In recent decades, the potential benefits of secondary use of health data have been highlighted in many different contexts, including health systems governance, quality improvement [1], point-of-care decision-making [2-4], and research [5]. While a unique, unanimously accepted definition of secondary use remains elusive, we will operate under the premise that it covers any use beyond the initial intent of data collection. Given the resources required to generate health data, using them to their fullest extent to improve health is a laudable goal. Nevertheless, health data often describe an individual, and thus their autonomy regarding the use of data also needs to be considered [6]. For individuals to be able to weigh in on the use of data about them, they at least need to be aware that data generated for a primary reason (eg, during a consultation with a physician) have values above and beyond the specific context for which they were created and can therefore be used for secondary goals such as research projects or systems policy research. As mentioned previously, the secondary use of health data covers a broad landscape, and the scale of expected benefits (eg, How much? How soon? Who?) varies from one use to the other, which could modulate how individuals react to different types of secondary use.

Therefore, while the potential of secondary use of health data for research purposes is seen as promising to many stakeholders in the academic and health care worlds, trust regarding data sharing in novel contexts should not be considered as a given [7]. Indeed, despite substantial evidence that most individuals support the secondary use of health data, this support is neither unanimous nor unconditional [6,8-11]. Stakeholders including patients and members of the public express different needs that touch upon trust and agency [12]. Controversial initiatives such as care data in England [13], the National Electronic Health Record System in Australia [14], and other incidents involving unconsented and unauthorized use of data [15-17] have contributed to some mistrust in the population, which may then tend to develop greater sensitivity to the risks associated with data sharing and health data in general [7,12]. In addition, few people seem to be aware that the secondary use of their health data is already permissible in certain circumstances [18]. In all these situations, some stakeholders perceived a lack of transparency. Although transparency is unquestionably identified as an essential element of social license regarding the secondary use of health data, it is not always clearly defined. For example, Kisekka et al [19] showed that information accuracy and availability are important parameters for the adoption of e-portals; however, an open question is which kind of information is expected by the stakeholders and under which modalities of communication. Furthermore, technological advances offer new ways for transparency to become tangible and bring forth strategies for more open communication [20]. Therefore, there is a real need to better understand the actual expectations regarding transparency with regard to secondary use of health data.

The fact that a research project is targeting a common health issue such as heart attacks does not guarantee a rapid and substantial benefit as an important justification for valuing the secondary use of health data. For example, in 1982, studies identified a class of medication, namely β-blockers, as useful in reducing mortality in the context of heart attacks. Nevertheless, it took more than 20 years to confirm the widespread adoption of this simple and relatively inexpensive intervention [21].

To increase the pertinence of projects undertaken, the rapidity at which the resulting new knowledge is put into action, and therefore the benefits to patients, learning health systems (LHSs) have been proposed. They are a model that embodies how linking data from clinical care and research can create new knowledge and improve clinical decision-making [22]. Recent publications have reported the success of early LHSs [1,23,24], which is encouraging. Moreover, this integration of care delivery, research, and effective knowledge transfer can be implemented as described in the study by Faden et al in 2013 [25], in which a new ethical contract is based on a large engagement of all stakeholders (including patients), transparency (both on ongoing activities and data supporting them), and accountability to demonstrate benefits for the participants. While communicating about secondary data use is important, it is particularly essential in an LHS context.

This scoping review aims to describe how transparency is portrayed by various stakeholders regarding the secondary use of health data in a research context. Specifically, we wanted to address the following questions: (1) What information do patients and members of the public want to learn about the secondary use of their health data for research purposes (what)? (2) How do patients and members of the public want to interact with this information (how)? and (3) In which situations (when)? The objectives of the scoping review were elaborated by a team composed of an expert in health informatics (J-FE), an expert in research ethics (AC), a legal expert (J-FM), and an expert in ethics and philosophy of science (AB).

The study aims to capture a broad landscape of the literature to serve as a cornerstone of the upcoming consultations with patients and members of the public to identify the operational requirements of a transparency portal as LHSs are deployed in the province of Quebec. Therefore, resources specifically targeting an LHS context were identified when available, but the search was not restricted to only these as the LHS field is relatively young.

We addressed our research objectives by performing a scoping review following the 6-stage methodology framework by Levac et al [26] (based on the framework initially proposed by Arksey and O’Malley [27]), which we reported in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) framework [28] (the PRISMA-ScR checklist is available in Multimedia Appendix 1). Our aim was to obtain a large spectrum of articles, from original research to opinion papers, law articles, and workshop reports. Our review protocol was finalized in July 2021, but was not made publicly available.

We defined health data as data generated either in the context of health care, whether found in patient charts or connected objects, or in the context of health care research to capture perceptions about all kinds of secondary uses. In these secondary uses, we included approved future uses of biobank data because the literature on that matter was quite relevant to our research question.

The term “transparency” characterizes a process of information communication that follows certain norms about the amount, type, and framing of information that is shared. For example, the process of informing a person about access to one’s health data by third parties through a poster in a health professional’s office might be characterized as aiming at being transparent. By extension, the term “principle of transparency” can refer to a normative principle that instructs to communicate information in a transparent way (in the first sense). Thus, the 2 senses of the term “transparency” are closely linked. Since the 2 core notions of transparency are its relevance and its easiness to understand, this suggests that an investigation of transparency should concentrate on questions such as “What information should be conveyed?” “How should it be conveyed?” “In which circumstances should it be conveyed?”

An extensive literature search was performed for articles up to November 2022 in the following bibliographic databases: Medline, CINAHL, PsycINFO, Scopus, Cochrane Library, and PubMed. The search strategy was developed with the assistance of a university librarian and was adapted to each database (the detailed search strategies are provided in Multimedia Appendix 2). The 5 core concepts used in the search strategy were patient, public, citizen (and synonyms); health data, medical data, data use, data sharing (and synonyms); attitude, view, perspective, opinion, position (and synonyms); information, transparency, communication, dissemination, awareness, notification, education (and synonyms); and research, secondary use, LHS (and synonyms). We did not have to limit the time span.

Two reviewers (RD and FL) independently screened the titles and abstracts of all articles identified by the search strategy (after removing duplicates), followed by a full-text review of the remaining articles. To perform an in-depth and broad research, no article was excluded based on year, type, or location of publication. The eligibility criteria are presented in Textbox 1.

The 2 reviewers first performed a joint screening on a 10% sample of the articles to familiarize themselves with the selection criteria and to standardize their approach. Every discrepancy between the 2 reviewers was discussed to obtain a consensus. If a consensus could not be reached, a third reviewer (AC or J-FM) was consulted. The team met weekly to discuss the progress of the literature review. The search and screening process was initially completed in July 2021 but was updated up to November 2022 following the same methodology to capture and include recent publications.

The extraction, organization, and analysis of information from the articles included in this scoping review was performed using NVivo 11 Pro software (QSR International), a tool to conduct extensive literature reviews [29]. On the basis of the aforementioned research questions and objectives, the research team developed a codebook with an initial tree of nodes to chart the data into NVivo. This codebook also includes the coding procedure and definitions of the nodes for the reviewers responsible for the data extraction. All selected articles were imported into NVivo software. Four reviewers (RD, FL, AC, and 1 research assistant) coded all the articles, a process that was subsequently validated by 2 reviewers (AC and J-FM). All reviewers received detailed training in the software, project, and codebook. They all simultaneously coded a sample of the articles to familiarize themselves with the tree of nodes and coding method. During training, the reviewers’ coding was compared with determine whether their approach to data extraction was consistent, and discrepancies were discussed in team meetings. This process was repeated until all reviewers were comfortable with the coding. The coding was performed in an iterative manner to allow the emergence of new concepts and themes in the tree of nodes. Weekly team meetings were held between reviewers to discuss and validate coding decisions as well as emerging themes. The information was subsequently analyzed to examine what information, time frame, and means of communication were proposed, recommended, or desired for various research activities regarding the access and secondary use of health data for research purposes. We also aimed to identify topics that were not addressed in the literature for future work. The following 3 themes emerged from a text analysis: content to communicate, how to communicate, and situations in which communication becomes relevant. A fourth theme also emerged, namely, the impacts of communication, which were further divided into potentially negative, neutral, or positive impacts. This last theme was not directly linked to our research question and was the focus of a separate article. Note that the literature update from July 2021 to November 2022 did not bring any new concepts to light but rather confirmed that we had reached theoretical data saturation.

The detailed selection process is illustrated in Figure 1. After removing duplicates, 2964 articles were identified through the database search. The screening of titles and abstracts led to the assessment of 518 articles in a full-text review. In total, 152 original articles and 13 review articles met our inclusion criteria. The manual assessment of the references from the review articles added 26 original articles. Only original articles of those reviews that met our inclusion criteria were included in the final selection of this scoping review. Finally, 178 articles were included in this scoping review.

The main characteristics of the articles included in this review are presented in Table 1, and 178 complete references are presented in Multimedia Appendix 3 [6,8,11,18,19,30-202].

There were 2 distinct types of content that emerged from this review. We refer to them as “generic content” (communication on broader aspects than the secondary use of their own personal health data) and “specific content” (communication on the secondary use of their own personal health data).

This scoping review presents an extended analysis of the literature on patients, public, and other key stakeholders’ perspectives on transparency and secondary use of health data. It reports findings from 178 studies with various designs published over the past 2 decades. This extensive work is the first to map the views of a broad range of stakeholders on what information should be communicated to data contributors and how and in what situations to ensure transparency with the secondary use of health data. Figure 2 summarizes the findings of this study. There was a major overlap in the perspectives of different stakeholders. Some groups expressed opinions that reflected specific needs, such as data sovereignty in the context of Indigenous community data or access to results for patients with rare disorders. Overall, communication was deemed crucial for many purposes, including educating patients and members of the public on potential benefits, giving some control over data use as a form of reciprocity and as a condition to build and maintain trust. Elements that should be communicated include generic content, such as governance and regulatory frameworks, scientific aims, potential future uses of the data, and specific content that is relevant to each person with regard to the use of their data. Regarding methods of communication, broad approaches were generally favored, such as nationwide publicity campaigns, mainstream and social media for generic content, and mixed approaches for specific content, including websites, patient portals, and face-to-face encounters. Patients and members of the public rarely specified on whom the onus for these communications would fall, but we can imagine a shared responsibility according to intent and means between governmental, national, health care organizations, other large data custodians, and, to a lesser extent, individual researchers. Content should be customized to the individual as much as possible with regard to the length, avoidance of technical terms, cultural competence, and level of detail.

At this moment, in Western societies, sharing both generic and specific content seems necessary to achieve the goals of raising awareness, creating and maintaining trust by respecting the moral concerns of donors [110], and addressing the specific information needs of populations that have experienced racism and research violations [166,186,190]. Raised awareness and education are fundamental for members of the public to understand the legitimacy of various secondary uses of health data [58,79,134,163,194,195]. However, the data suggest that the current knowledge deficit is important [196,197]. In the international survey by Milne et al [85], providing transparent information about who will benefit from access to genomic data was the most important measure for increasing trust. There is an ongoing perceived need for control over the secondary use of health data and an aspiration for greater equity in the science-public relationship [31]. These findings confirm the conceptual framework by Brall et al [147] of individual willingness to participate in personalized health research and emphasize certain elements. The framework identified 20 elements pertaining to attitude, motivation, utility of results, data sharing and use, data management, and data governance that can have an impact on individual willingness. All these elements are presented in our review. Their relative importance could evolve with society’s greater understanding of data, measures taken to make their reuse secure, and confirmation of the benefits of secondary use. Achieving the objective of transparency needs to address its various facets whether it is “informational transparency” defined by Aitken et al [31] as “requiring disclosure of information on which decisions are based,” “participatory transparency” defined as “enabling public participation in decision-making processes,” and “accountability transparency” defined as “decision-makers are held accountable.” Any step toward greater communication and transparency should be carefully considered to ensure relevance and quality and decrease the risk of negative backlash [204].

A narrative review of the literature highlights the fact that social licenses for data-intensive health research cannot be assumed [205]. Trust and its absence had a considerable impact on the acceptance of the secondary use of health data without specific consent [136,191,198]. Muller et al [206] emphasized the merits of using the concept of social license as a reference for ethical governance and highlighted the strategic importance of dialogue with patients and members of the public. Research is ongoing to define what secondary uses of health data are within a social license and how perspectives could change with greater transparency [207]. On the basis of 8 focus groups with Canadian members of the public, Paprica et al [194] concluded that “[r]esearchers and organizations that hold health data should engage with members of the public to ensure that research is aligned with social license, particularly where there is private-sector involvement.” These findings are also reflected in a similar study of deliberative focus groups conducted with United Kingdom citizens [208]. Furthermore, we would do well to not assume and rather gather more evidence to support the idea that transparency and trust will necessarily go hand in hand [20].

Reciprocity was also a transversal theme in this review and is likely a foundation on which to build this common understanding with different communities and populations. Indeed, Sexton et al [34] linked trust and reciprocity and raised the concern that “[f]aceless processes of governance are increasingly foregrounded over traditional relational bonds.” Hobbs et al [83] explored the privacy-reciprocity connection, whereby participants attached certain conditions to their research donations. Patient engagement is another form of reciprocity that could increase trust [31]. Finally, reciprocity from the private sector appeared to be an important condition for sharing health data [136,191].

The return of individual results, particularly if perceived as potentially actionable, was a commonly cited example of a specific communication that patients and members of the public wanted, and this was substantiated by other research [209,210]. For some but not all, it was a motivation to donate health data, including genomic data [199]. What is less clear is whether, in the absence of the return of individual results, the need for information could be satisfied by communication on aggregated or global research results.

We also tried to distinguish whether individuals wished to receive specific communication about the secondary use of information by a single research project or by a group of projects with similar features. However, that distinction was not clearly present in our analysis. It could be interesting to confirm that an intermediate level of information is linked to the category of projects (eg, academic research, research with genomic data, and research with a private partner) that may or is using the data could meet individuals’ need for information without excessive use of resources [192]. This would also align well with a metaconsent model, a promising dynamic approach to information and consent in the context of LHS [11].

There will be no one size fits all—some individuals wanted regular communications by varied means and at specific times, for example, each time their data were used, whereas others wished for basic information to be available on demand so that resources can be allocated elsewhere. Individual factors, including age, ethnicity, gender, education, general and scientific culture, curiosity, and motivations to participate in research, were key to explaining the variety of views on the subject [56,59,83,151,158,166]. Biobank participants, for example, may have a more altruistic outlook and a reduced need for the return of information than members of the public at large who are less aware of the risks and benefits of research. It will also be important to pay attention to the “unheard voices” of people who are marginalized in society [169]. Public deliberation is an interesting approach to better understand specific population needs and preferences for notifications about data sharing [114]. Any communication strategy will need to be patient-centered and modifiable so that low information seekers are not overwhelmed by unwanted communication, and high information seekers can equally meet their needs [11,69]. Communications should also be adapted to the needs of people with disabilities, especially considering that technology is creating an information accessibility gap as it tends to replace face-to-face communication [101]. Interesting models of communication include the frequently asked questions section that provides layered information so that an individual could obtain as much or as little information as desired [103] and the OPT-IN framework by Nelson et al [176] “[...] to help guide communicators with selecting and presenting data to lay audiences, taking into account the broader communication perspective.” Guidelines on how to discuss the protection and use of personal health information would also be welcome [59].

There are other noteworthy initiatives to help us pave the way forward and find a balance between information needs, trust, and societal benefits from the secondary use of health data [200]. We would like to highlight 4 key features: patient or public engagement, harmonization of policies, the importance of user-centered informational tools, and innovations in data security.

First, transparent governance and patient engagement are increasingly being proposed as important avenues toward social license. Paprica et al [87] proposed the minimal specifications for data trust. Participatory governance is a key concept, as exemplified in the Academic Research Network of Diabetes Action Canada [72], the Research Data Alliance COVID-19 Working Group guidelines [201], and in 1 of the 7 recommendations from the work by Courbier et al [76] with patients with rare diseases. Participatory governance is also necessary to respect the Principles for Indigenous Data governance [201].

Second, the adoption of coherent policies and practices can help harmonize different initiatives and reduce risks, such as privacy breaches [44]. National or institutional standards such as Quebec’s Fonds de Recherche Standards for Data Access Centers [211], New Zealand’s integrated data infrastructure [166], and the United Kingdom’s data registry standard [212] are a few examples of initiatives that can serve as models.

Third, educational tools will be required to raise awareness among all stakeholders, with a focus on members of the public and tailored to specific populations. Such large-scale initiatives could address the frequently expressed need for information to help individuals control their use of data. Indeed, studies have demonstrated that once informed of regulatory safeguards, presented with conditions around use [6], informed of impacts [213], or sensitized to logistic hurdles, opinions regarding individual consent and control can shift.

Fourth, we should build on existing communication strategies. The Health Care Information Directive is an example of a patient decision aid that aims to delineate the level of health information that an individual is willing to share [100,214]. Caine et al [74] derived 6 implications for the design of a patient-centered tool to allow individual choices in the disclosure of health data from patient interviews. In their systematic review of the design of patient aids, Vaisson et al [215] proposed a user-centered design framework that could help design future aids to track the secondary use of one’s health data. The proposed concept interface enables contextual control by allowing patients to set their privacy levels in the context of viewing events within an electronic health record. Fagerlin et al [216] proposed 10 pragmatic recommendations to help patients make decisions when faced with complex information on risks and benefits that could be applied to making decisions about data sharing. On a more theoretical level, Zikmund-Fisher [217] proposed a standard taxonomy of risk concepts that could inform or evaluate future communications. Riso et al [70] outlined a framework with 6 core values to improve the ethical standards of data-sharing platforms. For example, the GA Registry [218] is patient-led and permits large-scale data sharing while letting users decide which data sets to share and with whom. Exemplary projects and patient portals include the tailored, ongoing communication of the CHRIS longitudinal study [202], NIH-funded All of Us Research Program [219], Connect Care (Canada) [220] that allows bidirectional interaction with the patient, Sundhed (Denmark) [221], and Health data hub (France) [222].

Finally, ongoing innovations in data security will be important to enhance privacy and trust in all these initiatives. Examples include CrowdMed, a blockchain approach proposed by Shah et al [180] and MiNDFIRL, a software to enhance privacy in secondary database studies [103].

Our search yielded most articles originating from North America and Europe and documenting Western contexts and attitudes; therefore, our review may not entirely represent other societies. In addition, although we cited research conducted with stakeholders from Indigenous communities, we cannot conclude that the views collected are representative of all Indigenous communities. Furthermore, many earlier articles focused on research conducted in the context of biobanking. We restricted our analysis to articles that explored stakeholder perspectives with a focus on the future use of samples or data. During thematic analysis, we were able to conclude that many of the perspectives overlapped with research conducted in other contexts, such as administrative databases, data mining, and mobile app. This suggests that for the person concerned by the secondary use of health data, there are similarities between the research contexts of biobanking and newer contexts.

This scoping review highlights the diversity and extent of patient and public expectations regarding information on the secondary use of health data. It also attests how and in which situations the communication of different types of information could contribute to the objective of greater transparency. These findings suggest that governing bodies should actively invest in widespread and targeted activities to increase public awareness and understanding about the secondary use of health data. Modern societies would do well to foster a culture of information and health literacy, which in turn is necessary for individual empowerment in the context of health care. Indeed, our results highlight the importance of patient, citizen, and community engagement in different secondary uses of health data [104]. This is particularly true for some populations such as Indigenous communities that have developed their own models of data governance [223].

Future studies should consider building on existing models to develop thoughtful strategies for communicating with the secondary use of health data. This review did not identify resources specifically considering the LHS context; therefore, the impact of such a framework on the desires of patients and members of the public remains to be explored. To support the implementation of LHSs in Quebec (through a transparency portal), the next steps would be to validate some of the results with patients and members of the public, including the following: (1) the need for both types of content (generic and specific); (2) the level of generic content to meet information needs and foster and sustain trust; and (3) the resulting most appropriate format considering resource use and development of web portals to support LHSs [224]. Other research could also validate how we can match actual citizen and patient needs with congruent types of knowledge and informational tools (Table 2 in the study by Zikmund-Fisher [217]), and how the personalization of communications meets the demands of different regulatory bodies, for example, the recommendations of the European Commission [203]. Another important question is whether transparency leads to unexpected outcomes, such as desensitization and “systematic exploitation” [111]. Finally, these strategies should be evaluated to ensure that they reach their objectives, notably with regard to transparency and even improving health outcomes at the individual and system levels. Beyond just informing members of the public on the secondary use of their data, we should aim to foster a culture in which all citizens are given tools to participate in their health care and determine research priorities [139].