Abstract
OBJECTIVE To determine whether use of clinical decision rules or rapid streptococcal antigen detection tests (alone or in combination) can lower the number of unnecessary prescriptions for antibiotics for adults with acute sore throats.
DESIGN Four-arm randomized controlled trial.
SETTING Family practice offices in eastern Newfoundland.
PARTICIPANTS Forty urban and suburban family practitioners.
INTERVENTIONS Participants were randomly assigned to one of 4 arms (usual practice, decision rules only, rapid antigen test only, decision rules and antigen test combined), and each recruited successive adult patients presenting with acute sore throat as their main symptom. Following usual care or use of decision rules or rapid antigen tests or both (where applicable), physicians were to record what they prescribed for each patient.
MAIN OUTCOME MEASURES Prescribing rates and types of antibiotics prescribed.
RESULTS The prescribing rate using decision rules (55%) did not differ significantly from the rate using usual clinical practice (58%). Physicians using rapid antigentests, both alone and with decision rules, had significantly lower prescribing rates (27% and 38%, respectively, both P < .001).
CONCLUSION Evidence-based clinical decision rules alone do not change family doctors’ prescribing behaviour. Use of rapid antigen tests might allow physicians to persuade patients that negative results (and hence, viral infection) mean antibiotic therapy is not required.
Acute respiratory tract infections are the reason for about a quarter of all visits to family doctors in North America, and a quarter of these visits are for acute sore throat.1 The large number of sore throats means that they account for 3% to 6% of all office visits.2,3
In adults, 85% to 90% of sore throats are caused by viral infections.4 Treating patients who have sore throats with antibiotics does not relieve symptoms very much, if at all. A study of patients with tonsillitis in 17 countries found that the mean duration of fever was 2 to 3 days, regardless of whether or not patients took penicillin.5 Patients who test positive for group A β-hemolytic streptococcus (GABHS) who are treated with penicillin have relief of symptoms about 16 hours earlier than those who test negative for GABHS.6 Treated or not, 85% of patients are completely free of symptoms at 1 week.7
Mentioning a sore throat to a doctor almost guarantees a prescription for antibiotics. In Australia, 89% of patients with sore throats got antibiotic prescriptions8; in the United States, the rate was 73%.9 A study of 73 Newfoundland family doctors found that 84% of them prescribed antibiotics to adults with sore throats.10 In Holland, which has a tradition of low prescribing rates for antibiotics, the rate was still 52%.11
Attempts have been made to derive simple sore throat decision rules (STDR) so that doctors can more appropriately prescribe antibiotics for patients likely to have GABHS. In general, 4 clinical features have been found useful in deciding which adults are most likely to have GABHS infection: fever, tonsillar exudates, anterior cervical lymphadenopathy, and the absence of cough (Table 112). A Canadian study found that using clinical rules would have reduced prescriptions for antibiotics among adults with sore throats by 88%13 if doctors had followed these rules. Another study showed that, while using the rules improved physicians’ estimates of the presence or absence of GABHS, it did not alter how they used antibiotics.14 We wanted to know whether using STDR would change doctors’ antibiotic prescribing practices appropriately.
Score of =1—no need for antibiotics, 3 or 4—antibiotics are required, 2—antibiotics might or might not be beneficial.
Rapid antigen detection tests (RADT) provide results much more quickly than the criterion standard of GABHS detection, agar plate culture, does. Rapid antigen detection tests have a specificity in the 95% range, so false-positive results are rare.15 The sensitivity of RADT increases with the number of positive clinical features a patient has.16 The primary advantage of RADT over throat-swab cultures is that results can be available in only 5 to 10 minutes, and the test costs as little as $5. Treatment, if needed, can be started before patients leave the doctor’s office.
A literature review using the terms “rapid antigen test,” “streptococcal sore throat,” and “primary care” found that RADTs have been used in various parts of the world: Denmark,17 New Mexico,18 the Netherlands,17 Israel,19 the Canary Islands,20 and Switzerland.21 In general, these studies concluded that the RADT was a valid test for diagnosing streptococcal sore throat in adults and that it had high specificity and positive predictive values. The value of RADT for children was less clear.
Using either STDR or RADT should allow family doctors to make rapid and more informed decisions than when they use usual clinical judgment. The objectives of this trial were to compare rates of diagnosis of likely GABHS infection, represented by prescriptions for antibiotics, using usual clinical judgment, STDR, RADT, and both STDR and RADT, and to assess whether STDR and RADT are better used alone or in combination.
METHODS
This randomized controlled trial was conducted in family doctors’ offices in eastern Newfoundland during February, March, and April 2005. The study was approved by Memorial University of Newfoundland’s Human Investigations Committee.
Physicians known not to be in family practice (either from the investigators’ personal knowledge or from the records of the Newfoundland and Labrador Medical Association) were removed from the list of non-specialist physicians in the study area. Successive random blocks of 40 family physicians were approached to take part in the study (using the Dillman technique) until 40 physicians were recruited.
The trial had 4 arms: a control group continuing usual clinical practice, a group using STDR, a group using RADT, and a group using both STDR and RADT. Physicians in the control arm were asked to follow their usual clinical practices and record their prescribed management. All others were required to administer the appropriate intervention first and then record their management.
We based our STDR upon those developed and tested for adults by Centor and colleagues at the University of Virginia,12 but modified them to reflect evidence from other studies of adult patients (Table 112).22,23 Once doctors assessed patients using STDR criteria, they made recommendations based on total scores. A score of =1 suggested there was no need for antibiotics, while a score of 3 or 4 suggested that antibiotics were required. A score of 2 indicated that antibiotics might or might not be beneficial. When using the combined STDR and RADT intervention, doctors were asked to use the RADT only when the score on the STDR was 2. We only suggested what physicians should do with each category of score; we did not command or expect obedience from family doctors.
The RADT used in the trial was the Clearview® Exact Strep A dipstick from Wampole Laboratories. This test is reported by the manufacturers to have a sensitivity of about 90% and a specificity of about 95%.
The 40 physicians who agreed to take part in the study were randomly allocated to 1 of 4 trial arms, and each was asked to recruit 20 successive adult patients (aged 19 years or older) who presented with acute sore throat as their primary symptom. We did not ask doctors to record clinical or demographic characteristics of the patients; we wanted the doctors to do as little extra work as possible. We oversampled to compensate for physicians who might not enter the total number of subjects requested. Figure 1 shows the trial design. To detect a 25% reduction in the rate of antibiotic prescribing, 98 patients were required in each trial arm (2-tailedα.05, β .20). χ2 comparisons using the Statistical Package for the Social Sciences for Windows, release 11.0.1, were used to test for differences in prescribing rates between the control group and the intervention groups. Analysis of covariance was used to compare differences in demographics among the groups of physicians.
Trial design and number of patients in each arm receiving and not receiving antibiotics
RESULTS
Of the 204 general practitioners in eastern Newfoundland, 157 were in active family practice. Three successive samples of 40 physicians were asked to be in the study until a total of 40 physicians was recruited. Of the 120 physicians in these random samples, 40 agreed to take part and 80 declined. The first, second, and third efforts recruited 11, 17, and 12 doctors, respectively. Table 2 shows some characteristics of the physicians by study group. There were no significant differences between doctors in the 4 groups, nor were the study group doctors different from the 80 physicians who declined to take part in the study (analysis of covariance, F = 2.12, P > .05). In all, 37 of the 40 recruited physicians actually entered patients in the trial.
Characteristics of physicians recruited to each study arm and of non-participating physicians: None of the differences between groups was significant (analysis of covariance: F = 2.12; df = 4,19).
The number of patients entered was 533, with a mean of 14.5 patients for each doctor (range, 4 to 20). The physicians wrote a total of 247 prescriptions for antibiotics (46.7% of patients received antibiotics).
Table 3 shows the number of patients recruited in each arm of the trial and the rates of antibiotic prescribing. Doctors in the usual-practice and STDR-only arms wrote prescriptions for approximately the same proportion of patients (58% and 55%, respectively). Doctors in the RADT-only and the combined STDR and RADT arms prescribed antibiotics to fewer patients (27% and 38%, respectively). Doctors in the RADT and combined STDR and RADT arms prescribed antibiotics to significantly fewer patients than doctors in the usual-practice group did (Table 3). Doctors in the groups who used RADT alone or in combination with STDR recruited fewer patients (120 and 102, respectively) than doctors in the usual-practice and STDR-only groups did (141 and 170, respectively).
Percentage of antibiotic prescriptions by study arm
Amoxicillin was the most commonly prescribed antibiotic; 115 of 247 (47%) prescriptions were for amoxicillin. Penicillin was second, with 50 prescriptions (20%). A variety of other antibiotics constituted less than 10% of prescriptions each.
DISCUSSION
Inherent in the design of this trial was a potential clustering of patients by physician. Clustering has the effect of widening confidence intervals. The 95% confidence levels of the proportions of antibiotic prescriptions in the higher-prescribing groups (usual care and STDR) did not overlap with those of the lower-prescribing groups (RADT and combined STDR and RADT). Accordingly, we think that differences we observed were so great that our results are robust.
In this trial, we found that use of the RADT, either as the only diagnostic aid or in combination with the STDR, was associated with a significantly lower rate of antibiotic prescribing. Use of the STDR alone did not appear to affect physicians’ prescribing rates. The lower antibiotic prescribing rates, of 27% in the RADT-only arm and 38% in the combined STDR and RADT arm, are similar to those found in a previous primary care study in Switzerland where use of the RADT reduced antibiotic prescribing from 60% to 37%.21
The apparent inability of the STDR to change doctors’ behaviour had been found in 2 previous studies.14,24 A recent survey of 1000 American pediatricians found that 42% would start antimicrobials before they knew diagnostic test results, and 27% of them would continue treatment even if test results were negative.25 We do not know why doctors are so resistant to changing their practice, even when they know the STDR and the test results.26 Our study suggests that doctors are more confident with the results of a diagnostic test than they are with decision rules. Negative RADT results seem to give physicians more confidence about explaining to patients why antibiotics are not needed for their throat infections.
Questions and concerns
Several intriguing questions remain about the use of RADT in doctors’ offices. Even after using RADT, the antibiotic prescribing rate was still 27%, higher than the known community infection rate with GABHS.4 Perhaps the GABHS infection rate in these patients was higher than usual, or perhaps some of the positive RADT results were from patients who were GABHS carriers. It is also possible that physicians in the RADT arms recruited patients with symptoms and signs more suggestive of GABHS. The relative inconvenience of performing the test might have discouraged them from enrolling those less likely to have GABHS and might help explain the lower overall enrolment in the RADT and combined STDR and RADT arms.
Of some concern was the finding that, when an antibiotic was used, it was likely to be amoxicillin (47%). Group A β-hemolytic streptococci are uniformly susceptible to penicillin V. Clearly patients who received amoxicillin were not allergic to penicillin, and we do not know why amoxicillin was prescribed so often. We are conducting a qualitative study of doctors who took part in this trial to gain more insight into their prescribing practices.
Conclusion
Our study found that use of the RADT in primary care offices (either alone or in conjunction with the STDR) was associated with a significant reduction in prescriptions for antibiotics among adults whose primary complaint was an acute sore throat.
Acknowledgment
Kathy Murphy was the research nurse on this study. Theresa Kerrivan coordinated the paperwork for the study and handled correspondence. Jennifer Braun helped with preliminary data analysis.
Notes
POINTS DE REPÈRE DU RÉDACTEUR
-
La consultation d’un médecin pour un mal de gorge entraîne presque toujours la prescription d’un antibiotique.
-
Les tests de détection rapide de l’antigène offrent une spécificité de 95% permettant ainsi d’exclure rapidement la plupart de faux positifs.
-
Ces tests se révèlent supérieurs aux lignes directrices pour convaincre le patient de l’inutilité de prendre des antibiotiques.
EDITOR’S KEY POINTS
-
Mentioning a sore throat to a doctor almost guarantees a prescription for antibiotics.
-
Rapid antigen detection tests have a 95% specificity, allowing rapid exclusion of false-positive results.
-
These tests are superior to guidelines for persuading patients that antibiotics will not be useful.
Footnotes
-
This article has been peer reviewed.
-
Contributors: Drs Worrall, Hutchinson, and Sherman and Mr Griffiths contributed to concept and design of the study; data gathering, analysis, and interpretation; and preparing the article for submission.
-
Competing interests
None declared
- Copyright© the College of Family Physicians of Canada
References
In this issue
Jump to section
Related Articles
Cited By...
- Awards season
- Impact analysis studies of clinical prediction rules relevant to primary care: a systematic review
- Developing an International Register of Clinical Prediction Rules for Use in Primary Care: A Descriptive Analysis
- PRImary care Streptococcal Management (PRISM) study: identifying clinical variables associated with Lancefield group A {beta}-haemolytic streptococci and Lancefield non-Group A streptococcal throat infections from two cohorts of patients presenting with an acute sore throat
- Clinical score and rapid antigen detection test to guide antibiotic use for sore throats: randomised controlled trial of PRISM (primary care streptococcal management)
- Incidence and clinical variables associated with streptococcal throat infections: a prospective diagnostic cohort study
- Should we treat strep throat with antibiotics?