Strengths and Weaknesses of the General Practice Research Database

STRENGTHS AND WEAKNESSES OF THE GENERAL PRACTICE RESEARCH DATABASE

The inherent strengths in the GPRD stem mainly from the NHS system of health care delivery, which essentially provides single UK cradle to grave healthcare delivery. General practitioners are the central health care providers in the United Kingdom, and thus, GPs have longitudinal medical records for their patients. The data subjects included in GPRD are broadly repre-sentative of the UK population as a whole with respect to age, sex, socio-economic class and UK region. Unlike databases based on health insurance claims, the GPRD includes a relatively stable population with good information on start and stop of data collec-tion. The possibility to obtain further information from GPs and validate computerised information and to collect prospective data and samples is an additional major strength of this data set. The planned external record linkages will further enhance the utility of the database. The GPRD is now used by several regula-tory authorities and numerous pharmaceutical compa-nies and as of the end of 2005 is available through a collaboration with the Medical Research Council (MRC) to UK academics.

Some of the traditional weaknesses of the GPRD have been associated with the level of completeness of data recording due to the way data were transferred between secondary and primary care. This weakness is rapidly diminishing due to massive IT changes in the UK NHS. Drag and drop data entry into a patients record is now becoming the norm for laboratory data as well as hospitalisations. In the future, it will become fully automated.

A limitation of the GPRD, in the same way as for most databases, is that the information on factors such as over-the-counter medication, diet and exercise is limited. Also, detailed information on disease severity may not always be available or may not be recorded in a routine and standardised manner. Another challenge is for researchers to understand the complexity of this data set and to take into account the huge variability of patient characteristics and drug use. Association is not causation, and simplification of analysis can mean complication of result.