Variation in recruitment across sites in a consent-based clinical data registry: lessons from the Canadian Stroke Network
1 Centre for Evaluation of Medicines, St. Joseph's Healthcare, Department of Clinical Epidemiology & Biostatistics, McMaster University Hamilton, Ontario, Canada
2 Department of Medicine, University of Toronto, Institute for Clinical Evaluative Sciences, Division of General Internal Medicine and Clinical Epidemiology, University Health Network, Toronto, Ontario, Canada
3 Centre for Bioethics, Clinical Research Institute of Montreal (IRCM), Centre francophone d'informatisation des organisations (CEFRIO), Montreal, Quebec, Canada
4 Institute for Clinical Evaluative Sciences, Toronto, Ontario, Canada
5 Division of Neurology, Department of Medicine, University of Toronto, Toronto, Ontario, Canada
BMC Medical Ethics 2006, 7:6 doi:10.1186/1472-6939-7-6Published: 23 May 2006
In earlier work, we found important selection biases when we tried to obtain consent for participation in a national stroke registry. Recognizing that not all registries will be exempt from requiring consent for participation, we examine here in greater depth the reasons for the poor accrual of patients from a systems perspective with a view to obtaining as representative sample as possible.
We determined the percent of eligible patients who were approached to participate and, among those approached, the percent who actually consented to participate. In addition we examined the reasons why people were not approached or did not consent and the variation across sites in the percent of patients approached and consented. We also considered site variation in restrictions on the accrual and data collection process imposed by either the local research ethics board or the hospital.
Seventy percent of stroke patients were approached, with wide variations in approach rates across sites (from: 41% to 86%), and considerable inter-site variation in hospital policies governing patient accrual. Chief reasons for not approaching were discharge or death before being approached for consent. Seventeen percent of those approached refused to participate (range: 5% to 75%). Finally, 11% of those approached did not participate due to language or communication difficulties.
We found wide variation in approach and agree rates across sites that were accounted for, in part, by different approaches to accrual and idiosyncratic policies of the hospitals. This wide variation in approach and agree rates raises important challenges for research ethics boards and data protection authorities in determining when to waive consent requirements, when to press for increased quality control, when to permit local adaptation of the consent process, and when to permit alternatives to individual express consent. We offer several suggestions for those registries that require consent for participation.