Community engagement in the development of health-related data visualizations: a scoping review

Abstract

Objective

This scoping review aims to address a gap in the literature on community engagement in developing data visualizations intended to improve population health. The review objectives are to: (1) synthesize literature on the types of community engagement activities conducted by researchers working with community partners and (2) characterize instances of “creative data literacy” within data visualizations developed in community-researcher partnerships.

Methods

Using the 2018 PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines, the review focuses on peer-reviewed journal articles from 2010 to 2022 in PubMed, Web of Science, and Google Scholar. A community engagement tool was applied to the studies by independent reviewers to classify levels of community engagement, social determinants, and vulnerable populations.

Results

Twenty-seven articles were included in the scoping review. Twelve articles worked with vulnerable populations. Four articles attempted to alleviate barriers to representation in their respective studies, with addressing language barriers being the most prevalent approach. Thirteen articles considered social determinants of health. Sixteen studies engaged in iterative approaches with intended users when developing the visualization or tool.

Discussion

Only a few significant examples of creative data literacy are incorporated in the studies. We recommend a specific focus on engaging intended users at every step of the development process, addressing language and cultural differences, and empowering intended users as data storytellers.

Conclusions

There is room for deeper and more meaningful community involvement in the development of health-related data visualizations geared towards them.

INTRODUCTION

Background and significance

In the past few decades, society has experienced an explosion of data and with it, a new need to ensure public engagement with the messages in the data. Hence, the use of data visualizations to disseminate information has become increasingly prevalent. Data visualizations have been used throughout history in a multitude of contexts, including civil rights and healthcare. For example, social activist W.E.B. Du Bois utilized data visualizations to combat stereotypes and disparities experienced by African Americans in 1900 through his work, “The Exhibit of American Negroes.”¹ More recently, data visualizations have been highly utilized in the COVID-19 pandemic to share information related to health education, health risks, and statistical data with a broad range of communities.² In this instance, data visualizations highlight their utility in facilitating communication about health and risk.³

Unfortunately, when communities experience differences in accessing data (such as with poor broadband internet access) or have varying skills to create visualizations, the meaningfulness of such visualizations can be jeopardized.^4–10 This chasm in access, skill, and action related to data in the context of the “digital divide”—is especially concerning when data visualizations relate to the very communities that are affected by such disparities. The digital divide encompasses disparities in who can access and use data; who is represented in data is equally important, especially when it forms the basis of policy decisions and funding for health priority populations. The digital divide’s importance in relation to health literacy and health outcomes has been well described.^11,12 There is growing concern that the interests of today’s data power brokers introduce biases that contribute to who is and who is not reflected in data. Discussions, therefore, of data discrimination—the unequal ways in which data is created and how it comes into societal spaces—are increasing.^10,13,14

Data discrimination means that much of the control over how data is provided, how it is presented, and how it is managed (the “data narrative”) is dominated by profit-based businesses, technically adept individuals, and national or regional governments. There is limited input from the diverse communities that may be excluded from the data and its outputs.¹⁴ This has important implications for health communication, as how messages “land” relates to how they are created and by whom. Previous studies have demonstrated that cultural and linguistic barriers have a significant impact on health messaging and, thus, health outcomes. Individuals with low English proficiency are more likely to experience negative health outcomes than those whose native language is English.^15,16 Additionally, individuals lacking access to linguistically and culturally appropriate health services also experience poorer access to health information.¹⁷ Ensuring that data visualizations surrounding health are equitable and accessible becomes an imperative issue. Therefore, any effort that seeks to use data visualizations as a tool to mitigate data-related inequities (eg, the data divide or data discrimination) would benefit from a data visualization literacy framework that considers not only the intersecting challenges posed by the digital divide and data discrimination, but also links the unique perspective of being both a data user (eg, using the data for action) and data maker (eg, shaping what data counts). These concerns have led to the emergence of a new field of study, critical data studies,¹⁸ and a growing demand for taking a participatory approach to establish what data is collected, how it is collected, and how it is used once it is collected.

Data visualization literacy and meaning-making

There are different views about what comprises data visualization literacy, varying as much in conception as in scope.^19–24 Some authors approach the conception of data visualization literacy as a standalone competency to be achieved at the individual level, such as digital literacy or the numeracy component of health literacy.²⁴ Others emphasize the use of a tool in their description of data visualization literacy, and therefore focus on the attributes of images that may increase patients’ engagement with health communication tools, such as patient-decision aids or medication adherence apps.^25–28 However, in line with the work of W.E.B. Du Bois’s data visualization work,¹ the “act” of visualizing data is, in itself, an important component of developing knowledge and agency,²⁹ and is more along the lines of meaning-making.³⁰

As a result of these changing approaches to data visualization, new types of data activism are emerging, for example, through the collection and presentation of data showing disparities or “hidden” realities, which enables groups to advocate for change. Groups are then able to confidently choose and act on the data issues that are most critical to them, demonstrating collective agency and supported by newly developed data visualization literacy skills.³¹ This approach is prevalent in the environmental justice literature that aim to address socioeconomic inequalities in relation to the environment that lead to disproportionate health outcomes in communities. There is a focus on data activism in the conceptualization of data visualization literacy, beyond an emphasis on understanding individual exposure risks, whereby community members’ understanding of data visualizations enables them to advocate for changes to those risks, using localized evidence.³¹

Creative data literacy

Another framework, creative data literacy, can also support data visualization literacy, particularly in developing a more critical approach to data and its uses. This framework emphasizes the process of making things with data, wherein visualizing and interpreting data empowers people to enact change, collectively, in their own communities. Creative data literacy offers those without technical backgrounds a more accessible avenue for learning about data. This approach considers what data are relevant to the community, how data originate, how data are collected and processed, how learner-centered tools can support data literacy development, and what types of outputs best resonate with community stakeholders.¹⁴ Applying creative data literacy concepts to health communication is an emerging field of study that stimulates new thinking about the practice of creating data visualizations.¹⁴ Consequently, the community’s involvement in this creation process can be evaluated. To our knowledge, a formal review assessing the aspects of community engagement in this space has not been conducted.

In this scoping review, we aim to: (1) characterize the state of community engagement and creative data literacy skill-building in data visualizations surrounding health communication, and (2) to highlight instances of creative data literacy skill-building throughout stages of the visualization process—conception, application/implementation, and evaluation—that empower communities to think critically about data and its applications. We first characterize (1) how community members are engaged in the process of data visualization development with community partners (including those who may or may not have strong technical backgrounds) around local health issues and (2) the integration of community input to improve tools for health communication throughout the stages of this process. We conclude with recommendations for future developments of health visualizations for the community that involve creative data literacy.

MATERIALS AND METHODS

This scoping review follows the approach described by Munn et al³² to characterize trends and gaps in existing research; and to map evidence to inform future research. We followed the 2018 PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) guidelines.³³

Search strategy and selection criteria

We developed 2 search strategies in collaboration with a health sciences librarian at the University of California, San Francisco between November 2021 and March 2022 for PubMed and Google Scholar. We created an additional search strategy for Web of Science in August 2022 to add the most recent publications. We repeated our searches in March 2023 to include all March 2022 to December 2022 studies. Our first search strategy was “data visualization literacy” and our second strategy was (“health literacy” OR “health literate”) AND (“data literacy” OR “data visualization literacy” OR “data literate”). We applied a search flow diagram across the searches to map the studies identified, those determined eligible and ineligible using our predetermined criteria, and included studies (Figure 1).

Figure 1.

Flow diagram of article screening.

Adapted framework for characterizing active community engagement coding studies for community engagement

We also applied a community engagement assessment tool³⁴ to standardize how we characterize the degree and quality of community participation in our included literature. Specifically, we used an adapted validated, reliability-tested guideline for evaluating community engagement in participatory research projects. Mercer et al³⁴ propose 25 guidelines to consider when evaluating research project proposals; adhering to the authors’ recommendations for adapting the guidelines, we determined that 8 were relevant to our scoping review, whereas the others were more relevant to the protocol stage. For example, the guidelines proposed by Mercer et al contain 8 questions regarding plans to engage with community stakeholders after research activities have concluded such as, “Do the researchers and intended users engaged in the research process have a formal or informal agreement (verbal or written) for acknowledging and resolving in a fair and open way any differences in the interpretation of research results?” and “Do the researchers and intended users engaged in the research process have a formal or informal agreement (verbal or written) regarding the dissemination (and/or translation or transfer) of research findings?” We determined that, while engagement efforts of this nature may occur, they are not typically reported in peer-reviewed journals. Therefore, we condensed multiple series of questions into single items to concisely capture the relevant points.

The studies were given a “1” if they complied with the community engagement principle and a “0” if they did not. Three coders contributed to assessing the guidelines for each paper included in the review. The remaining studies were divided among the coders, and any doubts were clarified during regular meetings. For the social determinants of health coding, reviewers categorized whether social determinants data were collected, and for which categories (eg, defined as including a social determinant in the sampling or reported it in the results), according to those established by Healthy People 2030, an initiative set by the U.S. Department of Health and Human Services.³⁵ For example, safe housing, transportation, racism, discrimination, education, job opportunities, access to nutritious foods and physical activity opportunities, pollution in air and water, and language and literacy skills are considered social determinants in the United States. To ensure agreement, the first 5 studies were evaluated by all 3 coders and disagreements were discussed until a consensus was reached.

RESULTS

Search results

One hundred twenty-seven articles were initially found across the 2 databases, PubMed and Web of Science, using our 2 search terms. No new records were found from Google Scholar. The repeated search in these 3 databases in March 2023 generated 6 additional articles for a total of 133 articles retrieved. Forty-four duplicates were removed and the remaining 89 articles were screened through their respective titles and abstracts using the exclusion criteria (see Figure 1). Ultimately, 27 articles were included in the final results (of which 2 were from the March 2023 search). Included studies ranged in year of publication from 2010 to 2022, with 22 of the 27 studies published more recently, between 2017 and 2022 (see Supplementary Tables S1 and S2).

Review results

The questions and results of the adapted guidelines for the community engagement tool are discussed throughout the results section (see Table 1). Articles were separated into 2 categories: those that developed and deployed community-based programs and tools for communicating health-related information using data visualizations, and those that tailored health information to improve understanding of patients’ personal medical information using data visualizations. Eleven of the articles fell into the first category of what we refer to as community-engaged data visualization (Supplementary Table S1).^{23,27,28,36–43} Sixteen articles fell into the second category of what we refer to as visuals for understanding clinical information (Supplementary Table S2).^{21,25,26,44–56} Out of all of the selected articles, Beyer et al³⁶ and Stonbraker et al⁵⁰ met the greatest number of criteria within our adapted framework for active community engagement (6 out of 8 guidelines).

Table 1.

Adapted guidelines for community engagement

Guideline	N (Yes)	% (Yes)
1. Are the intended users of the data visualization described adequately enough to assess their representation in the project?	22	81.5
2. Is the mix of participants included in the research process sufficient to consider the needs of the project’s intended users?	16	59.3
3. Is effort made to address barriers to participation in the research process by intended users who might otherwise tend to be underrepresented?	4	14.8
4. Was the data visualization developed in collaboration with intended users in an ongoing, iterative fashion?	16	59.3
5. Does the research project consider social determinants of health in the development of the data visualization?	13	48.1
6. Does the research project provide intended users with the opportunity to participate in planning and executing the data collection (whether or not the intended users choose to take that opportunity)?	0	0.0
7. Does the research project involve intended users in interpreting the data visualization at a level that sufficiently reflects their knowledge of the particular culture, context, and circumstances in the interpretation?	22	81.5
8. Does the project describe any engagement efforts that occur after research activities concluded?	2	7.4

Studies focused on community-engaged data visualization

Of those studies that focused on community-engaged data visualization studies, 7 of the 11 studies worked with community health professionals of some kind (healthcare providers, public health officials, etc.),^{27,36–38,40,42,43} and 8 out of the 11 studies involved community residents.^{23,36–41,43} Four articles from this first category worked with participants who spoke another language other than or in addition to English,^23,36,38,39 with the most common non-English languages being Spanish³⁶ and Chinese.^23,38 Five of the 11 studies involved vulnerable populations.^23,36–39 The average number of community engagement guidelines met by the articles in the first category was 3.4 guidelines (out of 8) and the number of guidelines met ranged from 1 to 6. The mode was 4 guidelines.^37–40

Studies focused on improving understanding of health information

Of those studies that focused on understanding health information, 13 of the 16 articles met our criteria for engaging with patients from clinics or hospitals in the design process.^{21,25,26,46–53,55,56} Two out of the 3 articles that did not engage with patients focused on the general public instead of actual patients.^46,54 One out of the 3 articles that instead worked with caregivers of patients, but not directly with the patients themselves.⁴⁴ Three of the articles worked with participants who spoke another language other than or in addition to English, all being Spanish.^26,44,51 Seven out of the 16 studies involved vulnerable populations.^{26,44,47,50,51,55,56} The average number of community engagement guidelines met by the articles in the second category was 3.6 guidelines and the number of guidelines met ranged from 0 to 6. The modes were 3 guidelines^{25,45,46,48,51} and 5 guidelines.^{21,26,44,55,56}

Engagement with vulnerable populations and exploration of social determinants

Just under half (n = 12) of the total studies included in this review involved working with vulnerable populations, broadly defined as groups whose social conditions both promote various chronic diseases and make their management more challenging.^{23,26,36–39,44,47,50,51,55,56} Few studies explicitly made efforts to reduce barriers to participation in visualization projects among those they were recruiting, although there were some exceptions.^23,36,44,50 Notably, these 4 studies made available resources in another language in addition to English to encourage engagement. The mapping interface that participants utilized in the Wong et al²³ study was presented in English and Chinese for their users to navigate through the interface. Interface features and landmarks were included in traditional Chinese characters to support participant engagement. Beyer et al³⁶ also incorporated materials in multiple languages, including English, Spanish, and Lao, in their study which focused on educating participants about colorectal cancer and its risks, research methods for evaluating disease rates, and generating conversations around colorectal cancer risk within the community. The studies by Stonbraker et al⁵⁰ and Arcia et al⁴⁴ included materials in English and Spanish in order to accommodate participants’ comfort in different languages. Stonbraker et al held interviews in Spanish to elicit patient preferences and thoughts about the interface for a health application. Graph literacy was also evaluated with either a Spanish or English measurement tool.⁵⁰ Arcia et al⁴⁴ created infographics and held design sessions in both English and Spanish to facilitate discussions about the meanings of the infographics, preferences for design elements in the infographics, and reflections on participants’ own experiences with health. In the design sessions, participants were asked to elaborate on their perceptions of the infographics surrounding a certain health topic in the language they were most comfortable with.

None of the studies, however, offered users the opportunity to plan and/or execute the collection of data. Twenty-two articles did, though, allow for intended users to reflect on their own experiences in their interpretation of data visualizations.^{21,23,26–28,37–42,44–46,48–53,55,56} The most common method of soliciting feedback was through semistructured individual interviews.^{21,23,28,38,41,45,50,51,53,55} Examples of prompts for the feedback elicitation range from the type of graph (bar chart, pictograph, etc.) participants preferred²¹ to real-world applications of the intervention using new information learned.²³

A little less than half (n = 13) of the articles incorporated information about social determinants of health.^{21,23,26,36,37,39,40,43,47,50,53,55,56} Out of those articles that did, many focused on sampling key groups to explore in more depth the social and community context.^{23,26,36,37,39,47,50} For example, Wong et al²³ employed high school youths from the community as teachers to enhance the social connection between the youths and first-generation Chinese immigrant adults (who did not have fluency in speaking English) and to encourage the adults’ participation in the research project, “where social connections between young people and elders are culturally valued.” Lor et al utilized a similar approach with their focus group of older Hmong adults who were not fluent in English. The study utilized helpers that understood English and were recommended by the older adults. Lor et al³⁹ recognized that the older Hmong community typically had a trusted family helper to navigate them through English documents, so these helpers were included in the study to maintain cultural norms. Additionally, many articles discussed social determinants of health in the context of health care access and quality.^{21,39,43,47,50,53,55,56} For instance, Pack et al²¹ considered how lower health literacy combined with a lack of access to health data created difficulties in the communication of health data between patients and providers.

Health literacy was explicitly discussed in 15 articles.^{21,23,26,28,37–39,41,44,47,50,51,53–55} Out of those 15 articles, 6 articles actively evaluated the levels of health literacy within their participant populations,^{21,28,47,51,53,55} with the Newest Vital Sign being the most common measure of health literacy.^21,53,55 Data literacy was explicitly mentioned in only one article.⁴⁵

Visual formats in community-engaged data collaborations

Bar charts were the most common type of data visualization used, appearing in 9 studies.^{21,27,42,45,46,48,50,53,55} The most common type of output from articles involved data visualizations in some form of an infographic.^{26,38,40,41,44,47,48,54–56} Three articles presented data visualizations in geographic mapping tools.^23,36,43 Thirteen studies explicitly stated their focus on intended users’ preferences in tailoring data visualizations and tools.^{21,26,39,41,44–46,48–51,55,56} Out of those articles evaluating users’ preferences, 9 of the studies’ participants addressed the use or meaning of colors.^{21,26,39,44,45,48–51} Lor et al considered the cultural meaning that certain colors could convey. For instance, the color black was preferred over red to denote things with negative connotations. Participants revealed that in the Hmong culture (from which the participants were a part of), red represented a “bad omen/misfortune” and that people would refrain from incorporating the color into their daily lives, “including wearing red clothing and writing in red.”³⁹ Notably, Arcia et al examined participants’ preferences regarding word meaning and language in data visualizations. For example, in a visualization depicting levels of psychological distress, the Spanish word for distress, “trastorno,” is used. However, in Spanish, “trastorno” has a stronger connotation than does “distress” in English. The study found differences in preferences for including “trastorno” or an alternative word meaning “anxiety and depression” as the title of the visualization.⁴⁴ This language-concordant addition addresses the language barrier that could pose a challenge to non-English speakers in engaging with the development of data visualizations and tools.

Stages of data visualization engagement

Sixteen studies engaged in some type of iterative feedback in the development of data visualizations^{21,25,26,36,38,40,42,44,46–50,53,55,56} at the late conception and early implementation stages. Of these 14 studies, 8 involved intended users in more than just one round of feedback.^{25,26,40,42,44,48,49,56} Notably, Arcia et al⁴⁴ explicitly mentioned that their feedback sessions continued until reaching “design saturation.” In other words, the rounds of development ended when users’ interpretations of the data visualization matched the visualization’s message and no further changes to the graphics were needed. Only 3 studies gave intended users a voice in shaping the message of the visualizations, beyond simply changing its visual aspects.^40,43,56 Hammond et al allowed for intended users to identify pollutants for an environmental pollution map. Pollutants of interest were then explored for their origins, concentrations, and health risks in the community.⁴³ Also, participants in Snyder et al⁵⁶ offered input on what quality of life topics were most important to them and how prostate cancer has affected this quality of life. In addition to participating in iterative design sessions, the stakeholder advisory board for Sampson et al’s⁴⁰ study also prioritized what environmental health information should be presented to the community in the visualization.

Some studies also involved community members applying their own data or data they collected to the conception and implementation stages of creating data visualizations. In other words, they took a more personalized approach. Thompson-Butel et al⁵² created visualizations of patients’ own strokes in the form of virtual reality. Patients involved in the study stated that not only did the personal visualizations help to learn about mechanisms of their own strokes, but also to build initiative in improving their personal health. The study by Sandhaus et al is another example of utilizing personalized data; in this case, a results book with visualizations depicting environmental sample results. Similar to Thompson-Butel et al’s study, the community members from Sandhaus et al⁴¹ expressed their views about the initiative by sharing the information they learned with other members of the community outside of the study. There were no studies that engaged community members in the explicit evaluation of the data visualizations.

DISCUSSION

In this review we focused on exploring how community members were involved in developing data visuals that reflected key aspects of creative data literacy in health settings. Our review suggests that there are many ways that community engagement can be incorporated when developing data visualizations or tools related to the lived experience of the members of that community (see Figure 2). However, none of the studies we reviewed involved community members in the design and data collection development process, steps in creative data literacy.¹⁴ Some studies took a direct participatory approach that maps to the creative data literacy framework, while others focused on evaluation of the community’s ability to understand information from data visualizations and adjusted data visualizations accordingly to fit the community’s preferences or suggestions.

Figure 2.

Matrix for examining participatory engagement priorities and strategies in data visualization projects. This matrix reflects 2 key areas of decision-making that can shape the quality and degree of community involvement in the data visualization process. The horizontal axis represents the degree to which a project orients towards the data visualization process as a mechanism for improving health communication and knowledge dissemination to the community, or towards data visualization as a process of community transformation.^11,23 The vertical axis represents the degree of community involvement in shaping the data.

However, community participation in relatively peripheral roles (eg, responding to already prepared visuals) may limit their engagement with the data visualizations: this limited participation can be problematic if the data visualizations are being used to shape the data stories told about their own healthcare experiences. Only a few of these studies took participant preferences into account, for example, to respond to newly modified visualization in a 2-step process of input or evaluation. Even though some of the included studies used approaches such as semistructured interviews with participants to gauge their understanding of the data visualization and suggestions for improvement, few described the impact of that engagement on the creation of data visualizations. In the evaluation stage, the extent to which community members contribute to the creation and revision of data visualization are rarely specified, which raises questions about whether communities are authentically engaged as data visualizers in their own right. It is possible that these participatory processes did occur but were not described in the studies.

Approaches that facilitate community empowerment and strengthen the relationship between the community and data visualization researchers are needed. Only a few studies considered the importance of involving intended audiences in the drafting stages of the creative process. Most studies instead collected feedback and implemented revisions themselves, which can unintentionally reinforce the uneven power dynamic wherein the community member is the subject of the data and the researchers maintain their data maker/data user roles. Those projects that did explicitly promote data activism principles by facilitating deep community involvement encouraged creative data literacy skill-building, but also prompted calls to action and changes in data visualization practices.²³ Another strategy involved utilizing participants’ own data in creating the data visualizations, which can then shape the kinds of stories we tell with data and the relevance of the visualization work. This initiative to share with others is strengthened by participants’ interests, community-owned knowledge, and confidence in the subject. Visualizations from personal data also reveal an avenue to empowering community members to become advocates for appropriate health messaging in their own community. However, few studies highlighted participants’ own voices or narratives as part of the data visualization.

Noticeably, a handful of studies included material and visualizations in languages other than English to support a more diverse population of users. Future endeavors in visualization development should support and document data visualization work in multiple languages and explicitly characterize how participants “speak data” using multilingual approaches and the extent to which multilinguality strengthens engagement.

We also observed that few studies engaged the intended audience or stakeholders across the spectrum of visualization development. This includes engagement from the visualization’s initial conception, application/implementation, and eventual evaluation and dissemination to the community.

To summarize, we recommend greater investment in research and intervention agendas that address barriers to community involvement in data visualization work. This commitment empowers the community to influence the collection of data, the visuals created with that data, and the stories we tell about health disparities in their own communities. Active inclusion in the application/implementation stage ensures the process of visualization incorporates community’s needs and addresses cultural or linguistic barriers that may inhibit access to health communication pathways. Finally, strengthening community involvement in the evaluation of data visualizations supports the democratization of science communication, opening up more opportunities for sharing data stories in community spaces, beyond the conventional outlets such as academic journals and conferences. Consequently, the community has the opportunity to contribute as the data storyteller, rather than just as the subject of data. In our included studies, there was a bigger emphasis on the application/implementation phase than the conception or evaluation phases of the development process.

Lastly, future community projects should strive to document the level of meaningful interchange that took place after the intervention has occurred. It is possible that authors did not describe the next steps of their peer-reviewed papers, which may have been collaborative engagement with the community. Although we were unable to evaluate the extent of data storytelling present in the selected studies, this inclusion is an approach that would strengthen the use of any tools generated by community collaborations.

There are some important limitations of this scoping review. For example, we were only able to include articles that met our search criteria; given the relative newness of data visualization literacy as a health-related research topic, we may have missed important studies that did not fit our search criteria. Also, it is possible that some of the studies did not provide relevant details that would have led to their being categorized as more participatory, due to journal-imposed limitations of space in the methods. As noted, there are no guidelines for reporting on community engagement methods. Our approach to coding these studies using an established approach is an effort to explore a structure for this. Yet, there remains work to do to link the field of participatory research with creative data literacy. We believe that community-focused research in data visualization should incorporate more elements from creative data literacy, such as involving community members in the data collection and visualization process.

CONCLUSION

This scoping review examines the current state of community engagement in data visualization development in health and offers suggestions for future endeavors in empowering communities through data visualizations. While some notable aspects of creative data literacy are present in the included studies, there is still more that can be done to give the community an active voice in data storytelling. Approaches that prioritize communities’ cultures, languages, backgrounds, and needs may strengthen the communities’ responses to health messaging and improve the quality of health-related visualizations.

FUNDING

This research was supported by the Patrick J. McGovern Foundation (grant C-7000-112110-7030755-45) and by funds made available by the Centers for Disease Control and Prevention, Center for State, Tribal, Local and Territorial Support, under CDC-RFA-OT21-2103: National Initiative to Address COVID-19 Health Disparities Among Populations at High-Risk and Underserved, Including Racial and Ethnic Minority Populations and Rural Communities (grant A138216) through the San Francisco Department of Public Health. The findings of this research are those of the authors and do not necessarily represent the official position of or endorsement by the Centers for Disease Control and Prevention.

AUTHOR CONTRIBUTIONS

DC, JP, and MAH conceptualized the project and worked on data analysis. DC, JP, and MAH wrote the initial draft of the manuscript. MGS, MJB, and RK contributed to shaping the direction of the scoping review and to subsequent revisions of the manuscript. All authors agree to the published version of the manuscript.

SUPPLEMENTARY MATERIAL

Supplementary material is available at Journal of the American Medical Informatics Association online.

CONFLICT OF INTEREST STATEMENT

None reported.

DATA AVAILABILITY

Data available upon request.