THE PROBLEM OF DIAGNOSTIC VARIABILITY IN GENERAL-PRACTICE

Study objective-The aim was to examine the scale, source, and relevance of variation between general practices in respect of the rates with which patients consulted with illnesses falling in each of several diagnostic groups. Design-This study involved a general practice morbidity survey conducted over two years, 1970-72. All patients who consulted their general practitioners were identified and the number of these who consulted with diagnoses attributable to each of the 18 main chapters of the International classification of diseases were counted. Patients who consulted for more than one diagnosis within a chapter were counted once only; those who consulted for one or more diagnoses in each of several chapters were counted once for each chapter. Setting-This was a national survey involving general practitioners in England and Wales. Subjects-The study involved 214 524 patients from 53 selected general practices (115 doctors) who were registered with their general practitioners for the whole of the year 1970-71 and for whom their morbidity data had been linked with their social data from the 1971 census. Measurements and main results-Using the numbers of patients on the practice lists as denominators, practice patient consulting rates (PPCR) were calculated for each practice and for each ICD chapter. Variability in chapter PPCR was examined by calculating coefficients of variation and, after allowance for random variation, coefficients of residual variation. There were large interpractice (doctor) variations in all chapter rates. These variations were only marginally attributable to: chance; different age, sex and social class mixes of practice populations; geographical locations; and practice organisation. The rates were, however, consistent from one year to the next for any one practice. Approximately half of the interpractice (doctor) diagnostic variability was associated with overall patient consulting behaviour. When the effects of this behaviour were discounted, any major residual diagnostic variability was confined largely to ICD chapters I-V, XVI, and XVII, ie, those chapters where aetiology forms the basis of classification. Conclusion-Variations in recorded diagnostic rates are mainly due to the consistent but idiosyncratic and selective exclusion by practitioners of some components from the total set which often coexist in a new diagnosis. Because of the scale of interpractice diagnostic variability, the use of algorithms and information technology is largely precluded from outcome studies, auditing procedures, and studies of practice work loads in general. However, (1) the consistency of any individual doctor's pattern of diagnostic recording from one year to another permits studies of trends; and (2) given a reasonable number of recording practices, the population mean practice consulting rates can be estimated with sufficient accuracy for many epidemiological research and administrative uses.