Although all these registries have the common goal of improving the quality of stroke care, there is substantial variability between them in terms of their structure and organization. Only the Swedish RIKS-Stroke registry provides a complete census of all acute care hospitals in the country [34, 35]. All of the other registries include a sample of eligible hospitals (and therefore a sample of patients) but the number, type, and sampling mechanisms used to select them vary substantially. Thus the extent that these registries provide information that is generalizable to the total stroke population is somewhat questionable.

The funding sources also vary widely with most registries relying on federal and/or state/provincial level government support, with the exception being the US Get-With-The-Guidelines-Stroke (GWTG-Stroke) program, which relies heavily on pharmaceutical industry support through the American Heart Association (AHA) [36]. Other registries, such as the Australian Stroke Clinical Registry (AuSCR), have been supported by a combination of funding from nonprofit institutions, consumer donations, pharmaceutical industry, and grants [37].

All of these registries are able to report data on patient outcomes that occur during the in-hospital stay (e.g., complications, deaths) or at discharge (e.g., destination, functional status). Several registries such as the Swedish Stroke Register (RIKS-Stroke) and the Registry of the Canadian Stroke Network (RCSN) have been able to link the registry data to other data sources (including administrative, billing, vital statistics, and census data) to obtain data on longer term mortality, readmission, physician visits, and medication use [20, 24].

Some of the registries, for example RIKS-Stroke [35], AuSCR [38], SITS-MOST [25], and the UK Stroke Improvement National Audit (SINAP) [22], have been able to collect patient-reported outcome measures (PROM) following hospital discharge. This typically includes data on functional status (disability) and quality of life (QOL), and is most often collected 3 or 6 months after discharge. Follow-up data collection has been done using a variety of methods including face-to-face interviews [33], mailed questionnaires [39], or telephone interviews with either patients [23] or their caregivers [33]. Although patient-reported outcomes data are regarded as the gold standard of outcome measures, its collection represents considerable practical challenges in terms of the amount of resources (personnel time and costs) required and in obtaining individual patient consent [24].

Specific requirements of stroke registries to obtain informed consent from patients are highly variable and depend in part on the policies and preferences of local human subjects’ research oversight committees. The type of data collected also has a large influence on the need for consent; some registries have been required to obtain consent from patients prior to any data collection [23, 40], whereas others have been able to collect in-hospital data after obtaining a waiver of consent [21, 41, 42]. It is reasonable to expect that individual patient consent is required when registries collect post-discharge follow-up data directly from patients or caregivers, although confirmation of this from available published reports is often lacking [43, 44]. The requirement to obtain individual patient consent for follow-up of stroke patients post-discharge is one of the major reasons that data collection in US-based registries is limited to hospital discharge data only [21, 41].

A useful lesson on the pitfalls of requiring consent from patients comes from the RCSN, which has evolved in its approach to obtaining patient consent. During the first 2 years of the RCSN (Phase 1 and 2) informed consent was required on all patients even for a basic level of data abstraction. However the results indicated that substantial selection bias was introduced due to difficulty of obtaining consent. More than 50 % of eligible subjects were not consented either because of refusal or inability to contact the patient during the in-hospital stay [40]. Patients who were consented were different from those who were not consented in important ways; for example the in-hospital mortality rate was three times higher in patients who were not consented compared to those who were. Other factors such as age, level of consciousness on admission, race, preferred language, and length of stay were also different between consented and non-consented patients [40]. Subsequently in 2003, the registry was granted special status as a “prescribed registry” and the requirement for informed consent was dropped, thus allowing it to collect data on all patients without consent. This status was extended to also include the subsequent Ontario Stroke Registry (OSR) [24, 40]. The Australian registry (AuSCR) is an example of another approach to informed consent which utilizes an “opt-out” consent protocol to minimize dropouts that can occur when written informed consent is required. The “opt-out” consent protocol provides patients with information on the purpose of the registry, explanation of the data to be collected, and details of simple, cost-free ways available to exclude their data from the registry [23]. A similar opt-out approach is used in the UK, SINAP registry, where although a waiver of consent was obtained individual patients can request that their identifiable data not be included in the registry [42].

Perhaps the most liberal approach to consent is demonstrated by the Swedish RIKS-Stroke registry, which does not require informed consent from patients because quality monitoring by the Swedish healthcare system is mandated by law and so the registry is regarded as a component of the regular healthcare system rather than a research project [35]. This waiver of the need for consent even extends to the collection of patient data by survey or interview post-discharge. Despite many hospitals to provide patients with information about the purposes of the registry and the types of information collected, patients also are informed about an “opt-out” procedure should they wish to withdraw from participation [39].