Administrative health data are generated through the routine delivery of healthcare programs and services [1]. These data are primarily collected and curated to manage health care resources including the evaluation and planning of hospital bed numbers, occupancy and budgets, and payment of physician and drug dispensation claims [2, 3]. However, these data have also come to be recognized as valuable for secondary uses, including clinical, health services, and population health research [4–6].

Administrative health data are typically population-based, capture a wide range of variables, contain data captured over long periods of time, and contain unique individual-level identifiers, making them well-suited to use in health research [5, 7, 8]. The benefits of using administrative health data for research purposes include improved statistical power and reduced non-response or participant bias [9], the opportunity to examine temporal trends, and lower costs compared to primary data collection [7, 8, 10]. Another benefit of using administrative health data for research is the potential for data linkage, which has been defined as “the bringing together from two or more different sources, data that relate to the same individual, family, place or event” [7] (p.767). The linkage of patient-level data contained in multiple databases provides a more comprehensive view of their encounters with the healthcare system than would be possible if a single database was used. Where individual-level identifiers are available, administrative health databases can also be linked to other related databases (e.g. judicial, educational, social services) to study the social determinants of health behaviors and outcomes [5, 8].

Canada has a long history of administrative health data collection and is considered to have some of the most comprehensive administrative health data in the world [2, 3, 11]. While there is some variation in the specific data that are collected across provinces/territories, data that are commonly collected include inpatient hospitalizations, day surgeries, physician billing claims, and prescription medications [2, 3]. Since these are collected at the provincial/territorial level, the majority of provinces have established data repositories that contain various administrative health databases that may be accessed by researchers [12–18].

Despite having what has been deemed an “information rich” environment [19], infrastructure and resources dedicated to facilitating access to administrative health data (i.e. repositories and staff), and a regulatory framework (i.e. ethical guidelines and information legislation) that permits access to health information for research purposes [20–30], there is evidence to suggest that researchers across Canada are not able to access these data in a timely manner [5, 31–38]. Several groups of researchers who have undertaken multi-province studies using linked administrative health data have described delays and substantial interprovincial variations in the timeliness of data access approvals and receipt of data [34–37]. In a 2002 report, Kephart [34] described a multi-province study involving administrative health data where one province was ultimately excluded due to an inability to obtain data access approval after two years. Since then, similar accounts have continued to appear in the literature [34–37]. In 2016, a multi-site study conducted by the Canadian Network for Observational Drug Effect Studies (CNODES) reported that the British Columbia site was not included due to “lengthy timelines” for data access [35]. Elsewhere, variations in the time to obtain required data access approvals across provinces were reported to impact overall project timelines and hinder the ability for work to be carried out in parallel across provinces [36, 37].

While interprovincial variations in the timeliness of data access may be expected as a result of differences in data access processes, in some cases these variations may be indicative of barriers to timely data access. Currently, the extent to which researchers are experiencing barriers when attempting to gain access to administrative health data in Canada is not well understood. In a 2015 report sponsored by the Canadian Institutes of Health Research (CIHR), a broad range of potential barriers to timely access to health and health-related data were identified; however, minimal evidence was provided to establish the impact of these barriers on data access timelines in Canada, or that researchers in Canada were in fact experiencing barriers to timely data access [5]. More recently, a quality improvement study focused on examining local data access and usage practices was carried out at a single institution in British Columbia [38]. Through interviews with researchers, “lengthy turnaround times” were identified as the primary barrier to accessing and using data for research, although the factors contributing to these lengthy turnaround times were not addressed. Notably, only one study [32] has examined timely access to administrative health data specifically. This study reported wide variations in data access timelines across provinces (from 1 to 18 months) and identified a number of barriers and enablers to timely access. However, there were several important limitations: timeliness data were self-reported by provincial agencies and not confirmed by other means; the starting point for measuring timeliness varied across agencies; and no data were obtained from researchers regarding the experiences with accessing data. Moreover, given that this study was published in 2013, the findings may be of limited applicability to the current Canadian context. Robust evidence is therefore required to assess the extent to which researchers in Canada are experiencing barriers when attempting to access administrative health and the impacts of these on researchers’ ability to access data, including the timeliness of data access.

The overall aims of this study are to identify the factors (i.e. barriers and enablers) to accessing administrative health data for research within selected Canadian provinces, and to examine how variations in these factors across provinces contribute to reported interprovincial variations in data access timelines across Canada. The specific objectives are to:

This study will use case study methodology, underpinned by pragmatism. Case study methodology is particularly useful in situations wherein the phenomena being studied cannot be separated from its context [39]. This is relevant to the current study, wherein the factors affecting access to administrative health data for research purposes (i.e. the phenomena) can only be understood in terms of the interactions of the actors involved in research and research oversight, the organizations in which they are situated, and the broader regulatory landscape in which they operate.

There are different approaches to case study, including those published by Stake [40–42], Merriam [43, 44], and Yin [39, 45–47]. All three approaches focus on studying a case, or cases, within a real-life context through the use of multiple data sources, though each proposes different strategies for doing so. Consistent with the pragmatic worldview underpinning this study, we will use a case study approach informed by the works of all three authors, though Merriam [43, 44] and Yin [39, 45–47] will feature more prominently. Since pragmatism embraces both the objective and subjective [48], researchers are not tied to a specific set of methods based on their ontological assumptions, and free to use whichever method (or methods) that are best suited to answering a research question [48].

This study will employ a multiple case study design (also referred to as “multi-site” [44] or “collective” [40] case study) with embedded units of analysis as set out by Yin [39] (Figure 1). We will use a multiple case study design to enable the examination of variations in data access across jurisdictions, with each of the selected cases representing an example of the social phenomenon being studied [47]. Within each case, there will be two embedded units of analysis, which represent stakeholder groups of particular interest: 1) researchers and research staff who have accessed/sought access to administrative health data, and 2) individuals involved in the regulation and oversight of data access. These two groups differ in terms of their role/mandate, training, interests at stake, incentives, and how their positions within their affiliate institutions. As such, each will have unique insights into the factors affecting access to administrative health data.

Three cases will be included in this study, with each representing a “research system”. Each research system will encompass a provincial health data repository, as well as the two stakeholder groups described above. To enable the examination of contextual factors, including variations in provincial legislation, cases will be identified in three different provinces. As per guidance from Yin [39], Stake [40], and Merriam [44], cases will be selected in order to maximize the knowledge gained [39, 40, 44]. This means taking into account not only what can potentially be learned from them, but the extent to which they are accessible [40]. More specifically, cases will be chosen based on:

Case study research is characterized by the use of multiple data collection sources and methods [39, 40, 44, 51]. In this study, interviews will be the primary source of data, providing important historical and contextual information relevant to a case, explaining events and behaviours, and understanding participants’ opinions and attitudes [39, 44]. We will also use documentary evidence to provide historical and contextual information relevant to the case that may not otherwise be obtained [26, 30], and to confirm, corroborate, and supplement information gained from interviews [39].

For each case, we will carry out interviews with key informants (i.e. individuals who have specialist knowledge relevant to the case or specific aspects of the case [52]). Key informants will include members of the two aforementioned stakeholder groups: (1) researchers and research staff with experience accessing administrative health data for research purposes, and (2) those involved in the regulation and oversight of data access.

Interview format and the development of interview guides will be informed by the work of Patton [50] and Rubin and Rubin [53]. Thus, interviews will use a semi-structured approach, taking on the form of a guided conversation rather than a structured interview [39]. Each stakeholder group will be asked a common series of questions related to their own role and experience as well as the facts of the case (i.e. processes for accessing data, required approvals and documentation, actors involved, and relevant policies and legislation), along with a series of questions specific to each group. Researchers and research staff will be asked questions focusing on their experiences when attempting to access administrative health data for research purposes (Supplementary Table 1), while those involved in regulation and oversight of data access will be asked questions focusing on the use and regulation of administrative health data for research purposes (Supplementary Table 2).

Interviews will be carried out in person or via telephone and will be approximately one hour in length. Interviews will be audio-recorded and transcribed verbatim for analysis. Drawing on guidance from Stake [40], notes will be taken throughout interviews to capture impressions, particularly useful quotations, key ideas or concepts to be explored in subsequent interviews, questions that need to be revised or reframed, and to capture recommendations of other sources of evidence to be included or other key informants to be invited to participate.

Documents will be obtained from online sources and requested from key informants. Access to a wide array of documents relevant to each case will be sought, including: administrative documents (e.g. agendas and minutes from research ethics board (REB) and data access committee meetings); policy documents (e.g. provincial legislation and regulations relevant to research involving administrative health data, institutional policies and guidance documents for researchers and oversight bodies); data access documents (e.g. data access forms, data sharing agreements); and evaluations of the provincial health information legislation, and provincial or institutional research reports.

For interviews, we will recruit key informants from two stakeholder groups in each province:

We will identify participants using two strategies. First, we will use purposive sampling to ensure that study participants include those individuals from whom the most can be learned (i.e. key informants) [50]. Based on the research team’s knowledge of the research and regulatory landscapes in Canada, as well as information obtained from online searches and relevant published literature and reports, the team will compile a preliminary list of potential participants. As individuals are recruited and interviewed, we will also use snowball sampling [50], whereby participating key informants will be asked to identify other potential participants in their province.

Individuals who are selected for study inclusion will receive an email invitation to participate. A modified Dillman method [54] will be used to increase response rates. A consent form will be sent to all participants several days prior to the interview, and written consent will be obtained prior to the start of the interview. For each case, we will aim to recruit 10 key informants from each stakeholder group, totaling 20 interviews per case and 60 interviews for the entire study (10 individuals/group * 2 groups/province * 3 provinces = 60). For each case, recruitment will stop once theoretical saturation is reached [49, 55, 56].

As is common in qualitative research, data collection and analysis will occur concurrently [44]. For analysis, we will use an approach consistent with that of Merriam [44]. Since we are using a multiple case study design, analysis will occur on two “levels”. Each case will first be analyzed separately as a standalone study (i.e. “within-case analysis”), followed by the comparison of findings across cases (i.e. “cross-case comparison”).

A limitation of this study is the lack of comprehensive quantitative data to examine data access timelines in each province. Although provincial data repositories may capture or be able to calculate the time from the receipt of a researcher’s application to the date approval is granted, this represents only a portion of the total data access timeline, excluding other aspects of the overall process of data access (e.g. application preparation, obtaining REB approval, contracts and data sharing agreements, dataset preparation, etc.), the time to access data held by other data holders, as well as applications that remain incomplete. To obtain complete timeliness data, the research team would need to request detailed information from researchers and all relevant data holders in an attempt to assemble data access timelines for individual studies. This resource intensive process could only be undertaken for a small sample of studies in each province, which would ultimately provide limited insights into the overall timeliness of data access in a province, or into the factors affecting data access. With that being said, the primary focus of this study is not to measure timeliness, but rather to gain an understanding of the contexts and conditions under which researchers are able to gain timely access to administrative health data for research, which case study methodology is well-suited to address.

The inclusion of a limited number of cases may also be perceived as a study limitation due to lack of generalizable findings [65], however, the traditional notion of generalizability (i.e. “statistical generalisation” [66]) is not an appropriate criterion for assessing the value of case study research. Instead, the aim of this study is “transferability” [44] (also referred to as “naturalistic generalization” or “case-to-case transfer”[66]), which occurs when findings observed in one context can be applied in a different context. The onus is on the reader to determine the extent to which the findings from a particular case or cases may apply in another context, while it is the responsibility of the researcher to provide sufficient detail to enable this process [44]. In this study, transferability of findings will be facilitated through the process of “particularization” (i.e. by focusing on the details and uniqueness of each of the selected cases [40, 67]) and the use of “thick description” (i.e. rich, literal descriptions of each case and its context) to describe the context in which each case is embedded [44].