Investigating relationships between a range of potential indicators of general practice prescribing: an observational study

Objectives: (i) to describe the variation in a number of potential prescrib ing indicators across a large number of general practices, (ii) to determin e whether particular indicators are associated with higher or lower overall prescribing costs, (iii) to determine how particular indicators are correl ated with each other, (iv) to determine whether the indicators are stable o ver time. Design: Observational study based on the analysis of PACT (Prescribing Anal ysis and Cost) data for general practices. Setting: The Trent region of England. Subjects: 809 general practices. Main outcome measures: Variations in a range of potential prescribing perfo rmance indicators; correlations between these indicators and NIC (net ingre dient costs) per ASTRO-PU (Age, sex and temporary resident originated presc ribing unit); change in indicators between the financial year 1994-5 and 19 95-6. Results: Wide variations were found between practices in a range of potenti al prescribing performance indicators. There were positive correlation (P < 0.001) between NIC per ASTRO-PU and indicators of relatively high cost pre scribing within different therapeutic areas. There were negative correlatio ns (P < 0.001) with indicators of relatively low cost prescribing. There we re positive intercorrelations (P < 0.001) between indicators of relatively high cost prescribing. For most indicators there were significant changes ( P < 0.001) in values between the two years. Conclusion: There are considerable variations between general practices in prescribing performance indicators and the indicators show changes over tim e. We have demonstrated correlations between different indicators that lend support to the idea that such indicators might act as proxies for more gen eral prescribing patterns. However, most of the correlations we observed we re not strong. Further research is needed into the development and use of i ndicators based on PACT data.