Agreement between patient reports of cardiovascular disease and patient medical records

OBJECTIVE: To determine how well patient-reported cardiovascular disease (CVD) and CVD risk factor information agrees with medical record information. METHODS: Information from Patient/Family History (PFH) questionnaires completed between 1936 and 1999 by residents of Olmsted County, Minnesota, aged 20 years and older was compared with information available through the Mayo Clinic medical diagnostic index. Positive and negative agreement values were calculated by comparing agreement between the 2 data sources. Also, with the Mayo Medical Index serving as the criterion standard, sensitivity, specificity, and positive and negative predictive values of questionnaire information were calculated overall and by subgroups of sex, age, and years of education. RESULTS: Questionnaire responses were retrieved for 26,162 patients. Positive agreement values ranged from 31% for report of a medical problem or surgery related to arteries to the head, arms, or legs or the aorta to 78% for high blood pressure. Negative agreement values ranged from 30% for high cholesterol to 93% for medical problem or surgery related to arteries to the head, arms, or legs or the aorta. Sensitivity of questionnaire information ranged from 37% to 73%, whereas positive predictive values ranged from 27% to 83%. Positive agreement, sensitivity, and positive predictive values tended to increase with increasing age of the patient. Specificity and negative predictive values were 87% or greater, and negative agreement, specificity, and negative predictive values decreased with increasing age. CONCLUSION: Positive patient reports of CVD conditions and risk factors are relatively inaccurate. However, negative self-reports of these conditions are unlikely to be noted in the medical record. Further development of the PFH questionnaire is necessary to ensure accurate patient reporting of CVD data.