Details of Publication 650 for Project A434:

Jackson CA, Mishra G, Byles J, Tooth L, Dobson A. Agreement between self-reported stroke and hospital-recorded stroke in two cohorts of Australian women. , 2013; : 35(S3) 142

Background: There is uncertainty around using self-reported stroke in epidemiological studies, with conflicting recommendations from prior studies. We compared self-reported with hospital-recorded stroke, and investigated reasons for disagreement. Methods: We included women from the Australian Longitudinal Study on Women’s Health, who lived in New South Wales and returned 3-yearly surveys between 2001-10 (mid-age cohort; N= 2119, mean age 53) or 2002-08 (older cohort; N = 1556, mean age 78). We determined agreement between self-report and hospitalised stroke by calculating sensitivity, specificity, positive and negative predictive values (PPVs, NPVs) and kappa statistics, and determined factors associated with disagreement.Results: Women reported 102 and 23 strokes in the older and mid-age cohorts respectively. Sensitivity and specificity were high. Few women failed to report hospital-recorded strokes (NPV>99% in both cohorts) but many reported an unverified stroke, particularly in the older cohort where the PPV was 25% compared with 44% in the mid-age cohort. This was partly due to: reporting of transient ischaemic attack (TIA); strokes occurring outside the period of interest; and possible reporting of stroke-like events. A third of older women with an unverified stroke did have a cerebrovascular disease diagnosis (including TIAs) and 18% provided additional survey comments with details of the stroke. Agreement was therefore fair in older women (kappa 0.35, 95% CI 0.25 to 0.46) and moderate in mid-age women (0.56, 95% CI 0.37 to 0.75). In both cohorts higher education was associated with agreement. Recent poor mental health was associated with disagreement. Conclusion: Validity of self-reported stroke is fair to moderate in our study, but is probably underestimated due to comparison with hospital data only. Where routinely collected data are unavailable, self-reported stroke may be a reasonable alternative method of ascertainment for some epidemiological studies.