Introduction

As drug expenditure in the NHS increases each year, appropriate and cost effective prescribing is a high priority for the NHS1. In efforts to contain their prescribing budgets, the PCTs monitor prescribing trends using Prescribing Analysis and CosT (PACT) database which details all dispensed NHS prescriptions. The prescribing teams within PCTs provide prescribing advice and support through outreach activities to General Practitioner (GP) practices in their areas. Both research and experience identify that the difference in prescribing profiles between practices in the same PCTs can be large. Even in 1976 high-cost prescribers were identified as more often foreign-trained and working in small practices2. This abstract describes how prescribing varies within a PCT and how these differences relate to practices with different characteristics and demographics of the prescribing GPs.

Methods

K-means cluster analysis was undertaken using six variables that related to prescribing influences including peers, pharmaceutical advisors (PAs) and drug representatives (drug reps) (table 1)3. Three clusters of GP-identity were identified, to which prescribing data from the PACT database was linked. Key prescribing indicators from the PACT database were selected. Three years of prescribing data was retrieved from the PACT database. Time-series trend analysis was conducted.

Results and Discussion

Cluster 1 comprised smaller practices with significantly more men with older medical degrees. Cluster 3 comprised opposite characteristics, i.e. female doctors with more recent degrees working in teaching practices. Table 1 compares data on the cluster variables. Six prescribing indicators were included in the analysis, four of which were Commission for Health Improvement indicators. The prescribing areas included the volume ratios of antibiotics; benzodiazepines; ACE-inhibitors and atypical antipsychotics, cost of drugs of limited clinical value and generic prescribing. The prescribing by each cluster separated well as exemplified in figure 1.

Figure 1. Generic prescribing in the clusters

Table 1. Mean item scores of cluster variables

¤¤ =p<0.001 ¤=p<0.05

Across the prescribing areas there was a general trend for cluster 1 to correlate less well with the national and local prescribing targets whereas prescribing within cluster 3 moved steadily towards target. On each of the six indicators cluster 3 members were always nearest to target. Cluster 2 prescribers showed the most erratic prescribing; large fluctuations within a prescribing indicator, as well as little consistency between indicators. Whilst Cluster 3 was closest in line with national and local prescribing objectives, equating this with 'highest quality of prescribing' is debatable. From the analysis to date, we can predict prescribing based on GP demographics and practice characteristics. Our findings have implications for effective targeting and management of prescribing outreach in PCTs. Knowing what influences different prescribers can inform PCTs in targeting prescribing interventions differently to different practices.

(1). Prescription Pricing Authority Home page (www.ppa.org.uk) accessed on 20/11/03
(2). Eaton G and Parish P. Ch 9 "High-cost prescribing doctors" In: Prescribing in General Practice. London: RCGP.
(3). Åström K., Duggan C., Bates I. Influences on the prescribing of general practitioners. Int. Pharm Pract, 2003; 11(suppl): R89.