Variation in Antibiotic Prescriptions for Sore Throat
Variation in Antibiotic Prescriptions for Sore Throat
Table 1 shows the baseline characteristics of the study population. A total of 6394 patients with sore throat were recruited by 457 GPs, 24.4% of the patients were under 15 year old. Only in Denmark and Sweden, 100% of the GPs had access to strep A test. Request of antibiotics varied across countries from 0.1% of the patients in Denmark to 9% of the patients in Russia.
Table 2 shows the ORs for prescribing antibiotics controlled by patient and GP characteristics. In general, characteristics at GP level as well as demographic characteristics of the patients were not associated with prescription of antibiotics.
Only in Russia, 4 or more days with symptoms was positively associated with prescription of antibiotics (odds ratio 6.5, 95% confidence interval 2 to 20). Furthermore, in Argentina, Russia and Spain, patient request for antibiotics was positively associated with prescription of antibiotics.
Figure 1 shows the crude variation in the prescription of antibiotics within and across countries. The median percentage of patients being prescribed antibiotics varied across countries from 38% (interquartile range (IQR) 22%-62%) in Spain to 88% (IQR 50%100%) in Sweden. There was variation in prescription of antibiotics within the countries too, represented by the asymmetry of the interquartile ranges. In all countries some GPs always prescribed antibiotics to all their patients while other GPs never did.
(Enlarge Image)
Figure 1.
Crude variation in prescription of antibiotics per country. Box-and-whisker plot shows proportions of patients prescribed antibiotics per country. The horizontal line inside the box shows the median percentage of patients prescribed antibiotics for sore throat and the upper and lower end of each box give the 75 and 25 interquartile ranges, respectively. The area between the different parts of the box indicates the degree of dispersion and skewness of data. The ends of the whiskers represent the maximum and minimum percentage of patients that were prescribed antibiotics.
Figure 2 shows the multilevel analysis of the variance of GPs' prescribing style (model A) and the changes after adjusting for patient characteristics (Model B) and afterwards adding GP's characteristics (Model C).
(Enlarge Image)
Figure 2.
Unadjusted and adjusted Median Odds Ratios (MOR) per country. The diagram shows the multilevel analysis of the variance of GPs' prescribing style. Model A (light grey): prescription of antibiotics is only a function of GPs' prescribing style. Model B (medium grey): prescription of antibiotics is a function of GPs' prescribing style and patient characteristics. Model C (dark grey): prescription of antibiotics is a function of GPs' prescribing style, patient and GP characteristics. When MOR = 1, there is no variation in GPs' prescribing styles. The higher the MOR, the more variation in GPs' prescribing styles.
After adjusting by patient and GPs' characteristics (model C), the Median Odds Ratio was consistently greater than 1 within countries and varied across countries. The most heterogeneous group of practitioners was found in Russia (Kaliningrad) Median Odds Ratio (MOR 6.8, 95% CI 3.1; 8.8). It means, in Russia a randomly chosen patient has a median 6-fold risk of being prescribed antibiotics if consulting a GP with a higher tendency to prescribe antibiotics.
The group of Danish GPs had the most homogeneous prescribing style (MOR 2.6, 95% CI 2.2; 4.4).
Results
Table 1 shows the baseline characteristics of the study population. A total of 6394 patients with sore throat were recruited by 457 GPs, 24.4% of the patients were under 15 year old. Only in Denmark and Sweden, 100% of the GPs had access to strep A test. Request of antibiotics varied across countries from 0.1% of the patients in Denmark to 9% of the patients in Russia.
Table 2 shows the ORs for prescribing antibiotics controlled by patient and GP characteristics. In general, characteristics at GP level as well as demographic characteristics of the patients were not associated with prescription of antibiotics.
Only in Russia, 4 or more days with symptoms was positively associated with prescription of antibiotics (odds ratio 6.5, 95% confidence interval 2 to 20). Furthermore, in Argentina, Russia and Spain, patient request for antibiotics was positively associated with prescription of antibiotics.
Figure 1 shows the crude variation in the prescription of antibiotics within and across countries. The median percentage of patients being prescribed antibiotics varied across countries from 38% (interquartile range (IQR) 22%-62%) in Spain to 88% (IQR 50%100%) in Sweden. There was variation in prescription of antibiotics within the countries too, represented by the asymmetry of the interquartile ranges. In all countries some GPs always prescribed antibiotics to all their patients while other GPs never did.
(Enlarge Image)
Figure 1.
Crude variation in prescription of antibiotics per country. Box-and-whisker plot shows proportions of patients prescribed antibiotics per country. The horizontal line inside the box shows the median percentage of patients prescribed antibiotics for sore throat and the upper and lower end of each box give the 75 and 25 interquartile ranges, respectively. The area between the different parts of the box indicates the degree of dispersion and skewness of data. The ends of the whiskers represent the maximum and minimum percentage of patients that were prescribed antibiotics.
Figure 2 shows the multilevel analysis of the variance of GPs' prescribing style (model A) and the changes after adjusting for patient characteristics (Model B) and afterwards adding GP's characteristics (Model C).
(Enlarge Image)
Figure 2.
Unadjusted and adjusted Median Odds Ratios (MOR) per country. The diagram shows the multilevel analysis of the variance of GPs' prescribing style. Model A (light grey): prescription of antibiotics is only a function of GPs' prescribing style. Model B (medium grey): prescription of antibiotics is a function of GPs' prescribing style and patient characteristics. Model C (dark grey): prescription of antibiotics is a function of GPs' prescribing style, patient and GP characteristics. When MOR = 1, there is no variation in GPs' prescribing styles. The higher the MOR, the more variation in GPs' prescribing styles.
After adjusting by patient and GPs' characteristics (model C), the Median Odds Ratio was consistently greater than 1 within countries and varied across countries. The most heterogeneous group of practitioners was found in Russia (Kaliningrad) Median Odds Ratio (MOR 6.8, 95% CI 3.1; 8.8). It means, in Russia a randomly chosen patient has a median 6-fold risk of being prescribed antibiotics if consulting a GP with a higher tendency to prescribe antibiotics.
The group of Danish GPs had the most homogeneous prescribing style (MOR 2.6, 95% CI 2.2; 4.4).