Improving inter-rater reliability by coding scheme reorganization : managing signs and symptoms

• The aim of this paper is to study the potential for improving inter-rater reliability in general practice registries through the use of a semantic terminology model that enables diagnostic labels to be separated into symptoms and diseases, i.e. into different levels of diagnostic precision. Cases coded as symptoms according to the ICD-based coding system currently in use in Swedish general practice were reclassified with the help of the model, and inter-rater variability was measured through divergences of observed coding distributions from expected distributions. 40 percent of the symptom cases were candidates for reclassification; half of these could actually be reclassified. This decreased inter-rater variability, but the difference was not statistically significant. Diagnostic categories with large variation in utilization rates were foWld, which calls for careful selection of topics for medical audit. Although reclassification of symptoms may improve reliability, no straightforward association was found between a chapter's diagnostic precision and its contribution to overall variability. Nor could differences in diagnostic precision explain all variation within a chapter. Further research on other dimensions of the coding system is needed before symptom reclassification can be recommended as a general reliability-improving tool.

Nyckelord

Statistics

Data collection

Family practice

Observer variation

Signs and symptoms

Publikations- och innehållstyp

vet (ämneskategori)

ovr (ämneskategori)