Background: General practitioner (GP) records are increasingly being used as sources of information on potential confounders such as smoking use and alcohol intake in epidemiological studies. The aim of this study was to assess the accuracy of GP records on smoking use and alcohol intake compared with data from patient questionnaires.
Methods: Patients registered with 42 practices in Oxfordshire that agreed to take part in a post-marketing surveillance study of omeprazole were sent a postal questionnaire that included questions about alcohol and tobacco use. Two years later, data on these aspects of lifestyle were abstracted from the GP records.
Results: A total of 892 patients agreed to take part in the study; 804 (90 per cent) completed the postal questionnaire, and the records of 856 (96 per cent) were reviewed. Information on smoking and alcohol use was present in 74 per cent and 63 per cent of GP records, respectively. Agreement between the two data sources was moderate for both smoking (kappa = 0.50) and alcohol use (kappa = 0.52). With regard to smoking, the main discrepancy between the two data sources was that 46 per cent (94/206) of patients who reported themselves as exsmokers were recorded as being never smokers in the GP record. With regard to alcohol, there were no systematic differences between the two data sources.
Conclusion: Data from GP records on smoking status and alcohol use are incomplete and subject to some misclassification. This is a source of potential failed adjustment for confounding, which should be considered in epidemiological studies that make use of these records.