Supplementary Material for: Role of Proxy Respondents in International Stroke Research: Experience of the INTERSTROKE Study

Introduction
Measuring patient-reported information in stroke research is challenging. To overcome this, use of proxy respondents is often a necessary strategy. In this study, we report on use and effect of proxy respondents on patient case-mix in a large international epidemiologic stroke study (INTERSTROKE).

Methods
This was a cross-sectional study of 13,458 cases of acute first stroke in 32 countries. A standardised study questionnaire recording behavioural cardiovascular risk factors was administered to the patient, and if unable to communicate adequately, a valid proxy, or both. We used logistic regression to evaluate the association of age, sex, education, occupation, stroke severity and region with need for proxy respondent, and report odds ratio (OR) with 95%CI.
Results
Among 13,458 participants with acute stroke, questionnaires were completed by patients alone in 41.4% (n=5573), combination of patient and proxy together in 21.7% (n=2,918) and proxy alone in 36.9% (n=4967). Use of proxy alone was greater in participants with severe stroke (4.7% with modified-Rankin score of 0 versus 80.5% in those with score 5; OR 187.13; 95% CI 119.61-308.22) older persons (43.8% of those aged 80 years and over versus 33.2% of those aged less than 40 years; age per decade OR 1.09; 95% CI 1.06-1.12), women (40.7% versus 34.3% of men; OR 1.32 95% CI 1.22-1.43); and those less educated (58.9% of those never educated versus 25.7% of those who attended third level education; OR 7.84; 95% CI 6.78-9.08).
Conclusion
Use of proxy respondents enhances the generalisability of international research studies of stroke, by increasing representation of women, patients with severe stroke, older age, and lower education.