REPRODUCIBILITY OF DATA COLLECTED BY PATIENT INTERVIEW

This study examined the reproducibility of data, not generally conside red at risk. for poor patient recall, obtained on two separate occasio ns. Our study used data collected for a register of heart attacks in t he Lower Hunter Region of New South Wales, and included 1675 patients who were registered at least twice. Reporting inconsistencies between occasions were assessed for eight data items. We found that the sex of five patients had been recorded differently on the two occasions. Amo ng patients interviewed on both occasions, between 0.5 per cent and 2. 0 per cent of patients had inconsistent reports for marital status, co untry of birth, smoking status and height, 2.7 per cent for date of bi rth, 13 per cent for education level and between 1.6 per cent and 9.6 per cent for the history of various medical conditions. Patients not f rom an English-speaking background, over 60 years of age or without te rtiary education tended to have higher rates of inconsistent reporting . Time between occasions, marital status and sex were not associated w ith increased rates of inconsistent reporting. We concluded that appar ently straightforward data items, such as date of birth and education level, were not perfectly reproducible when obtained by patient interv iew on separate occasions. Our results provide a starting point for se nsitivity analysis in other studies if the potential inaccuracies in r eporting of such data should be of concern.