of antibiotics in the management of a population of patients, rather than changes in the use of antibiotics.

Prior work using the National Ambulatory Medical Care Survey (NAMCS)¹⁸ suggested that the use of certain classes of broad spectrum antibiotics was rising in the ambulatory care setting. Additional work¹⁹ based on this survey demonstrated that as recently as 1992, most patients diagnosed as having colds, upper respiratory tract infections, and bronchitis were treated with antibiotics. We hypothesized that this pattern of rising antibiotic use would be even more apparent in the management of patients with cough, despite the recent increase in awareness regarding the rise in antibiotic resistance. The specific aims of this study were (1) to examine national trends in the use of antibiotics by primary care physicians in the care of adult patients with cough and (2) to identify factors that may influence antibiotic use for these patients.

Data for this study were from the NAMCS for 1980, 1981, 1985, and 1989-1994.²⁰ Conducted in these years by the National Center for Health Statistics, Hyattsville, Md, these surveys provide an ongoing assessment of the practices of US office-based physicians, including both urgent and scheduled visits. We began with 1980 because it was the first survey year in which drug prescribing information was included at each sampled visit. We included all subsequent NAMCS survey years through 1994, the most recent year available at the time of this study. The sampling process makes use of the American Medical Association and American Osteopathic Association master lists of all US primary care physicians. A sample of physicians from these lists was then selected by random, stratified sampling by geographic area and specialty. Among eligible physicians, annual participation rates varied from 72% (1992) to 78% (1981). For each participating physician in each year, patient visits during a randomly selected week were sampled systematically.

For each selected patient visit, physicians completed encounter forms detailing the specific clinical services provided during the visit, as well as patient demographics, reasons for visit, and diagnoses. The first listed reason for visit and first listed diagnosis represent the principal reason for visit and associated principal diagnosis, respectively. We chose to focus on cough as the principal reason for visit because of its explicit linkage to the principal diagnosis, thus permitting a better understanding of the clinical indications for antibiotic use. In fact, secondary reasons for visits were listed in only 37% of all visits and a third reason for visit was listed in only 11% of all visits to primary care physicians. Continuing and newly ordered over-the-counter and prescribed medications also were listed for each visit. For each visit record, the National Center for Health Statistics provided a visit weight calculated from the physician and visit sampling rates, adjusted for nonresponse. Statistical aggregation using these visit weights allows extrapolation to national patterns of practice.

We restricted this study to visits by adults (aged ≥18 years) to primary care physicians, defined as physicians identified in 1 of the following 3 American Medical Association categories: internal medicine, family practice, or general practice. This portion of the NAMCS database includes from 5627 (1993) to 19977 (1985) annual outpatient visits by adults to primary care physicians.

Visits by adults to primary care physicians for cough were identified based on the coding of patient reason for visit. Those visits in which cough was coded as the first (primary) reason for visit were identified as primary care visits for cough. A total of 3416 visits by patients with a chief complaint of cough were identified: 431 in 1980; 335 in 1981; 555 in 1985; 378 in 1989; 487 in 1990; 412 in 1991; 234 in 1992; 264 in 1993; and 320 in 1994.

Patients receiving antibiotic therapy were identified by the coding of generic or proprietary names for all antibiotics in the following classes: penicillins, cephalosporins, macrolides, tetracyclines, quinolones, and trimethoprim-sulfamethoxazole drugs. We did not exclude parenteral forms of antibiotics, although they contributed only a small proportion of all antibiotics prescribed since only outpatient visits are included in the NAMCS.

Primary diagnoses for patients with visits for cough were identified based on the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)²¹ diagnostic code listed as the first diagnostic code for each visit. Based on the frequency of ICD-9-CM codes, we identified the following primary diagnoses for visits by patients with cough: bronchitis, acute or not otherwise specified as acute or chronic (ICD-9-CM codes 466 and 490), influenza (code 487), acute upper respiratory tract infection (code 465), acute nasopharyngitis (common cold) (code 460), acute and chronic sinusitis (codes 461 and 473), bacterial pneumonia (codes 481-483 and 485-486), asthma (code 493), and acute pharyngitis (code 462). We coded the cough visit as an initial visit for the problem or a follow-up visit based on whether the physician identified the primary reason for visit as a new problem (or a first visit with the patient) or as a problem for which they had previously seen the patient.

A number of patient characteristics reported in the NAMCS were evaluated to determine their association with antibiotic use. Particular characteristics that were tested included patient age (categorized as 18-44 years, 45-64 years, and ≥65 years), sex, and race and ethnicity (categorized as white, non-Hispanic vs other).

National estimates of the number of visits by adults to primary care physicians for cough and the number of antibiotics recorded at those visits were calculated based on the sampling weights provided by the National Center for Health Statistics. For each estimate, a 95% confidence interval was calculated based on the relative SEs provided in the documentation for each survey year. These relative SEs are provided for visit estimates and drug estimates for individual physician specialties. Because our study of primary care physicians merged 3 specialties (internal medicine, family practice, and general practice) in each year, we chose the largest relative SE associated with one of these specialties to determine the confidence intervals around the aggregated sample. As recommended by the National Center for Health Statistics, significance of linear trend for the annual frequency of visits for cough and the annual frequency of antibiotic drug prescriptions at these visits was based on a 2-sided t test for trend, incorporating the relative SEs for each year in the calculation.²²

National patterns in the rate of patients with cough receiving antibiotics were assessed by cross-tabulation of medication information by year of visit, weighted by the National Center for Health Statistics visit-sampling weights. Rates of antibiotic use were expressed as the proportion of visits for cough in which any antibiotic was listed. Both unadjusted and adjusted tests for significance of trends over the study years were based on logistic regression analysis with the natural logarithm of the odds of antibiotic use in patients with cough as the dependent variable. These analyses were performed with SAS statistical software (Release 6.11, SAS Institute Inc, Cary, NC) using modified sample weight to reflect the actual-sampling probabilities of each visit and the original sample size of the final study group. In the logistic regression analysis, calendar year was entered as a continuous variable to test for significance of linear trend.

All analyses used the patient visit as the unit of analysis and did not adjust for the clustering effect within individual sampling units (eg, physician practices) because unique physician identifiers were not available for this study for years prior to 1991. However, using general estimating equations and the physician identifiers for years 1991-1994 demonstrated that adjustment for the clustering effect in those years had minimal effect on the parameter estimates and increased the SEs for each predictor by 20% to 30%. Moreover, the mean number of office visits by patients with cough to the sampled population of primary care physicians was only 2.2 (median, 2).

The frequency of visits to primary care physicians by adults with a chief complaint of cough rose steadily from 1980 through 1994. In the United States, there were an estimated 5.5 million office visits for cough in 1980 rising to 10.2 million office visits in 1994 (P<.001 for trend) (Table 1). Over this period, the number of total office visits to primary care physicians remained relatively constant (225 million in 1980 and 219 million in 1994). As a result, the proportion of visits to primary care physicians by adults with cough nearly doubled from 2.5% of all visits in 1980 to 4.7% in 1994. Of note, cough was listed as a nonprimary reason for visit much less commonly and the frequency of listing cough as a secondary or tertiary reason for visit did not demonstrate a similar increase in frequency during the study period (data not shown). During this study period, the frequency of other primary reasons for visits did not demonstrate the same pattern of increase. For example, sore throat represented 2.6% of primary care visits in 1980 and 2.3% of visits in 1994. Head cold represented 2.0% of visits in 1980 and 1.6% of visits in 1994, and a chief complaint of lung congestion or chest cold represented 0.3% of visits in 1980 and 0.4% of visits in 1994. Finally, 58% of visits with a chief complaint of cough included additional reasons for visit. The leading additional reasons for visits were nasal congestion, fever, and sore throat, representing 12.5%, 8.5%, and 7.4% of all visits, respectively.

Most patients who presented with cough were women (60%), younger than 65 years (73%), and white, non-Hispanic (82%). Fifty-two percent of the patients had been previously seen by the surveyed physician for the same problem. The race and sex distribution of patients did not change significantly during the study years. In contrast, there was a slight increase in both the proportion of visits by patients 65 years or older (23% in 1980 to 29% in 1994; P =.03 for trend) and the proportion of visits by patients previously seen for cough (42% in 1980 to 51% in 1994; P=.004 for trend).

The relative distribution of primary diagnoses for patients with a chief complaint of cough did not change much during the study years (Table 2). In particular, bronchitis, including acute bronchitis and bronchitis not specified as acute or chronic, remained the leading diagnosis in each study year (average, 42% of visits). Other leading diagnoses included upper respiratory tract infection, asthma, sinusitis, and pneumonia. Fifty-nine percent of the visits for cough had only a single diagnosis listed.

The number of antibiotics prescribed as a result of visits for cough to primary care physicians increased over the survey years from 1980-1994, in parallel to the increase in frequency of visits for cough (Table 1). The lowest annual number of antibiotics was prescribed for cough in 1981 (5,128,000 annual prescriptions) and the highest annual number of antibiotics was prescribed in 1992 (11,440,000 annual prescriptions).

The average rate of antibiotic use for patients with a chief complaint of cough was 66% of office visits during the entire study period. Although the measured rate of antibiotic use showed considerable year-to-year variation, the use of any antibiotic for patients with cough demonstrated a statistically significant increase during the study period (P=.002). The 3 highest rates of use were in 1991, 1992, and 1994 (70% of visits) compared with the 3 lowest rates of use in 1980 (59%), 1985 (61%), and 1993 (63%) (Table 3).

In every survey year, physicians prescribed antibiotics for younger patients more often than for older patients. For example, in the most recent year (1994) physicians prescribed antibiotics for cough 74% of the time for patients aged 18 to 44 years, 72% of the time for patients aged 45 to 64 years, and 63% of the time for patients 65 years or older. Similarly, white, non-Hispanic patients were more likely to receive antibiotics compared with all other patients in each study year. There was no difference in the rate of antibiotic use for male and female adult patients (Table 3).

We next considered the potential influence of physician diagnosis on the use of antibiotics for these patients. Overall, patients with a diagnosis of bronchitis were prescribed antibiotics at 81% of office visits, which was more than patients with upper respiratory tract infections and all other diagnoses combined, who were prescribed antibiotics at 68% and 51% of office visits, respectively. In addition, patients who were seen for an initial visit for cough were more likely in every study year to receive an antibiotic compared with patients with cough who had been seen previously by the physician for this problem (Table 3).

Finally, we considered the independent effects of each of these predictors of antibiotic use in a multivariate logistic regression analysis model. Younger patients, white patients, and patients diagnosed as having bronchitis were significantly more likely to receive antibiotics for their cough (Table 4). In addition, the increased rate of antibiotic prescribing for patients seen during an initial visit for cough compared with patients previously seen for this problem remained significant in the multivariate analysis (odds ratio, 1.5; 95% confidence interval, 1.3-1.8). After adjusting for these clinical and demographic variables, the increasing trend of antibiotic use during the survey years remained statistically significant, although the adjusted size of the increase itself was modest (odds ratio, 1.03 for antibiotic use for each unit increase in year; 95% confidence interval, 1.01-1.04). Adjustment for secondary symptoms such as fever and sore throat had only minimal effects on the calculated odds ratios for the factors listed in the table (data not shown).

The rate of visits by adult patients with cough to primary care physicians and the rate of antibiotic use for these patients increased from 1980 through 1994. As a result, the absolute number of antibiotic prescriptions for patients with cough increased by 125% from 1980 through 1994. In addition, during this period, younger patients and white, non-Hispanic patients were more likely to receive antibiotics for cough, even after adjusting for the clinical diagnoses listed by the physician.

A shift in the frequency of office visits by patients with a chief complaint of cough may be multifactorial in origin, including both actual changes in the incidence of the underlying diseases in the population and changes in the thresholds for patients to seek medical care for these illnesses. In support of the former explanation, results from the National Health Interview Surveys from 1982 and 1993 indicate that patients are reporting increasing rates of acute respiratory conditions during this study period.^23- 24

While the absolute increase in the rate of antibiotic use during the study period is modest, the corresponding absence of any evidence that antibiotic use is falling is significant. Although there have been increasing efforts to call attention to the rise in antibiotic resistance and the role of antibiotic use in this problem, most of these efforts have occurred only recently. Whether these efforts at increasing awareness alone, or in conjunction with local or nationwide programs to reduce inappropriate antibiotic use, will actually lead to measurable reductions in antibiotic use remains to be seen. Indeed, considerable barriers including patient preferences and market advertising may continue to prevent significant reductions in antibiotic use.

The association of certain patient characteristics with the increased use of antibiotics by primary care physicians suggests that there may be important nonclinical barriers to promoting more judicious use of antibiotics for patients with cough. Alternatively, some demographic variables, such as age, may be markers for important clinical variables that more appropriately determine which patients with cough will benefit from antibiotic therapy. In addition, patient expectations and physician perception of those expectations may be additional important determinants of antibiotic use,²⁵ although these factors cannot be addressed in this study.

The findings of this study are largely in agreement with earlier reports, based on the NAMCS database, which noted significant increases in the rate of antibiotic use for adults with sinusitis from 1980 to 1992,¹⁸ and high rates of antibiotic use for adult patients with upper respiratory tract infections and bronchitis in 1992.¹⁹ Our study extends these observations to adult patients with cough illnesses, noting an overall increase in antibiotic prescribing by primary care physicians through 1994. Similar high rates of antibiotic use for respiratory illnesses also were noted in a study using a cross-sectional sample of Kentucky Medicaid claims that indicated that 60% of patients diagnosed as having common cold and 75% patients diagnosed as having acute bronchitis subsequently filled prescriptions for antibiotic medications.^26- 27

One limitation of this study is that the rates of antibiotic use are based on physician recording of new or continuing medications at the time of the visit, not whether the antibiotic prescriptions were ever filled or taken by the patients. Rates of antibiotic use would be underestimated when physicians failed to record antibiotic prescriptions given to patients during sampled visits, while rates would be overestimated when patients failed to fill or take antibiotic medications prescribed by the physician. However, as noted earlier, studies using Medicaid drug claims have noted similarly high levels of antibiotic use for patients with acute bronchitis or a common cold.^26- 27

Finally, the NAMCS database does not provide the clinical data necessary to assess the appropriateness of each antibiotic associated with a cough visit. For example, it is difficult to clearly identify patients with chronic cough illnesses (particularly chronic bronchitis) who might be appropriately treated with antibiotics. Moreover, we believe that there is still a strong need for future studies to provide better information on the appropriate indications for antibiotic use for adult patients with cough.

In conclusion, the rate of antibiotic use by primary care physicians for adult patients with cough increased from 1980 through 1994, a period during which there was substantially increased detection of drug-resistant respiratory pathogens. As programs to decrease inappropriate use of antibiotics are initiated, continued monitoring of the rate of antibiotic use will be necessary to detect if these efforts have been successful. In addition, future efforts to promote judicious antibiotic use should consider the importance of nonclinical factors in influencing current patterns of antibiotic use, and design interventions that address these barriers.

Accepted for publication December 2, 1997.

Dr Metlay was supported by a General Medicine Research Fellowship, National Research Service Award, National Institutes of Health, Bethesda, Md, grant 5T32PE11001-08. Dr Stafford is supported by a Mentored Clinical Scientist Development Award from the National Heart, Lung, and Blood Institute, Bethesda.

Presented in part at the 1997 Annual Meeting of the Society of General Internal Medicine, Washington, DC, May 2, 1997.

Corresponding author: Joshua P. Metlay, MD, PhD, Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Room 712 Blockley Hall, 423 Guardian Dr, Philadelphia, PA 19104 (e-mail: jmetlay@cceb.med.upenn.edu).