Comparison of patient and proxy responses on risk factors for stroke

Background - For studies on chronic risk factors and trigger (i.e. acute risk) factors, stroke researchers often have to rely on proxies. The reliability of proxy responses regarding trigger factors for stroke is unknown. Methods - Thirty patients with stroke and their proxies were interviewed about chronic risk factors and trigger factors. We assessed the completeness of proxy-derived data by calculating the level of non-response and the level of agreement using Cohen kappa statistics. Results - For most chronic risk factors and trigger factors, the response rate to whether or not exposure had taken place in the past year was 87% or higher. If couples agreed on exposure, patient and proxy could also provide a comparable estimate of the average frequency of exposure. Although the non-response on last time of exposure was higher, proxies who could answer provide a reasonably good estimate for most trigger factors. Conclusions - Proxies provide reliable information on exposure to chronic risk factors and trigger factors for stroke. For exposure and average frequency of exposure, non-response is low and the level of agreement is high for most chronic risk factors; for last time of exposure non-response is higher, but proxies who could respond provided reliable estimates of last time of exposure to most trigger factors.