Abstract
PURPOSE: Self-report or electronic health records can be used to calculate number of chronic conditions for study participants. Although agreement for specific conditions can be found in the literature, there is a lack of information on how the total number of conditions compares between the two sources. METHODS: Using a long-standing cohort study, the number of chronic conditions was estimated for 351 participants using self-report and data from their electronic health record. Agreement for each condition, and for total number of conditions was estimated using a weighted kappa with 95 % Confidence Interval (CI), and using Lin's concordance coefficient, and a Bland-Altman plot. Predictors of discord were identified using generalized linear models with concordant pairs as reference. RESULTS: Discord varied by condition but ranged from 3.7 % (diabetes, kappa=0.86) to 33.2 % (hyperlipidemia, kappa=0.31). While the mean number of chronic conditions was similar, it showed poor agreement [weighted kappa= 0.46 (95 % CI: 0.41, 0.52)], and did not show pattern of discrepancy as the average number of chronic conditions increased. Discrepancies were related to condition, male sex, and lower education. CONCLUSION: There are opportunities to ensure that electronic health records are complete, and that patients receive adequate health education to become aware of the conditions that they have.