Agreement between GPRD smoking data: a survey of general practitioners and a population‐based survey

Abstract

Cigarette smoking is a common habit that is associated with many diseases. Smoking is often an important confounding variable in pharmacoepidemiological studies. The General Practice Research Database (GPRD) is widely used in pharmacoepidemiological research. In this study, we compare data recorded in the GPRD with the smoking history obtained from direct query of general practitioners (GPs) and from a population-based survey. We completed a mailed survey of GPs caring for a random sample of 150 patients with inflammatory bowel disease. The survey asked the GP to categorize the patients smoking status on a specified date. These results were then compared to the data recorded in the GPRD. Smoking status of 225,308 randomly selected GPRD patients without inflammatory bowel disease was compared to the results of a population-based household survey. Completed surveys with usable data were received from GPs on 136 of the 150 patients (91%). The sensitivity and positive predictive value of the database for current smoking were 78% (95% CI: 52-94) and 70% (95% CI: 46-88) respectively. The sensitivity and positive predictive value of former smoking were 53% (95% CI: 28-77) and 60% (95% CI: 32-84) respectively. Current and former smoking rates in the GPRD were 79% and 29% respectively of expected rates according to the population-based survey. Current smoking is more completely recorded in the GPRD than former smoking. These data need to be considered when planning GPRD studies where smoking is an important exposure variable.