Linder JA, Stafford RS. Antibiotic Treatment of Adults With Sore Throat by Community Primary Care PhysiciansA National Survey, 1989-1999. JAMA. 2001;286(10):1181-1186. doi:10.1001/jama.286.10.1181
Author Affiliations: General Medicine Division, Massachusetts General Hospital, Boston (Dr Linder); and Stanford Center for Research in Disease Prevention, Palo Alto, Calif (Dr Stafford).
Context Most sore throats are due to viral upper respiratory tract infections.
Group A β-hemolytic streptococci (GABHS), the only common cause of sore
throat warranting antibiotics, is cultured in 5% to 17% of adults with sore
throat. The frequency of antibiotic use for pharyngitis has greatly exceeded
the prevalence of GABHS, but less is known about specific classes of antibiotics
used. Only penicillin and erythromycin are recommended as first-line antibiotics
Objectives To measure trends in antibiotic use for adults with sore throat and
to determine predictors of antibiotic use and nonrecommended antibiotic use.
Design, Setting, and Subjects Retrospective analysis of 2244 visits to primary care physicians in
office-based practices in the National Ambulatory Medical Care Survey, 1989-1999,
by adults with a chief complaint of sore throat.
Main Outcome Measures Treatment with antibiotics and treatment with nonrecommended antibiotics,
extrapolated to US annual national rates.
Results There were an estimated 6.7 million annual visits in the United States
by adults with sore throat between 1989 and 1999. Antibiotics were used in
73% (95% confidence interval [CI], 70%-76%) of visits. Patients treated with
antibiotics were given nonrecommended antibiotics in 68% (95% CI, 64%-72%)
of visits. From 1989 to 1999, there was a significant decrease in use of penicillin
and erythromycin and an increase in use of nonrecommended antibiotics, especially
extended-spectrum macrolides and extended-spectrum fluoroquinolones (P<.001 for all trends). In multivariable modeling, increasing
patient age (odds ratio [OR], 0.86 per decade; 95% CI, 0.79-0.94) and general
practice specialty (OR, 1.54 compared with family practice specialty; 95%
CI, 1.10-2.14) were independent predictors of antibiotic use. Among patients
receiving antibiotics, nonrecommended antibiotic use became more frequent
over time (OR, 1.17 per year; 95% CI, 1.11-1.24).
Conclusions More than half of adults are treated with antibiotics for sore throat
by community primary care physicians. Use of nonrecommended, more expensive,
broader-spectrum antibiotics is frequent.
Sore throat accounts for 2.1% of ambulatory visits in the United States
and is the second most common symptomatic reason for seeking care, after cough.1 Most sore throats are due to upper respiratory tract
viruses such as rhinovirus, coronavirus, and adenovirus.2
The main bacterial cause of sore throat is group A β-hemolytic streptococci
(GABHS), which is cultured in 15% to 36% of children and 5% to 17% of adults
with sore throat.3- 9
Group A β-hemolytic streptococci is the only common cause of sore throat
warranting antibiotic treatment.10
Ideally, the proportion of patients with sore throat receiving an antibiotic
would approximate or perhaps slightly exceed the prevalence of GABHS in those
with sore throat. However, 76% of adults and 71% of children diagnosed with
pharyngitis in 1992 were treated with an antibiotic.11,12
Therefore, pharyngitis represents 5.6 million annual antibiotic prescriptions
and is responsible for 9% of all antibiotics used in adults, over half of
which are likely unnecessary.11
Compared with the frequency of antibiotic prescription, less is known
about the classes of antibiotics prescribed for patients with sore throat.
The Infectious Diseases Society of America recommends using penicillin or
erythromycin (for those who are penicillin allergic) as the first-line agent
for patients with sore throat due to GABHS.13
These recommendations were recently reiterated in the guidelines for the treatment
of adults with acute pharyngitis issued by the Centers for Disease Control
and Prevention and the American College of Physicians—American Society
of Internal Medicine.14,15 Penicillin
and erythromycin are recommended because of proven efficacy, narrow spectrum,
safety, and low cost.
We used the National Ambulatory Medical Care Survey (NAMCS) to determine
if antibiotic prescribing practices for adults with sore throat have changed
over time. We tracked changes in the use of different antibiotic classes for
sore throat by community primary care physicians from 1989 to 1999. In addition,
we evaluated predictors of patients receiving an antibiotic for sore throat
and evaluated predictors of receiving nonrecommended antibiotics for sore
The NAMCS has been administered by the National Center for Health Statistics
annually since 1989. The NAMCS collects information on patient visits to nonfederally
funded, office-based physician practices throughout the United States and
has a multistage probability design in which sampling is based on geographic
location, physician specialty, and visits within individual physician practices.
Each visit is weighted by the National Center for Health Statistics to account
for location, specialty, and practice nonparticipation. These weights allow
extrapolation to national figures for all aspects of the survey.
The NAMCS has collected information on 355 354 patient visits between
1989 and 1999. The participation rate of physician practices in the NAMCS
has declined over the last 11 years from 74% in 1989 to 63% in 1999.16,17
At each visit, patient, physician, and clinical information is collected.
Patient information includes demographics, insurance status, and up to 3 reasons
for the visit. Physician information includes self-selected specialty, geographic
location, and if the practice is in a standard metropolitan statistical area
(SMSA). An SMSA approximates an urban region and is defined as an area that
includes a city of at least 50 000 people or that has a total area population
of at least 100 000 people. Information about SMSAs has been included
in the NAMCS since 1990. Visit characteristics include up to 3 diagnoses (coded
using the International Classification of Diseases, Ninth
Revision, Clinical Modification), 6 entries for medications, and the
type of visit (eg, acute, chronic follow-up).
We limited our analysis to adults with a primary reason for their visit
of "symptom referable to the throat" (reason for visit code, 1455; over 97%
complained of "soreness" or "pain" in the throat). We included only visits
to primary care physicians, including those who identified themselves as belonging
to the specialties of family practice, general practice, internal medicine,
adolescent medicine, geriatrics, and general preventive medicine. In all analyses,
adolescent medicine physicians were included with family practice physicians;
geriatric and general preventive medicine physicians were included with internal
We excluded patients younger than 18 years and patients immunosuppressed
with diabetes mellitus, human immunodeficiency virus infection, or cancer.
We also excluded patients with a primary diagnosis of sinusitis, an alternate
cause of sore throat that might prompt antibiotic treatment (7% of the sample
in a preliminary analysis). We excluded visits described as nonacute; chronic,
flare-up; chronic, routine; preoperative; postoperative; or visits that were
the result of an injury. This final sample consisted of 2244 patient visits.
We calculated annual national rates of antibiotic prescriptions for
patients with sore throat and the rate of nonrecommended treatment for patients
with sore throat. We assessed changes in the distribution of different antibiotic
classes over 11 years. For ease of interpretability, we also grouped changes
in the use of antibiotics and nonrecommended antibiotics into the periods
of 1989-1992, 1993-1996, and 1997-1999.
Antibiotics were identified according to the National
Drug Code Directory18 classification
of "antimicrobial agents." We included only nontopical antibacterial agents
in the analysis. Because the National Drug Code Directory has only 8 categories for antimicrobial agents, we further subdivided
antibiotics into 37 different classes based on pharmacology and route of administration.
Recommended treatment was defined as the use of penicillin, including
intramuscular preparations, or erythromycin for patients who received antibiotics.
If more than 1 antibiotic was used in a single visit (2.5% of visits), each
antibiotic was considered prescribed in its respective class. For all other
calculations and statistical tests, use of multiple antibiotics within a single
visit was counted as a single antibiotic use. If a patient received both a
recommended and a nonrecommended antibiotic, they were considered to have
received only a recommended antibiotic.
To ensure the validity of our main analysis, all weights and analyses
were recalculated after excluding any diagnosis—primary, secondary,
or tertiary—that could account for an antibiotic prescription other
than pharyngitis. Excluded diagnoses included acute bronchitis, acute exacerbation
of chronic bronchitis, staphylococcal infection, bacterial infection, gonococcal
infection, lymphadenitis, otitis media, otitis externa, pneumonia, urinary
tract infections, and skin infections. Exclusion of these diagnoses reduced
the sample size by 17.5% from 2244 to 1852 for this secondary analysis.
Ninety-five percent confidence intervals (CIs) for percentage estimates
were obtained by calculating relative SEs as recommended by the National Center
for Health Statistics.1 Coefficients for these
calculations are dependent on physician specialty and year. Because our analysis
spans 11 years and 3 specialties, we used the single largest relative SE available.
We used modified sample weights, according to the method of Potthoff et al,19 to derive effective sample sizes that account for
clustering by physician practice.
We used linear trends to assess changes in antibiotic use over time.
Multivariable logistic regression was used to evaluate independent predictors
of antibiotic prescription and use of nonrecommended treatments. In the first
model, the use of any antibiotic was the dependent variable. In the second
model, use of a nonrecommended antibiotic among those receiving an antibiotic
was the dependent variable. Because patient age had a roughly linear effect
on these 2 outcomes, we modeled patient age as the odds ratio (OR) associated
with each increasing decade of age. We evaluated interaction terms, none of
which were significant or included in the final models.
All analyses were done with SAS statistical software, version 8.01 (SAS
Institute Inc, Cary, NC). P values less than .05
were considered significant, except for univariate trend tests by year, where,
because of multiple comparisons, P values less than
.001 were considered significant.
There were an estimated 6.7 million (range by year, 5.1 million to 8.7
million) annual visits to office-based, community physicians by adult patients
with sore throat between 1989 and 1999 that met our inclusion and exclusion
criteria. There was no significant change in the number of estimated annual
visits over 11 years. The most frequent physician-reported diagnoses were
acute pharyngitis (41%), upper respiratory infection (21%), acute tonsillitis
(8%), streptococcal infection (6%), acute bronchitis (5%), and acute laryngitis
The sample had a mean age of 38 years (SD,13) and was 35% male and 79%
white, non-Hispanic (Table 1).
Forty-one percent of patients had private insurance, 29% made some type of
self-payment at their visit, and 23% belonged to a health maintenance organization.
Family practice physicians, internal medicine physicians, and general practice
physicians accounted for 45%, 29%, and 25% of visits, respectively. Twenty-three
percent of visits occurred in non-SMSA areas.
Over the 11-year period studied, antibiotics were prescribed in 73%
(95% CI, 70%-76%) of visits (Table 2).
Recommended antibiotics were prescribed in 23% (95% CI, 20%-26%) of visits.
Nonrecommended antibiotics were prescribed in 49% (95% CI, 46%-52%) of visits.
Among those receiving antibiotics, 68% (95% CI, 64%-72%) received nonrecommended
Recommended antibiotic use decreased from 32% of visits in 1989 to 11%
of visits in 1999 (P for trend, <.001). There
was a significant decrease in the use of both penicillin (17% of visits in
1989 to 6% in 1999; P for trend, <.001), and erythromycin
(15% of visits in 1989 to 5% in 1999; P for trend,
Nonrecommended antibiotic use increased from 45% of visits in 1989 to
56% of visits in 1998, but decreased to 46% of visits in 1999 (P for trend, <.001). There was a significant increase in the use
of extended-spectrum macrolides (P for trend, <.001)
and extended-spectrum fluoroquinolones (P for trend,
Antibiotics were prescribed in 76% of visits in 1989-1992, 71% of visits
in 1993-1996, and 69% of visits in 1997-1999 (P for
trend, .01; Table 3). The proportion
of visits in which nonrecommended antibiotics were given to patients who received
an antibiotic increased from 60% in 1989-1992, to 74% in 1993-1996, to 80%
in 1997-1999 (P for trend, <.001).
Internal medicine physicians significantly decreased the proportion
of visits in which an antibiotic was prescribed from 74% of visits in 1989-1992,
to 64% in 1993-1996, to 60% in 1997-1999 (P for trend,
.01). However, when prescribing an antibiotic, internal medicine physicians
significantly increased their use of nonrecommended antibiotics: from 66%
of visits 1989-1992, to 73% in 1993-1996, to 88% in 1997-1999 (P for trend, .003). Family practice and general practice physicians
also increased their use of nonrecommended antibiotics, but without a significant
change in their overall use of antibiotics.
In multivariable modeling, antibiotic use was independently predicted
by patient age (OR, 0.86 per increasing decade; 95% CI, 0.79-0.94) and general
practice specialty (OR, 1.54 compared with family practice specialty; 95%
CI, 1.10-2.14; Table 4). Among
patients who received antibiotics, nonrecommended antibiotic use was predicted
by calendar year (OR, 1.17 per year; 95% CI, 1.11-1.24). Nonrecommended antibiotic
use was less likely with health maintenance organization coverage (OR, 0.66;
95% CI, 0.44-1.00).
To ensure the validity of our findings, we reanalyzed the data after
removing 392 visits with alternate diagnoses that could potentially account
for antibiotic use. The percentage of visits in which antibiotics were used
decreased from 73% to 71% (95% CI, 68%-74%). The trend in use of extended
fluoroquinolones no longer met our criteria for significance (P = .002). In the multivariable models, the results changed modestly
from the main analysis to the reanalysis. In the first model, age (OR, 0.83
per decade; 95% CI, 0.75-0.91), general practice specialty (OR, 1.45 compared
with family practice specialty; 95% CI, 1.01-2.07), and self-payment (OR,
1.62; 95% CI, 1.07-2.47) were significant predictors of antibiotic use. In
the second model, calendar year (OR, 1.18; 95% CI, 1.10-1.25) was the sole
significant predictor of nonrecommended antibiotic use.
In a national community sample of adults presenting to primary care
physicians with a chief complaint of sore throat, 73% were treated with antibiotics.
This greatly exceeds the 5% to 17% prevalence of GABHS in adults with sore
From 1989 to 1999, there has been a marginally significant decrease in the
use of antibiotics overall, but an increase in the use of more expensive,
broader-spectrum antibiotics. Encouragingly, there was a decrease in the use
of almost all antibiotic classes in 1999.
Predictors of antibiotic use for sore throat were younger patient age
and general practice specialty. Prescribing antibiotics more frequently for
younger patients makes biologic sense. Younger patients are more likely to
be infected with GABHS, and some treatment algorithms for sore throat incorporate
younger patient age as a risk factor for GABHS infection.4,13
We also found in multivariable analysis that general practice physicians
prescribed antibiotics about 50% more frequently than family practice physicians.
According to the masterfiles of the American Medical Association, those identifying
themselves as general practice physicians tend to be older than other physicians:
47% are older than 65 years, compared with 17% of all physicians.20 In our analysis, general practice specialty may represent
an older group of physicians accustomed to prescribing antibiotics for upper
respiratory tract infections.
The only predictor for the use of nonrecommended antibiotics was calendar
year. This effect persisted despite controlling for patient demographics,
location, physician specialty, and after eliminating potential competing diagnoses
in our reanalysis. We found a significant increase in the use of the extended-spectrum
macrolides and a small, but statistically significant increase in the use
of the extended-spectrum fluoroquinolones.
While many broad-spectrum antibiotics are effective in eradicating GABHS
from the throat,21,22 there are
2 main concerns in the use of these agents. The first is cost. A course of
azithromycin is roughly 20 times more expensive than a course of penicillin
($39.32 vs $2.31), and a course of oral cephalosporins ranges in price from
$4.41 to $80.05.23
The second concern with the use of broad-spectrum agents is development
of bacterial resistance. Extended-spectrum fluoroquinolones are capable of
inducing resistance among GABHS isolates in vitro.24
In vivo, macrolide resistance among GABHS isolates varies between 2% and 17%25,26 and is proportional to local macrolide
use.27 Fortunately, interventions aimed at
lowering macrolide use decreases the prevalence of macrolide-resistant GABHS
In contrast, GABHS has remained exquisitely sensitive to penicillin.
Over the past 80 years, there has never been a GABHS isolate resistant to
penicillin found in Europe25 or North America.26,29,30 Despite a small risk
of serious allergic reaction, penicillin is generally well tolerated, inexpensive,
and effective. Penicillin reduces the duration of symptoms31
and is the only antibiotic proven to prevent rheumatic fever in patients with
sore throat.32- 34
Our analysis has limitations that deserve comment. Detailed historical
information such as duration of symptoms or history of allergy is not available.
We also do not have access to detailed clinical information, such as vital
signs, physical examination signs, or the test results for GABHS. As such,
we cannot definitively assess the appropriateness of antibiotic use for any
individual visit. Despite these limitations, the disparity between the known
prevalence of GABHS among adults with sore throat and the proportion of visits
in which an antibiotic was used is striking. This disparity persisted in a
reanalysis of our data after removing 17.5% of visits that had a potential
alternative reason for antibiotic use.
Our analysis of the NAMCS shows that over half of adult patients with
sore throat are treated with antibiotics by community primary care physicians.
Educational and policy efforts to reduce unnecessary antibiotic use may be
having some effect, as evidenced by a decrease in the proportion of patients
receiving antibiotics in 1999. However, the use of expensive, broad-spectrum
antibiotics that can induce resistance is still frequent. Efforts should be
continued to encourage appropriate antibiotic use by both patients and physicians.