Variation in clinical coding lists in UK general practice: a barrier to consistent data entry?

BACKGROUND: Routinely collected general practice computer data are used for quality improvement; poor data quality including inconsistent coding can reduce their usefulness.
OBJECTIVE: To document the diversity of data entry systems currently in use in UK general practice and highlight possible implications for data quality.
METHOD: General practice volunteers provided screen shots of the clinical coding screen they would use to code a diagnosis or problem title in the clinical consultation. The six clinical conditions examined were: depression, cystitis, type 2 diabetes mellitus, sore throat, tired all the time, and myocardial infarction. We looked at the picking lists generated for these problem titles in EMIS, IPS, GPASS and iSOFT general practice clinical computer systems, using the Triset browser as a gold standard for comparison.
RESULTS: A mean of 19.3 codes is offered in the picking list after entering a diagnosis or problem title. EMIS produced the longest picking lists and GPASS the shortest, with a mean number of choices of 35.2 and 12.7, respectively. Approximately three-quarters (73.5%) of codes are diagnoses, one-eighth (12.5%) symptom codes, and the remainder come from a range of Read chapters. There was no readily detectable consistent order in which codes were displayed. Velocity coding, whereby commonly-used codes are placed higher in the picking list, results in variation between practices even where they have the same brand of computer system.
CONCLUSIONS: Current systems for clinical coding promote diversity rather than consistency of clinical coding. As the UK moves towards an integrated health IT system consistency of coding will become more important. A standardised, limited list of codes for primary care might help address this need.