Accuracy of recall of hip fracture, heart attack, and cancer: a comparison of postal survey data and medical records

Abstract

Accuracy of recall of hip fracture, heart attack, and cancer was evaluated by comparing data obtained from postal surveys and medical records of participants in the Leisure World Disease Prevention Study. The study cohort comprised 13,897 residents of Leisure World Laguna Hills who responded to a questionnaire first mailed in 1981. Follow-up questionnaires requesting information on recent disease history were mailed in 1983 and 1985 to all living cohort members and were returned by 9,734 (86%) and 8,884 (82%) persons, respectively. The proportion of false-positive answers was about 40% for acute myocardial infarction, but less than 10% for hip fracture and cancer. Many of those who gave a false-positive answer on the questionnaire had a condition that the lay population might easily confuse with the condition of interest (e.g., heart attack vs. other cardiovascular disease). Among persons with the disease of interest recorded in the medical record, the percentage who failed to report their events on the questionnaires was highest for those who had an acute myocardial infarction, followed by those who had cancer and those who had hip fracture. The highest rates of confirmation of cancer were for cancers of the breast, bladder, prostate, and uterus. For closely related sites (e.g., colon and rectum), the reliability of reports decreased. The amount of information obtained varied according to the wording of the questionnaire and the specificity of the questions asked. Improved accuracy of recall was observed when the occurrence of specific diseases was asked in addition to a general question about recent hospitalizations.