The growth of big data allows researchers to use and link large, multisource health data sets for research. Big data is still an evolving field [1], and disagreements remain on precisely what the term stands for in health research [2]. Other terms used include routinely collected data [3] and data-intensive research [1,4]. For clarity, throughout this paper, we will refer broadly to the term big data as it is used in the literature and easily understood by the public. We follow the definition by Aitken et al [1], recognizing that the main feature of big data is the ability to link large data sets for analysis. They name sources for such data as patient records, administrative, registry biobanking, social media, and digital application data. Big data research in health can be used for multiple purposes with the aim of improving health care services and reducing health inequalities [5,6]. These include service management, evaluation or audit of services, statistics, and exploring connections between health and non–health-related outcomes [1]. Often, these purposes differ from the original intent of data collection (eg, health care or statistical purposes). In other words, big data is often used for secondary research purposes.

Big data research offers new opportunities for academics. However, reusing big data for research faces ethical challenges [7]. Previous big data initiatives suggest that the public must have confidence that their data will be used in an acceptable way if they are going to be supportive of big data research [8]. This means moving outside what is legally required and establishing a social license for research [9]. Carter et al [9] proposed 3 conditions for establishing a social license for big data research. First, reciprocity is essential, as there is a need for 2-way communication and improving public awareness of big data research as well as improving researchers’ understanding of the public’s concerns and expectations. A lack of transparency could make it challenging to secure public trust [10], and the public has a right to be informed about the progress of the research [11]. Second, the process should empower, not disempower, the public; in big data research, this could include members of the public involved in the governance of data linkage and the design of big data projects. Third, big data research should benefit the public; thus, researchers need to understand what the public might perceive as public benefit.

Public involvement and engagement could be used to bridge the gap between researchers and the publics’ understandings of the benefits of big data research [12]. There is evidence in the literature (outside big data) that public involvement can provide legitimacy for research [13]. Public contributors could be a part of the process of creating research norms for big data research [14]. Research norms consist of governance and regulation that could guide research. These might not be popular among some academics, but they could help secure a social license for research [15]. Aitken et al [1], in their consensus statement on public involvement with big data research, go a step further and argue that “the public should not be characterised as a problem to be overcome but a key part of the solution to establish beneficial data-intensive health research for all.” There is emerging evidence that public contributors can be meaningfully involved in big data research projects [16-18]. However, there is a need to understand how public involvement and engagement takes place in big data research comprehensively.

Previous reviews have examined literature around public trust and attitudes toward big data research [19-22]. Despite public involvement and engagement being seen as one of the ways to improve public trust, as far as we are aware, there have not been any previous reviews exploring public involvement and engagement in big data research and there have not been any reviews registered on the PROSPERO and Cochrane databases. Therefore, this review aimed to synthesize what is known about public involvement and engagement in big data research. Using scoping review methodology [23-25], we mapped key issues in the research to find evidence of how public involvement and engagement were carried out in big data research. Understanding how to involve and engage the public in big data research could be used to formulate guidance for researchers and policy makers on how to do this effectively, as there are field-related challenges, especially regarding the abstraction and complexity of big data [26].

The protocol for this scoping stage review was published previously [27]. The protocol outlines the parameters of the review and provides a justification and explanation of all the methodological steps and decisions taken. To ensure rigor further, we used the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist [28] and reported it as Multimedia Appendix 1.

In the literature, the terms involvement, engagement, and participation are used interchangeably, but these do not always have the same meaning [29,30]. This makes research and discussion about public involvement challenging, as it can be difficult to identify papers for review [31-33]. Hence, there is growing recognition that more consistent terminology is needed [13]. The diversity of types of involvement can be seen in the ladder by Arnstein [34] that determines types of involvement by constructing a typology based on the amount of power given to the public. It identifies from the bottom (lowest extent of people’s influence) to the top (highest extent of people’s influence) the following steps: therapy manipulation, nonparticipation, informing, consultation, placation, partnership, delegation, and full citizen control. The author herself called the ladder “provocative.” One of the health-specific definitions of public involvement has been developed by INVOLVE [35]. It has been used broadly by funders and researchers and embedded in the public involvement reporting checklist [33]. It offers a nuanced perspective on 3 types of activities: involvement, engagement, and consultation, which researchers can use when working with members of the public. One is not better than the other, but rather, each offers a different approach. INVOLVE defines involvement as research carried out with or by members of the public rather than to, about, or for them. This recognizes shared ownership of research with members of the public. Engagement is providing information about big data research and disseminating it to the public. Consultation happens when the research is discussed with the public, but there is no shared ownership. Thus, engagement and consultation are “to,” “about,” or “for” rather than “with” or “by” them. However, these activities can provide an understanding of the public views.

Owing to the diversity of definitions of public involvement and engagement used in the literature, we mapped all included papers using the INVOLVE definition, identifying whether they were involvement, engagement, or consultation.

Public involvement in reviews can improve their quality by contributing to defining the scope, appraising the papers, and interpreting results [36,37]. In total, 2 public contributors (SA and NT) were involved in the review from the initial design stage and contributed at each stage (screening, data extraction, and analysis). They are both experienced public contributors and previously copublished papers around public involvement and engagement in big data research. SA and NT ensured the relevance of review results to the public. This was achieved by relating results to their experience as public contributors in other research projects. The details of the involvement process and what was put in place to support them (eg, training) are reported elsewhere.

Following the search strategy developed with the support of a university librarian, the CINAHL, Health Research Premium Collection, PubMed, Scopus, and Web of Science databases were searched for papers in September 2021. The search strategy, as published in the protocol paper, is included in Multimedia Appendix 2. The search covered papers published after 2010 until the search completion in September 2021. Additional manual searches were conducted. These included the screening of the first 100 results from a Google Scholar search, journals that aim to publish public involvement research (BMC Research Involvement and Engagement and Health Expectations) or had special editions on public involvement in big data (International Journal of Population Data Science), and gray literature (the first 100 results from the Patient Outcome Research Institute database were screened). A call for potential papers to be included was posted on X (previously known as Twitter) to reach experts in the field.

The review included papers that met the following criteria: (1) discussed public involvement or engagement in big data research (those that appeared more as consultations were not excluded, but a note was taken of this), (2) focused on patient- or health-related research, and (3) were published in English. All study designs and nonempirical discussion papers were included.

PT took the lead by screening all papers. SA, NT, and EJ jointly screened at least a random 20% of papers at each stage (title, abstract, and full paper). Any discrepancies were discussed by the research team. The reasons for exclusions at a full paper stage were recorded and reported in the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist.

The data extraction form development was iterative and tested by the whole research team. The final data extraction form is available in Multimedia Appendix 3. PT extracted data from all papers in the first instance. Then, all extraction was double checked by the rest of the research team, thus ensuring each paper was considered by 2 researchers. The research team met regularly to discuss any discrepancies and discuss initial findings. PT organized the extracted data in a descriptive and narrative way under key headings based on the data extraction form. This was discussed with the research team.

The analysis was supported by a prior system logic model that we published in the protocol paper (Figure 1 [27]). It was initially developed by a preliminary scoping of the literature, research team discussion, and input from the public contributors. The logic model assisted us in identifying relevant elements of public involvement and engagement in big data research. We mapped our findings under the model and present them using headings from the logic model.

The database searches produced 4054 papers. Additional manual searches added a further 11 papers. After the removal of duplicates, 3540 articles were screened for inclusion in the review. A total of 3342 papers were excluded based on the title and abstract. The full-text screen took place for 198 papers, and 53 were included in the review. Figure 2 [38,39] shows the PRISMA flowchart of the screening process. We first discuss the study characteristics and thereafter present findings as mapped under the revised system logic model (Figure 3 [27]).

The most prevalent type of papers were discussion papers (nonempirical, including conceptual or ethical papers; 28/53, 53%), followed by review papers (5/53, 9%); qualitative study design (5/53, 9%); opinion, letter, commentary, or editorial (4/53, 8%); evaluation (3/53, 6%); protocol (2/53, 4%); ethnographic or descriptive case study (2/53, 4%); public deliberations (1/53, 2%); action research (1/53, 2%); quantitative (1/53, 2%); and mixed methods (1/53, 2%). The papers were from the United Kingdom (19/53, 36%), the United States (10/53, 19%), Canada (7/53, 13%), New Zealand (3/53, 6%), the Netherlands (1/53, 2%), Portugal (1/53, 2%), France (1/53, 2%), South Africa (1/53, 2%), Australia (1/53, 2%), Germany (1/53, 2%), and Africa (1/53, 2%). In total, 12 papers did not specify a geographical location, and some papers included more than one. The most prevalent type of involvement and engagement activities carried out with the public (following INVOLVE definitions) were involvement (45/53, 85%), followed by engagement (25/53, 47%) and consultation (7/53, 13%). Some papers discussed >1 type of activity. Table 1 presents the characteristics of the included papers.

The demographics of the public or communities involved and engaged in big data research were diverse. These included patients (including consumers and service users; 12/53, 23%); affected groups or groups considered vulnerable (8/53, 15%); Indigenous communities (6/53, 11%); articles focusing on specific characteristics (eg, gender, age, income, education, or geography; 5/53, 9%); carers (4/53, 8%); the general public (3/53, 6%); ethnic minority groups (3/53, 6%); patient representative or community leaders (3/53, 6%); and research study participants (1/53, 2%).

Deciding who should be on advisory boards, how they should be selected, and what their role should be remained a challenge for researchers [82]. An important issue was representativeness; advisory boards were unlikely to represent all the public views [66,69,87]. No single committee could represent all communities (because of their diversity) [58,76]. Identifying the relevant communities was seen to be difficult [82]. This created the challenge of ensuring legitimate group representation [72]. Advisory groups often did not reach a broader population [68]; hence, involvement and engagement need to move away from the “usual suspects” [16,18,66,76]. There was the risk that more vocal individuals could dominate the discussion [82]. Public contributors could be chosen arbitrarily, for example, based on personal contracts, and thus, the process might not be transparent to the public [72]. This could lead to involving financially and politically motivated [49] or well-connected contributors [42]. The way to overcome these issues could be to recruit public contributors from the study participants; for example, participants could elect their own representatives or a marketing company could conduct the recruitment [72,81].

Researchers should respect local and seldom-heard groups’ traditional structures and ethical perspectives. Papers focusing on Indigenous communities showed already existing governance mechanisms supporting research with these groups [59,79]. Researchers should incorporate Indigenous culture, for example, traditional ceremonies, when involving the community [58]. Formalized agreements with Indigenous organizations could improve the relationship with that community [55]. This more nuanced approach to big data research could assist researchers in establishing trust with Indigenous communities rather than merely convincing them that this is the right thing to do [59].

Political situations or public perspectives and attitudes could influence how and why members of the public get involved in big data research. Secrecy could be a challenge [11]. Organizations might not want to share controversial information, and private companies may argue that sharing it might be against their commercial interests [42]. Involvement and engagement could have the potential to improve public trust in big data research but not necessarily in the research institution [51]. There could be historic mistrust from underserved communities, for example, African Americans, Indigenous communities, and people living with homelessness [11]. There was no guarantee that it would always be possible to maintain public trust in big data research [67].

Some included papers in the review claimed that involvement and engagement should have clear outcomes. First, it could identify gaps in knowledge and priorities for research [70]. Second, it could align researchers’ and institutional perspectives of public interest with public views [44], for example, by bringing together charity workers, service providers, elected politicians, and members of the public [54,70]. Third, public contributors involved in governing bodies could have the effect of improving trust and accountability [84]. Fourth, improving public awareness of big data might democratize health research [62]. For example, Vayena and Blasimme [84] argued further that blending citizen science and participatory models could offer more democracy in governance.

However, measuring the impact of involvement and engagement in big data research was challenging [3,64,82,83]. A scoping review by Luna Puerta et al [64] recognized that there was no consensus about the objectives of public involvement in big data research, which undermines the ability to measure impact. Another review by Tindana et al [82] found that the papers included in their review on community engagement did not evaluate the effectiveness of engagement activities.

Engagement through genuine public debate could help demonstrate that the public sector could be a trustworthy steward of patient data [42]. This should include any negative comments toward the initiative; these should be publicly shared, and justification should be provided as to why their feedback was not implemented [42]. Dankar et al [47], when discussing biomedical databases, suggested that sharing research findings should include reaching individuals with personalized research results; these need to be valuable and benefit individuals (eg, they could go for health tests or make life changes that improve their health).

This scoping review provides an overview of how public involvement and engagement have been used in big data research or how it has been argued that it could be applied. This is the first review exploring this issue. The review has shown that the public can and, many articles argue, should be involved and engaged in big data research in terms of individual initiatives and data governance. However, the findings indicate that there is no one right way to involve and engage the public in big data research. Those responsible for working with the public should consider what type of activities are most relevant to their work and should use multiple approaches (involvement, engagement, and consultations) to reach different communities. Some papers suggested using modern technology when engaging the public (eg, through a website or digital newsletter). However, most included papers were not primary studies.

The review indicates that many believe that public involvement and engagement have the potential to improve public trust and accountability for big data initiatives. However, there is limited literature on how public involvement and engagement might influence it. Future research should attempt to measure the impact of involvement and engagement in securing social license for big data research with the broader public. The initial step to improve this situation could be to ensure reporting by using standardized reporting guidance for public involvement, such as GRIPP2 [88].

References to public involvement and engagement guidance or legal documents in the included papers were limited. The consensus statement on public involvement and engagement with data-intensive health research [1] is relatively new. However, INVOLVE (now incorporated into the NIHR) has been active in the United Kingdom since 1996. This indicates that many included papers replicate similar discussions around principles involving and engaging the public rather than referring to already established standards. However, more big data–specific guidance is being developed by the Public Engagement in Data Research Initiative in the United Kingdom [89].

The findings of this review indicate that some challenges are particularly relevant for involvement and engagement in big data research. However, the review has also shown that public involvement and engagement in big data research are not dissimilar to other research fields, as they share aspects of involving and engaging the public, such as working with seldom-heard communities and addressing power balance. This suggests that big data researchers could also use generic public involvement resources, such as the National Standards for Public Involvement in the United Kingdom [90].

The main challenge is that big data research is a complex topic. It might not be easy to explain it briefly (or in accessible language) to potential public contributors or the public. The papers offered some suggestions on how these barriers could be overcome. Researchers need to ensure that they allocate sufficient time and resources when discussing big data research with members of the public. This finding aligns with another review that examined patient involvement in cancer research, where the authors identified time-consuming involvement as a primary challenge in that context [91]. This review suggests that involving and engaging the public in big data research might be even more time consuming than in other fields. If these challenges are overcome, there is a higher chance that involvement and engagement in big data research is not tokenistic, but this might mean additional time and financial resources. Researchers should budget for these resources as they design any involvement or engagement activities. However, they should be supported to do it by research institutions and funders.

Bailey et al [92] reported that Black and South Asian communities in the United Kingdom have less trust in the health system, and because of this, there might be concerns within these groups about how the public bodies use their data. Researchers need to recognize how trust and attitudes toward big data research could influence public involvement and engagement. This review has offered some indication of how to achieve this from the literature that explored working with Indigenous communities, such as recognizing communities’ beliefs and way of life.

The protocol that this review was based on presented the priori system logic model for public involvement and engagement in big data research [27]. On the basis of the review findings, the model was revised. Within the context section, Indigenous standards were added to recognize that big data research needs to consider the perspective and views of Indigenous communities that might differ from previous dominant perspectives. In the intervention theory section, the execution of involvement activities could be divided into project-specific aspects (eg, focusing on 1 big data research project) and governance bodies that look into granting approvals into data linkage (for other projects). These 2 purposes might influence how researchers involve and engage the public. In intervention delivery, the bullet point around public-led activities was added, as some papers suggested that it was important to ensure that the public voice is equivalent to professionals’ views during voting and should have equal or even more influence (eg, by cochairing meetings or being coinvestigators). Furthermore, a new bullet point was added in intervention delivery to recognize big data–specific barriers, especially jargon, and how complex big data research could be to members of the public.

Most of the elements included in the model were discussed in the included papers. The only exception is that it does not reflect on the involvement and engagement of people who are not personally affected by big data research (or do not perceive themselves as such). The coverage of most of the issues raised in the papers for involvement and engagement in big data research suggests that the logic model could support researchers who intend to design and deliver these activities to the public.

Textbox 1 provides a summary of the key recommendations around public involvement and engagement in big data research based on the review findings.

The first limitation is the use of terminology. The review explored public involvement and engagement in big data research. These terms are used in different ways by researchers. This parallels the experience of Brett et al [93] in their review, where they found that the variability in wording used to describe involvement complicated literature searching. The search strategy was developed with an experienced librarian and included additional manual searches. However, this did not guarantee that all relevant papers were included. This could have influenced the search results, as potentially some relevant papers might not have been picked up by the search as the authors used different terms. The second limitation was that only information included in the papers was extracted. The authors of included papers were not approached for more details. As academic papers have a word limit, it is possible that some additional information about involvement and engagement may have not been included in the published paper. In contrast to the initial plan, the references of included papers were not screened for potential inclusion. This was because screening of references of included papers in the scoping review was considered impractical because of the high number of papers. Moreover, only papers published in English were included. Finally, owing to the number of papers identified through the searches, only a random sample of 20% was screened by all coauthors.

This review offers a snapshot of evidence on what public involvement and engagement in big data research could look like. It is limited, as it was largely based on discussion papers, but useful, as evidence on how these involvement and engagement activities could be delivered and what type of outcomes they could produce was provided. The field would benefit from further research and evaluation of involvement and engagement activities in big data through primary research. Owing to the ongoing development of big data research, it is likely that these would need to be updated on a regular basis, but nevertheless, such research could provide further insights into how to meaningfully involve and engage the public in big data research.