Purpose: Various data sources may be used in pharmacoepidemiological studies. When they cannot be obtained from valid databases, medical data must be obtained from physicians or patients. In the CADEUS study, both patients and their prescribers reported medical data allowing investigation of the concordance between these sources.
Methods: CADEUS is a French national cohort study of traditional non-steroidal anti-inflammatory drug (NSAID) and coxib users conducted between September 2003 and August 2004 in France that employed self-administered questionnaires to obtain medical data from patients and their prescribers. The Kappa statistic (kappa) was used to measure concordance between patients and prescribers in 18 530 pairs with regard to previous medical history and index NSAID indication.
Results: For previous medical history, the proportion of agreement ranged from 70.7 to 99.2% and concordance was: substantial (kappa = 0.61-0.80) for hypertension, myocardial infarction, stroke and diabetes; moderate (kappa = 0.41-0.60) for angina pectoris, peripheral arterial disease and hypercholesterolaemia; fair (kappa = 0.21-0.40) for unstable angina, cardiac insufficiency, dyspepsia, gastroesophageal reflux and gastric ulcer; slight (kappa < 0.21) for upper gastrointestinal haemorrhage. For index NSAID indication, the proportion of agreement ranged from 84.3 to 99.4% and concordance was almost perfect (kappa = 0.81-1.00) for inflammatory rheumatism, flu-like symptoms, dysmenorrhoea and dental pain; substantial for arthritis, back pain and headache; moderate for osteoarticular pain.
Conclusions: Concordance was better for specific or serious conditions both regarding previous medical history and indication. Prescriber or patient perception and understanding may reduce concordance for certain items.
(c) 2010 John Wiley & Sons, Ltd.