Variation in antibiotic prescribing in primary care practices

Problem

This study aims to describe antibiotic prescribing for sore throat and LRTI in UK general practice in relation to illness severity and to explore the extent of potentially inappropriate prescribing and whether this varies by GP practice.

Approach

Data is taken from the Descarte cohort for sore throat (12,829 participants) and the Cough Complications Cohort Study (28,883 participants). We describe variations in prescribing rates according to the severity of patient illness at baseline. Accounting for individual patient baseline presentation, sociodemographic characteristics and practice-level deprivation, we calculated the median odds ratio to quantify the practice-level variation in prescribing rates. The median odds ratio represents the probability that two randomly selected practices will prescribe to an identical patient.

Findings

There was substantial variability in prescribing rates by practice. For sore throat the range was 0%-97% (median 46%, IQR 33%-61%) of patients receiving an immediate antibiotic. For cough, the range was 7%-100% (median 62%, IQR 47%-75%). There was some evidence that higher prescribing practices do see a higher proportion of unwell patients. At the individual level, more severely ill patients are more likely to receive a prescription but prescribing levels for those will low severity scores are still high – 30% of sore throat patients with a Centor score 0 or 1 received a prescription and 50% of those with no signs/symptoms associated with adverse outcome for cough. The median odds ratio indicated that a sore throat patient seen at one practice would be 2.5 times more likely to receive a prescription than an identical patient seen at another practice. For cough, the figure was 2.8 times. This suggests that there is substantial practice level variation in prescribing, which cannot be explained by the illness severity of the patients that they see.

Consequences

Higher prescribing practices do see more unwell patients. However, this does not fully explain the differences that we observed in prescribing rates. The odds were substantially higher of an identical patient receiving an immediate prescription in some practices than in others. Focussing on patients with low risk of adverse outcome and where antibiotics are very unlikely to provide benefit could result in substantial reductions in prescribing.