Objectives General practitioners are increasingly required to practice in a paperless environment and to collect clinical data electronically on electronic patient record (EPR) systems. A principal step in meeting general practice information needs continues to be the establishment of disease registers and consequently the identification of patient populations within primary care databases is a prerequisite. This study aims to identify and validate the optimal search strategy for coronary heart disease (CHD). Methods A multiple logistic regression model for the identification of CHD patients was developed in one site using electronic data, the receiver operating characteristic (ROC) curve and Bayesian statistics. The model was tested on two trial sites. Results Young male CHD patients are more easily identified by generic searches than older females. The optimal search strategy for CHD was found to be the diagnostic code for CHD, nitrate and digoxin but this was dependent on the disease description, age and sex of the study population and the coding system used within the database. Diagnostic code for CHD identified 80.6% (95% confidence interval (CI) 77-83%), 90.0% (CI 88-92%) and 95.9% (CI 94-97%) of local, national and international definitions respectively, with 100% positive predictive values (PPVs) for all definitions. Conclusion Generic queries may inadvertently perpetuate inequalities in health care. Queries should be bespoke and mindful of the conceptualisation of disease by the clinicians recording these data.
|Number of pages||10|
|Journal||Informatics in Primary Care|
|Publication status||Published - 1 Jan 2009|