Summary of main findings
This study found that antibiotic prescribing volumes varied twofold between practices in the highest and lowest deciles of antibiotic prescribing. Location in the north of England was the strongest predictor of high antibiotic prescribing.
Practices serving populations with higher morbidity and a higher proportion of white patients prescribed more antibiotics, as did practices with shorter appointments, non-training practices, and practices with a higher proportion of GPs who were male, aged >45 years, and non-UK-qualified.
Group status and social deprivation were only weak predictors of antibiotic prescribing. There was no detectable association between antibiotic prescribing and list size per FTE GP.
Strengths and limitations of the study
In this study, data from 8057 practices covering 97% of the registered population in England were analysed. The study regression model explained around one-sixth of the variance in antibiotic prescribing. Data on consultation rates and on the proportion of consultations resulting in antibiotic prescribing might help to explain the remainder of the variance, but were not obtainable in this study.
Other factors that the current methodology did not incorporate include indications for antibiotics, types, doses, and durations of antibiotic courses prescribed, and clinical outcomes.
Other limitations of this study relate to the level of detail available from the data sources. It was not possible to adjust analysis for patient comorbidity levels based on practice-level chronic disease prevalence data because many patients feature on multiple chronic disease registers. Although this study did control for patient morbidity levels, these were not standardised for age or sex and only provided a subjective measure of health status.
This study was also unable to perform a detailed analysis of smoking data because the proportion of smokers in each practice was unknown.
Comparison with existing literature
The fivefold difference in antibiotic prescribing between practices at the extremes of the study dataset is consistent with that reported by Majeed and Moser.12 However, the range is not generally regarded as a satisfactory measure of variation, particularly as it increases without limit for larger samples.22,23 It is included here only for comparison with other studies. The twofold variation between the 10th and 90th centiles is considered to be more typical of the variation in antibiotic prescribing than the fivefold variation between the highest and lowest prescribing practices. Taken as a whole, the top one-fifth of practices prescribe more than twice the volume of antibiotics prescribed by the bottom one-fifth of practices (28% compared to 13% of all antibiotic prescriptions). However, antibiotic prescribing in England has only decreased by about 25% since 1995.1
This study confirms the findings of some regional studies. It also found that practices' non-training status and areas of greater social deprivation were associated with higher antibiotic prescribing, although after adjustment for confounders these appeared to be weaker predictors of antibiotic prescribing than previously suggested.13–15
The finding that practices with shorter appointment times prescribe more antibiotics is consistent with reports that GPs who perceive their consultation times to be more limited prescribe more antibiotics.24 These GPs may feel they have less opportunity to discuss patients' concerns and expectations during a consultation, resulting in more medication being prescribed.25
Doctors also often overestimate patients' expectations regarding antibiotics,26,27 and some GPs still prescribe antibiotics in an effort to maintain good doctor–patient relationships.28 However, several studies have shown that patient satisfaction is more strongly associated with the time GPs spend listening to patients,29 and providing information and reassurance.26,28,30 GPs who perceive greater pressure from patients to prescribe are more likely to be overseas-qualified,31 male,32 and more years from qualification.32 More specifically, overseas-qualified GPs15,33 and GPs who have been practising for longer32–35 are reported to prescribe more antibiotics and to offer fewer non-pharmacological treatments.36 This is consistent with the current findings that practices with a higher proportion of GPs who are male, >45 years old, and non-UK-qualified prescribe more antibiotics.
However, there is little evidence in the existing literature to help explain the observation that practice location in the north of England is the strongest predictor of high antibiotic prescribing, even after adjustment for social deprivation and other confounders. Higher antibiotic prescribing in the north may be partly driven by higher consultation rates for RTIs,37 possibly as a result of lower temperatures,38 higher levels of environmental pollution (particularly in industrial areas),16,39 and higher smoking prevalence.40 The latter may result in greater prevalence of chronic respiratory conditions41 and cardiac failure secondary to coronary heart disease,19 further increasing the likelihood of requiring antibiotics for intercurrent RTIs. However, antibiotic prescribing for RTIs only accounts for around 35% of all antibiotic prescriptions.4 Complex interactions between environmental, occupational, and social factors, which are difficult to measure, are likely to help further explain the current findings.16
Detailed exploration of the attitudes and expectations of patients from different ethnic groups regarding antibiotics is also largely unreported in the literature. It was observed in the present study that practices serving populations with a higher proportion of white patients prescribe more antibiotics. However, a previous study reported no ethnic differences in the proportion of patients receiving antibiotic prescriptions after consulting with upper RTIs.42
Various cultural factors may determine the likelihood of patients from different ethnic groups consulting their GPs with acute illnesses or receiving antibiotics. Afro-Caribbean patients are reported to self-administer medication more frequently than white or Asian patients before consulting.42 Older Asian Gujarati patients have also been found to consult with their GPs less often than older white patients because of poorer understanding of health services and greater availability of alternative sources of support.43 Therefore, it is likely that the relationship between patient ethnicity and antibiotic prescribing is highly context-dependent.
A similar principle may apply to the association between social deprivation and antibiotic prescribing. Previous practice level studies of antibiotic prescribing used measures of social deprivation (such as the Low Income Scheme Index14 and Townsend score15) which focus mainly on low income, unemployment, and poor living conditions. However, these are less comprehensive than the IMD-2004 score, which measures social deprivation across seven domains.20 Some GPs report a lower threshold for prescribing antibiotics for patients if they have poor nutrition and live in poor, overcrowded housing because of concerns about their increased susceptibility to bacterial complications.44 However, the current study found that increased deprivation in the education, skills, and training domain of the IMD-2004 had a stronger association with high antibiotic prescribing than deprivation in other domains, although European studies have shown that this association varies according to geographical location.45,46
Implications and further study
Several countries have undertaken national campaigns to reduce inappropriate antibiotic prescribing.2 The present findings suggest that if further campaigns are to reach practices that are high antibiotic prescribers, it may be useful to focus on practices in the north of England, non-training practices, and practices with higher proportions of male GPs, GPs aged >45 years, and non-UK-qualified GPs. Qualitative studies may be useful in helping to explore the attitudes, experiences, and backgrounds of these GPs and practices in relation to antibiotic prescribing. However, given that antibiotic prescribing only varies twofold between practices in the highest and lowest deciles, the need for GP practices as a whole to reduce unnecessary antibiotic prescribing should still be emphasised.
This study highlights the importance of adjusting estimates of the relative level of antibiotic prescribing according to the morbidity of the practice population, as practices serving populations with high morbidity levels would be expected to prescribe more antibiotics.
Further study using national consultation level data is needed to examine whether increased antibiotic prescribing in some regions of England is linked more strongly to consultation rates for certain infective conditions or to an increased likelihood of being prescribed antibiotics during a consultation.
Practice location in the north of England was found to be the most important predictor of high antibiotic prescribing volumes at practice level, even after adjustment for confounding factors. However, practice characteristics and patient demographics only explain about one-sixth of the variation in antibiotic prescribing between practices.
Further examination of antibiotic prescribing decisions at consultation level would improve current understanding of interpractice variation in antibiotic prescribing and guide future interventions aiming to reduce inappropriate antibiotic use.
