Geographical and organisational variation in the structure of primary care services: implications for study design
Aims: To evaluate the extent to which structural variation between English general practices is accounted for at higher organisational levels in the National Health Service (NHS). Methods: We analysed data for 11 structural characteristics of all general practices in England. These included characteristics of general practitioners (GPs), the practice list and the services provided by practices. A four-level random effects model was used for analysis and components of variance were estimated at the levels of practice, primary care group (PCG), health authority and region. Results: The proportion of single-handed practices ranged from 0% to 74% at PCG level and from 14% to 43% in different regions. The proportion of practices providing diabetes services ranged from 0% to 100% at PCG level and from 71% to 96% in different regions. The list size per GP ranged from 1314 to 2704 patients per GP at PCG level and from 1721 to 2225 at regional level. Across the 11 variables analysed, components of variance at general practice level accounted for between 43% and 95% of the total variance. The PCG level accounted for between 1% and 29%, the health authority level for between 2% and 15% and the regional level for between 0% and 13% of the total variance. Adjusting for an index of deprivation and the supply of GPs gave a median 8% decrease in the sum of variance components. Conclusion: Geographical and organisational variation in the structure of primary care services should be considered in designing studies in health systems such as the English NHS. Stratified designs may be used to increase study efficiency, but variation between areas may sometimes compromise generalisability.