A comparison of self-reported and record-linked blood donation history in an Australian cohort

Reference details

Bertalli NAA, K. J.; McLaren, C. E.; Turkovic, L.; Osborne, N. J.; Constantine, C. C.; Delatycki, M. B.; English, D. R.; Giles, G. G.; Hopper, J. L.; Anderson, G. J.; Olynyk, J. K.; Powell, L. W.; Gurrin, L. C. (2011) A comparison of self-reported and record-linked blood donation history in an Australian cohort. Transfusion 51:2189-2198


BACKGROUND: Questionnaire-based studies investigating blood donation history rely on the accurate recall of information from participants for results to be valid. This study aimed to retrieve electronic records from a national blood donation service and link them to self-reported history of donation to assess agreement between the two sources. STUDY DESIGN AND METHODS: Between 2004 and 2006, a sample of participants of northern European descent was selected from the Melbourne Collaborative Cohort Study (n = 31,192) to participate in the "HealthIron" study (n = 1438). A total of 1052 participants completed questionnaires that included questions about blood donation history. In 2009, consenting participants' records were linked to the Australian Red Cross Blood Service (ARCBS) to provide information on blood donations made between 1980 and follow-up (2004-2006). Those who commenced blood donation before 1980 were excluded. RESULTS: A total of 718 participants were available for analysis. Of these, 394 (55%) provided signed consent, including 182 (82%) of the 227 participants who self-reported ever donating blood. The two data sources were concordant for 331 (87%) of participants, with a kappa statistic of 0.74 (SE, 0.05) indicating a high level of agreement. Participants tended to overstate by a factor of 2.0 (95% confidence interval, 1.7-2.2) the number of donations they had made when compared with ARCBS records. CONCLUSION: Participants in studies assessing self-reported blood donation history are likely to correctly indicate whether or not they have ever donated blood. Quantitative estimates are potentially inaccurate and could benefit from validating a sample of records to quantify the bias.

LinkPubMed Link

You are not logged in.