Desperately Seeking Intersectionality in Digital Health Disparity Research: Narrative Review to Inform a Richer Theorization of Multiple Disadvantage

Introduction

Background

COVID-19 has thrown a spotlight on digital health disparities. Before the pandemic, patients with poorer self-reported health, of older age and lower incomes, and from certain minority ethnic groups were less likely to access health care by technological means [1,2]. The proportion of health care consultations conducted remotely (telephone, video, and web-based) increased dramatically during the pandemic [3], chiefly because many face-to-face consultations were canceled owing to the risk of transmitting the virus. Therefore, the shift to remote consultations affected a greater number of patients with increased access burdens because of being disadvantaged through poverty, low health literacy, limited English proficiency (LEP; in countries where English is the main language), or lacking digital skills or devices [4,5].

There is a vast amount of research on health disparities in general; this literature falls into 3 broad categories corresponding to 3 longitudinal phases. The first phase, detection, involves defining health disparities, identifying vulnerable populations, and developing valid measures for studying both. The second phase, understanding why disparities exist, involves identifying factors that explain gaps in health care between vulnerable and less vulnerable groups. The third phase involves the development, implementation, and evaluation of interventions that reduce or eliminate health disparities. These different kinds of research are all relevant to the study of digital disparities as well, although the literature on the latter is currently sparse in both volume and depth.

Digital health is sometimes presented vaguely and futuristically as having the potential to strengthen health systems and public health, improve efficiency, and increase equity in access to health services [6,7]. Video consultations in particular have been extensively researched (often in randomized controlled trials in comparison with face-to-face consultations) and depicted with promissory claims of delivering efficient care without compromising safety or patient satisfaction [8-10]. However, as the pandemic showed, digital solutions intended to reduce inequalities may actually widen them [11-13]. Broadly speaking, and with some notable exceptions [14-16], technologies are least frequently and least readily used by limited English–speaking communities, those with low income, and older adults—and especially by those in the triple jeopardy of all 3 groups.

Uptake of video consultations, for example, is known to be low among various disadvantaged groups [17]. This and other digital disparities have been explained by multiple factors, including lack of access to technology, low digital literacy, suboptimal internet coverage, and power differentials within the home in terms of who has access to digital devices [18-20].

Several recent publications have proposed strategies to ensure that the emergence of digital services does not exacerbate disparities in access to health care and health outcomes [11,21-23]. In total, 2 broad approaches have been taken. One approach speculates that digitally driven efficiency savings could free up staff to attend to the disadvantaged, who would continue to consult in traditional ways. However, there is limited evidence that such savings occur even with telephone consultation services and—to our knowledge—no evidence that they occur with video services (which have higher setup costs and require staff and patients to learn new skills) [24].

Another approach centers on identifying and removing barriers to video access among disadvantaged groups—for example, ensuring that people are equipped, competent, and confident in using the video modality where appropriate. This strategy is founded on an individual deficit model that depicts the disadvantaged as deficient in certain things (eg, knowledge, confidence, and bandwidth) and assumes that these deficiencies can be rectified by specific inputs (eg, training, practice, and digital upgrades). Thus, it tends to overlook the pervasive impact of multiple layers of structural disadvantage. Of particular interest is how key risk factors for digital exclusion such as LEP, poverty, and older age are often mutually reinforcing, an effect that some have called intersectionality [25].

These approaches are discussed throughout this paper, which starts by outlining the aim, scope, and research questions of the review along with definitions of important terms and concepts. We then explain our methodological approach to the review and the details of our methods. Our findings show that, in the relatively sparse literature uncovered on the topic, substantial digital disparities are reported and that this research to date in relation to video consultations has been almost entirely descriptive rather than explanatory. We also describe how we identified 3 candidate theories of multiple disadvantage that could further enhance our understanding of digital disparities. We conclude by proposing 3 candidate theories that may have particular relevance in explaining and helping address digital health disparities in people with multiple disadvantage.

Aim, Scope, and Research Questions

In this narrative review, we sought to explore how published studies of disparities in digital health consultations have defined, theorized, and empirically tested the concept of multiple disadvantage.

To sharpen our focus in a potentially vast field, we chose to restrict our sample of empirical digital health research to studies of video consultations between patients and health care professionals as this is where the most dramatic differences in digital access have been documented in the literature [8]. In contrast, there has been little research on digital disparities in electronic consultations [26], and research on telephone consultations suggests that digital disparities are less marked [27]. On the basis of findings from our previous work on remote consulting [28], we decided to focus particularly on studies that provided insights relevant to an underresearched group: older adults with low income and LEP.

Our research questions were as follows: (1) How have the intersecting effects of age, socioeconomic status, and LEP been conceptualized, theorized, and studied empirically in relation to digital consultations (especially video)? (2) What interventions have been developed and tested within the context of digital consultations to try to overcome the effects of multiple disadvantage? (3) What were the findings of these studies and how can they help us extend theory and inform future research? and (4) What are the implications for policy and practice?

Definitions

This review covers a number of closely related terms and concepts, which we define and discuss in Textbox 1.

Methods

We undertook a narrative review of the literature published from 2012 to 2022 focusing on digital health disparities in disadvantaged groups, with a specific focus on older, low-income, limited English–speaking individuals and on video consultations. Various combinations of search terms, including those regarding age, language, and income, were trialed as part of the initial search strategy but, because of the unique combination of terms being used, they yielded no relevant results. Following discussion with an expert librarian and coauthors, the search strategy shown in Textbox 2 was applied and updated with further terms as the study progressed to reflect the developing field. Although the search range extended back to 2012, only one study in our final sample was published before 2020. Earlier studies were clearly superseded by later work because of technologies having been developed at pace and the evolving field of digital health care. An evaluation of a video consulting service in Scotland by Wherton et al [41] reported that, as late as 2017, the platforms used for video consultations were designed for videoconferencing rather than video consulting and were expensive, clunky, unreliable, and poorly aligned with clinical workflows. A few years later, bespoke video technologies for health care encounters had been developed; they were cheaper, more agile, and better designed around key workflows. Accordingly, studies undertaken before the development of mature, fit-for-purpose technologies were less relevant. Similarly, when looking to the literature for explanations of digital health disparities, little had been published before 2012, and earlier studies reflected challenges that are no longer relevant today, such as website provision and public health dissemination through digital television [42].

We drew in particular on 3 methodological sources. First, we aligned with Greenhalgh et al [43], who highlighted the purpose of narrative review (to achieve clarification and understanding across a broader field of inquiry) and distinguished this from that of quantitative systematic review (to identify, summarize, and synthesize data on a narrowly focused topic, typically through meta-analysis). The latter relies on largely technical processes (eg, data extraction and the use of risk-of-bias tools), whereas the former requires the progressive development and refinement of an argument through interpretative methods.

Second, we engaged with the methodology by Boell and Cecez-Kecmanovic [44] for hermeneutic review, which applies the hermeneutic circle (progressively adding parts to the whole) to secondary research. The hermeneutic review begins with a close reading of key texts—those known to the research team and those identified on an initial scoping search. These texts are mapped and classified according to coherence, adequacy, and relevance (with relevant extracts as quotes) in interim summaries. Some summaries are deep dives on specific themes (eg, how a particular theory has been applied). Some are broader but less deep (eg, an early draft of the review findings to be progressively refined as the study unfolds). The researchers move back and forth between further searches and a progressively richer overall summary.

Third, we took note of methodological guidance from a group of journal editors in the health care field [45], who developed a structured critical appraisal tool (SANRA [Scale for the Assessment of Narrative Review Articles]). SANRA defines quality in narrative reviews in terms of strong justification for the importance and aims of the review, a well-described and well-justified literature search, claims backed up by referencing relevant primary studies, the quality of scientific reasoning (including a nonselective approach to inclusion of studies and study designs appropriate to the research question), and appropriate presentation of data.

The following search terms were developed by author 1 and discussed with the coauthors: remote consultations (or virtual consultations or video consultations or video visits or telemedicine or telehealth) and digital health inequality (or digital divide or inequity/ies or inequality/ies or health disparity/ies or disadvantage). These terms were then checked by the expert university librarian. Following some pilot searches, these were developed into formal search strings (Textbox 2).

The search was first completed in July 2022 and then supplemented with an additional search with revised search terms in October 2022. The results of all these searches were combined into a single data set, and irrelevant studies were excluded by screening titles and abstracts. All papers were reviewed by author 1, and each paper was second reviewed by at least one coauthor. Rigor was strengthened by reflexivity and the application of the Grading of Recommendations Assessment, Development, and Evaluation Confidence in the Evidence from Reviews of Qualitative Research checklist [46] to determine confidence in the findings.

To gain familiarity and aid data management, author 1 extracted and tabulated the following information from the included sources: (1) study design, setting, and sample; (2) key findings in relation to health inequalities and web-based care; (3) key issues raised by the authors; (4) authors’ recommendations for how to promote health equity in the context of remote care; and (5) any theories or frameworks used by the authors to study digital disparities.

Following the hermeneutic circle, we worked back and forth between individual papers and our overview of the findings, progressively adding detail. The process of interpretive synthesis was aided by discussions among the coauthors. When we realized that the initial data set of 27 papers included very little theoretical analysis, we conducted a further search for theories of multiple disadvantage that were potentially relevant to digital health disparities and had the potential to add a theoretical depth to our original sample of papers. This second search was deliberately not exhaustive; it included asking experts in the field, using key sources known to the authors, and searching the PubMed database for the term theor* along with the selected terms listed in Textbox 2.

A near-final version of the synthesis was shared with a wider group of peers, including experts in various aspects of health disparity or digital health, and with laypeople with lived experience of accessing care. The synthesis was refined in response to their feedback.

Results

Description of the Data Set

In this section, we describe the main findings of the empirical studies included in this narrative review (Table 1) as well as key points from commentaries on the theme of inequity of access (Table 2). All the authors of both commentaries and empirical studies offered a list of proposed solutions to digital disparities. These are further summarized and categorized (Textbox 3). Figure 1 presents a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flowchart demonstrating the number of papers identified, included, and excluded.

The 17 peer-reviewed research papers (summarized in Table 1) comprised 1 (6%) systematic review [48], 1 (6%) narrative synthesis [49], 7 (41%) retrospective audits of medical records [56,57,59-62,70], 3 (18%) quantitative surveys [5,50,58], and 5 (29%) qualitative studies based on semistructured interviews [5,51,52,54,55]. A further 37% (10/27) of the articles (Table 2) commented on others’ research, reflected on findings from clinical practice, or proposed measures to reduce digital health disparities [11,21,22,63-69].

‎View this figure

Descriptive Findings

Formal research studies on this topic in the early months of the pandemic were few in number, mostly small in size, and conducted rapidly. As such, our data set included no in-depth, theoretically informed empirical studies that had set out to explain how different dimensions of disadvantage combined to affect digital health disparities. During 2020, several clinical authors were moved to write urgent editorials and commentaries on the theme of inequity of access as frontline services shifted from face-to-face to remote modes. The summary of the review findings using the Grading of Recommendations Assessment, Development, and Evaluation Confidence in the Evidence from Reviews of Qualitative Research checklist [46] to determine confidence can be found in Table 3.

Systematic and Narrative Reviews

Studies published in late 2020 and early 2021 included rapid systematic reviews comparing the impact of face-to-face and remote consultations on a range of predefined variables. Parker et al [48], for example, reviewed quantitative studies from before the pandemic to mid-2020 across socioeconomic and disadvantaged groups in the United Kingdom. A total of 13 studies met their inclusion criteria, and they found that phone consultations were used more by young people of working age, nonimmigrants, and women. Asynchronous web-based consultations were used more by affluent and educated people. Findings in relation to socioeconomic status and ethnicity were inconsistent across primary studies, and none of the included studies reported on quality of care or clinical outcomes.

In a rapid review and narrative synthesis of UK primary care by Litchfield et al [49], 9 studies were identified that explored various constructs within three domains of the digital divide: (1) digital access, within which one study described continuing issues with internet connectivity among vulnerable patients in the United Kingdom; (2) digital literacy, where 7 studies described how ethnic minorities and older adults were less likely to use digital technologies for accessing care; and (3) digital assimilation, where one study described how video technologies can reduce feelings of isolation and another described how older Black men were the most likely group to share information about COVID-19 on social media platforms. This review also found a large number of opinion pieces and editorials on digital disparities.

Quantitative and Mixed Methods Studies

Nouri et al [22] audited the uptake of remote consultations by demographic group in 3 clinics at an academic medical center in San Francisco. Their paper included a literature review and empirical data from the authors’ own 2 general practices. They showed that the shift to remote consulting had been associated with a decrease in the number of consultations with older adults, those of low socioeconomic status, low–health-literacy groups, limited-English speakers, and Black and Asian minority ethnic groups. The authors offer a framework for addressing inequities, which includes goals such as improving digital literacy and resource barriers, removing health system–created barriers, and advocating changes to support sustained and equitable access.

Rodriguez et al [56] also found lower use of video versus telephone visits among older Black, Hispanic, and Spanish-speaking patients, which extends previous findings that showed decreased telemedicine use among patients with LEP [70]. This study also found that clinicians and practices largely drove this variation in the use of video versus telephone visits, suggesting an important target for intervention.

Hsueh et al [57] similarly found that patients with LEP chose video consultation options less often than those without LEP. However, they also found that, once patients with LEP had video visit use experience, they were not different from patients without LEP in likelihood to reuse video visits.

The survey of small primary care providers in New York City by Chang et al [50] uncovered the percentage of encounters undertaken by telephone, video, or web-based patient portal or face-to-face as well as barriers to remote consultations. A key finding was that the higher a patient’s social vulnerability index, the more likely the consultation occurred by phone instead of video. This was similarly echoed by Broffman et al [61] and Zachrison et al [62], who also found that patients of color, who have historically experienced the greatest access disparities, are significantly more likely to use smartphones to access telehealth compared with a computer. This provides insight into the extent to which low-bandwidth telehealth is accessible under certain conditions.

Although Chang et al [50] conducted a relatively small survey with a low response rate (exact figures not given), their findings align with those of other studies. For example, a retrospective case note review of 80,000 American patients from March 2020 to May 2020 who had completed a telemedicine visit (from a total of 140,000 who had scheduled such a visit) found that, overall, 46% of teleconsultations were by video and 54% were by phone [70]. Factors independently associated with fewer completed telemedicine visits included older age, preference for languages other than English, Asian ethnicity, and Medicaid insurance (indicating low income). Higher rates of telephone consultations were associated with female sex, Black or Latino ethnicity, older age, and lower household income. This was also the case with the study by Schifeling et al [59], where more than half of the older patients did not use video visits, especially if they were from racial or ethnic minority backgrounds or Medicaid beneficiaries, and with the cross-sectional analysis by Sachs et al [60] showcasing seniors, non-English speakers, and Black patients to be more reliant on telephone than video for care. Although these large studies provided important quantitative information about individual risk factors for a failed telemedicine visit (and for a telephone visit over a video visit), they were not designed to explore the interaction between different independent variables. Further theoretical explanations of how different risk factors, individually and in combination, contributed to digital inequities in certain vulnerable population groups are necessary to fully understand these data.

Similarly, a mixed methods study based in the United Kingdom by Fu et al [5] found that there was a reduction in video consultations in older users, undocumented migrants, and individuals with poor health, which could mean that those in the greatest need were being excluded, perhaps because of the digital divide evidenced in some groups of migrants before the pandemic. Yu and Hagens [58] echoed similar findings in older adults in Canada, with results highlighting socioeconomic disparities among older adults that could potentially explain this trend, including lower income and education levels that may act as barriers for older adults to acquiring the skills and technologies necessary to use more complex solutions such as video. As Fu et al [5] stated, “those in the greatest need of health care appeared to be less able to access remote services.”

Qualitative Interview Studies

The qualitative rapid health needs assessment by Steven et al [51] across the United Kingdom found that all groups studied experienced challenges in accessing and following COVID-19 information and government guidance, attributed variously to lack of access to digital technology, lack of translated resources, absent or inadequate tailored support, and inadequate housing. Changes in the organization and delivery of health care services, including closure of outreach and drop-in services, remote consultations, and web-based patient registration, worsened existing barriers to accessing health care.

The semistructured interview study by Kaihlanen et al [52] sought to explore and explain the challenges related to the use of digital health services in Finland through the lens of a digital health equity framework. They found that access to digital health services was hampered by insufficient digital or local language skills.

Knights et al [53] found that digitization and web-based consultations appeared to have amplified existing inequalities in access to health care for many migrants in Finland because of a lack of digital literacy and access to technology compounded by language barriers and indirect discrimination (eg, telephone-only booking services become inaccessible to those without a phone). Health care professionals perceived low digital literacy among migrants and were concerned that web-based consultations resulted in difficulties building trust and risked missing safeguarding cues. These semistructured interviews were conducted by phone, and the sample of migrants was small and skewed; the study was not designed to capture individual, contextualized narratives, and findings were largely impressionistic.

The qualitative study by Alkureishi et al [54] of patient perspectives on the digital divide in US primary care settings explained the concept to participants as follows: “there are people that have and can use technology like computers and the internet. But there are also people that do not have or cannot use this kind of technology. So there is a split or a divide, between people that have and know how to use technology, and those that do not” [61]. These authors found that patients were very aware of the digital divide and described the impacts beyond health care, including employment, education, community and social contexts, and personal economic stability. These participants viewed access to technology and digital skills as important influencers of health disparities.

The clinician-patient relationship was a theme in a qualitative study of patients’ and clinicians’ experiences with video consultations in general practice in Lothian, Scotland [55]. Although the sample size of this study was comparable with that of the study by Knights et al [53], purposive sampling was used in this study to include both sexes, a range of ages and socioeconomic statuses, and those with and without technical problems during their video consultations. The study stated that although the demand for video consultation services in primary care is likely to rise, for complex or sensitive problems that require touch, face-to-face consultations remain preferable for patients and clinicians.

Commentaries

A commentary from 2 directors of free clinics in the United States described how their patients were unable to access their web-based system and offered the solution of a combination of technology and face-to-face services to help address some of these disparities [63]. Their discussion highlighted various upstream societal and social factors (such as mistrust of technology, internet availability regionally, and housing insecurity, to name a few) that were being exposed across hospital systems in the country at a critical time in a public health crisis with no measures in place to address them.

Mehmi et al [64] further discussed how they expected video consultations to increase health inequality if the correct infrastructure is not put in place. Thronson et al [65] considered the underlying cause of this “pandemic of health care inequity” [50] to be access—the disadvantaged simply cannot access telemedicine or home monitoring tools. Ramasawmy et al [66], by contrast, highlighted 3 areas in which existing knowledge and evidence can be translated into cross-sectoral action to avoid further ethnic and digital health inequalities: data and measurement, improved communication, and embedded equality impact.

Gray et al [67] offered 5 strategies to prevent the exacerbation of health disparities for low-income, rural, disabled, ethnic minority, and older adult populations in the United States. They considered sociocultural barriers to digital inclusion, including limited digital skills, low health literacy, disability, low income, and LEP, and structural barriers such as geographic isolation, broadband capacity, and technical hardware.

Crawford and Serhal [21] offered a new digital health equity framework to identify the digital determinants of health and their links to digital health equity, which requires additional evidence and empirical application. Gallegos-Rejas et al [69] proposed practical steps to reduce the digital divide and encourage equitable access to telehealth through improvements in digital health literacy, workforce training in clinical telehealth, co-design of new telehealth-enabled models of care, change management, advocacy for culturally appropriate services, and sustainable funding models.

Another 20% (2/10) of the commentaries, by Rodriguez et al [11] and Eruchalu et al [68], also explored strategies for the digital care of vulnerable patients during the pandemic and further discussed concerns about inequities in digital health access. They reiterated the finding that ethnic minority patients had significantly lower chances of attending telemedicine visits because of inequities in broadband access, lack of available technology, and mistrust of health care professionals.

All the authors of the aforementioned studies offered a list of proposed solutions to digital disparities. These are summarized and categorized in Textbox 3.

Although the changes proposed in Textbox 3 have some face validity, they remain largely untested.

A reviewer of a previous draft of this paper suggested that the taxonomy offered in Textbox 3 (which was our own way of making sense of the data we extracted from papers in our sample) reflected a particular theoretical perspective, namely, the social-ecological framework, which “considers the complex interplay between individual, relationship, community and societal factors” [71]. We agree that this lens could potentially provide an overarching framework within which to synthesize middle-range theories in a future paper.

Our second search identified 3 candidate theories that helped explain the effects of multiple disadvantage identified in our data set: fundamental cause theory, digital capital theory, and intersectionality theory, all of which are discussed in the next section.

The fundamental cause theory and intersectionality theory were the most common theories cited in general health disparity research; digital capital theory was mentioned in studies and commentaries in digital disparity research. Various other theories recurred in the literature but were unhelpful in analyzing our data set. We found the aforementioned theories helpful as (1) intersectionality worked as an overall guiding principle for understanding how people’s lives and characteristics stem from and lead to multiple axes of disadvantage; (2) digital capital theory helped us understand how these axes of disadvantage played out in terms of access to and use of digital resources; and (3) fundamental cause theory sensitized us to the pervasive impact of poverty, which operates through multiple intermediate mechanisms.

Discussion

Principal Findings

Findings from our narrative review of digital health disparities in relation to video consultations highlighted that the available literature reports substantial digital disparities. Formal research studies on this topic in the early months of the pandemic were few, mostly small, and rapidly conducted. Research in relation to video consultations to date has been almost entirely descriptive, and our data set included no in-depth, theoretically informed empirical studies that were able to explain how different dimensions of disadvantage combined to affect digital health disparities.

Our narrative review, which focused on video consultations, produced 3 key findings in particular. First, the literature was sparse, comprising only 7% (2/27) of reviews and 63% (17/27) of empirical studies, most of which were published since the start of the COVID-19 pandemic in 2020. Of these studies, most (25/27, 93%) were relatively small and undertaken quickly and under pressure during the pandemic, for example, qualitative studies that comprised one-shot semistructured interviews on convenience samples.

The second finding is that, despite the limitations of the literature, substantial digital disparities were reported [5,22,50-53,56,57,59]. Low household income, older age, and ethnic minority background (especially limited-English speakers) were all independently associated with lower uptake of video consultations [5,50,52,53,58,60,61]. Proposed explanations include lack of digital devices and infrastructure, low health and digital literacy, and inability to understand written resource materials [11,65,66,68]. The disparities found were sometimes dramatic and contrasted strikingly with studies of video consulting undertaken before the pandemic. These studies had framed this as an innovative service model that might increase service efficiency.

The third major finding was that research on digital health disparities, at least in relation to video consultations, has to date been almost entirely descriptive rather than explanatory. All the quantitative studies (audits and surveys) in our data set were designed to generate knowledge of the association between particular patient characteristics and the uptake and outcome of video consultations. Although such knowledge is essential in identifying a problem, there is a risk that such studies reduce the complex and interacting aspects of disadvantage to simple variables. Variable-focused research has been criticized by social scientists for oversimplifying context, removing key content (eg, unmeasured variables), overlooking historical path dependencies, and failing to explore how different variables combine and unfold over time to produce complex and sometimes unpredictable outcomes for individuals [72,73]. Explanations generated from such studies and from the superficial and qualitative designs included in our data set tended to couch findings in terms of barriers and enablers, which were depicted as having more or less fixed effects (negative and positive, respectively) on outcomes. The result is a body of literature that is desperately in need of theorization; for example, none of the 63% (17/27) of empirical studies cited any theory of disadvantage, and none of the well-intentioned ideas listed in Textbox 3 are couched in a well-developed theory of change.

Descriptive, variable-centered research is common when studying health disparities as it allows the so-called social determinants of health (eg, income, education, and gender) to be manipulated by quantitative techniques such as aggregation and correlation. However, such approaches are inherently problematic as they require research participants to be placed into categories that can then be manipulated as variables (older adults, those with low income, and ethnic group X), and findings tend to be presented in terms of what has been called “single-axis” analyses [25].

As Zheng and Walsham [74] have argued, these notions and categorizations do not consider the multifaceted and complex interplay of factors that contribute to digital disparities, nor how a characteristic that disadvantages an individual in one setting may have little adverse impact or even a positive impact in another. In terms of multiple disadvantage, for example, limited-English speakers are more likely to be older and lack basic digital devices and skills [75]. Although these factors may in some cases be mutually reinforcing, some older adults from some minority ethnic groups may be more likely than White British older adults to live in intergenerational households with good internet connection and family member support—hence, a characteristic (non-White ethnicity) that acts as a barrier in one setting may act as an enabler in another.

In sum, the current literature on digital health disparities is not only sparse but also in need of richer theorization to generate explanations of how different dimensions of disadvantage interact. We have argued elsewhere that the overemphasis in evidence-based medicine on empirical research at the expense of explanatory theory on the causes of phenomena (what some have called EBM+) may produce impoverished findings [76]. This builds on earlier work emphasizing the crucial importance of theory in selecting which hypotheses to test and how when studying disparities [77,78].

We have begun to explore the wider literature to identify theories of multiple disadvantage that have a potential bearing on digital health. In the following sections, we consider the 3 most relevant theories that emerged in our search to date: fundamental cause theory, digital capital theory, and intersectionality theory.

Fundamental Cause Theory

Link and Phelan [79] define a fundamental cause of health disparity as anything that involves resources that influence the extent to which people are able to avoid risks of mortality and morbidity. Socioeconomic status operates as a fundamental cause as it (1) involves access to resources (in particular, wealth, income, education, and racial privilege) that allow individuals to avoid diseases and their consequences and (2) affects multiple risk factors (eg, health literacy, quality of medical care, and diet) and disease outcomes that change over time. In short, those with financial resources and high social status can use these resources to avoid disease, seek treatment, and adopt healthy behaviors. The higher risk of heart disease in those of lower socioeconomic status, for example, can be explained by a combination of less access to money, health care, healthy food options, opportunities to exercise safely, and social support. These fundamental upstream disparities operate through multiple mediating factors at both the individual level (eg, diet, physical activity, and attendance at screening programs) and the metabolic level (eg, cholesterol, blood glucose, and stress hormones). An intervention that successfully changes one risk factor (such as BMI) will have limited impact as the fundamental cause will still operate through other mediating factors. This theory is often invoked when authors talk of the structural determinants of health disparities [80-82]. Although human behavior, lifestyle choices, knowledge, and beliefs may mediate the link between social determinants and adverse health outcomes, these factors are inadequate as explanatory causes of disease.

Fundamental cause theory offers a plausible explanation for the powerful and persistent link between multiple disadvantage and digital disparities. Applied to digital health disparities, it would depict the fundamental cause of these disparities as low socioeconomic status and that this cause operates through flexible resources such as access (or lack thereof) to money, knowledge, power, prestige, and beneficial social connections. A key hypothesis based on this theory is that addressing any one proximal cause—for example, by supplying a person with low income with a digital device—will not solve the fundamental problem as, although this intervening mechanism may change one risk factor (money in this case), it may have limited impact overall as the fundamental cause will continue to cause disparity through other mediating factors such as lack of knowledge (ie, not knowing how to use the digital device given).

Digital Capital Theory

Bourdieu [83] applied the idea of capital to signify the internal (eg, abilities and attitudes) and external (possessions and attributes) resources that people mobilize to achieve their goals in social life. He highlighted cultural capital as a form of capital that can be accumulated and transformed into other capitals. Digital capital is an extension by Ragnedda and Ruiu [84] of the theory of cultural capital by Bourdieu [83], made up of “both digital competencies and digital technologies.” They argue that digital capital is a form of capital in its own right and is essential for building social, economic, and cultural resources in the digital world that we live in today. Disparities involving digital skills originate in inequalities of access but are mediated by orientations that can only be understood in relation to total life contexts (eg, education, income bracket, age, location, and social support all influence a person’s access to digital technologies and the level of digital skills that they can acquire) [85]. Digital capital is a relatively new concept that scholars have begun to explore empirically using various methodological approaches [86,87]. Digital capital may be estimated, for example, at the individual level by assessing a person’s digital literacy and skills, at the organizational level by measures of digital infrastructure (including the digital competence of personnel), and at the locality level in terms of the quality of the area’s IT infrastructure.

Digital capital theory points us to the hypothesis that traditional forms of capital (such as economic, cultural, and social capital) are converted into digital capital and vice versa and provides the conceptual tools to examine how and to what extent this occurs, thereby illuminating how social inequality relates to digital inequality. If digital spaces—because of social inequality and underlying power structures—become increasingly stratified, there will be significant impacts on how individuals from differing backgrounds gain accumulated forms of capital through the digital realm. In other words, digital capital theory seems to offer an explanation as to why people who already experience health and other disparities find that these disparities widen when services are digitized.

Intersectionality Theory

The Black feminist scholar Kimberle Crenshaw [25] critiqued traditional studies of Black women’s oppression for offering what she called “single-axis” analyses focusing on either race or gender but failing to integrate the 2 categories. Subsequent authors have extended the original concept of race-gender intersectionality by Crenshaw by adding categories, including nationality, class, age, sexual orientation, and disability [88,89], revealing “crosscutting and mutually reinforcing systems of domination and subordination” that “may construct multiple, uneven and contradictory social patterns” according to Anthias [90]. Intersectionality has been invoked to explain disparities in outcomes within minority ethnic groups in the context of the pandemic [91]. Intersectionality has been studied in many different ways [39]. Most relevant to our own data set is what we call lived-experience intersectionality research, which seeks to elucidate (through qualitative methods such as narrative interviews, ethnography, and arts-based methods) the complex and unique experiences of individuals whose identity crosses the boundaries of traditionally constructed groups [40,89].

Intersectionality theory applied to digital health disparities suggests the hypothesis that each individual’s identity and lived experience is unique and multifaceted and that individuals will use (or will not use) digital services based on their own unique identity and circumstances—rather than as members of a single category such as asylum seeker, Black individuals, or older adults. This theoretical approach would support detailed small-scale case studies to see how different aspects of disadvantage interact in individual lives.

Strengths and Limitations

Although our narrative approach to this review allowed for a comprehensive overview of the wide range of literature spanning digital health disparities, there was no evaluation of selected articles for validity. Nonetheless, quality was not jeopardized as the methodology by Boell and Cecez-Kecmanovic [44] for hermeneutic review, an explicit methodology and accepted standard, was used alongside other methodological guidance for quality judgment. Narrative reviews are also often criticized for the subjective weighing of the studies chosen for the review. We sought to mitigate this through discussion between coauthors and the study team for an investigation focused on using remote care as part of a wider study [28]. To account for any additional selection bias on the authors’ part, all the included studies were second reviewed for relevance.

A major limitation of the review was the sparse and undertheorized nature of the primary literature. We tried to remedy this by beginning to explore the wider literature to identify theories of multiple disadvantage that have a potential bearing on digital health. It should be noted that the 3 theories that we describe in the discussion are not a result of an exhaustive search, and there may be others of relevance. We plan to develop this stream of theoretical research in a future paper.

Conclusions: Suggested New Research Directions

Studies published since the COVID-19 pandemic began have shown that the move to digital forms of access and care provision has widened health disparities.

This recent literature contrasts strikingly with research on digital health services undertaken before the pandemic, which was largely focused on demonstrating noninferiority of digital modalities in terms of acceptability, safety, and subsequent use of services, often using randomized controlled trials with highly selected samples (stable, compliant, digitally equipped, and digitally confident patients recruited mostly from outpatient settings). The denominator population for digital health research expanded rapidly when such services became the default option for everyone for infection control reasons [92], revealing the (previously largely hidden) problem of wide digital disparities linked to multiple aspects of disadvantage. However, as this review has shown, in-pandemic research to date has been descriptive and superficial and has revealed few insights into how digital disparities emerge and play out both within and across categories of disadvantage.

This lack of attention to multiple disadvantage in digital health research to date represents both a unique opportunity and an important challenge for us to engage more curiously and theoretically with this core subject matter. We now need to turn our attention to specifically developing and using interdisciplinary theories of health disparities and technological innovation to inform our studies. Theories, including (but not limited to) fundamental cause theory, digital capital theory, and intersectionality theory, can provide a distinctive foundation for digital health inequality research and serve to guide ongoing research on this topic area. We suggest 3 complementary empirical approaches, as evidenced by previous studies that have conceptualized the aforementioned theories of health disparities.

First, quantitative studies using electronic patient record data should move beyond the current focus on single-axis analyses framed around a reified notion of the digital divide to produce category-focused intersectionality research. Such studies would need to be large, prospective, and hypothesis-driven to explore questions about the interaction between different categories of disadvantage; for example, how do gender, education, and ethnicity influence digital disparities in older people? The Digital Health Equity Framework [21] or the eHealth Equity Framework may provide a frame to think comprehensively about multifaceted approaches [93]. Using the eHealth Equity Framework approach in an initial scoping exercise, for example, can illuminate the proximal factors that need to be incorporated to address health inequities while also drawing attention to possible unintended consequences through distal interactions.

Second, qualitative studies of disadvantaged patients’ digital access and experiences should move beyond the one-shot, theory-free semistructured interview on a convenience sample to achieve richly theorized, in-depth longitudinal studies of lived experience in maximum-diversity samples within particular categories of intersectionality. As noted previously, extended narrative interviews and ethnographic techniques should focus on the unique and particular experiences of individuals. Through rich description and the use of literary devices (eg, metaphor and dramatization), such lived-experience studies will reveal how multiple intersections play out over time in members of a particular broad category of intersectionality (eg, people who have low income, are older, and with limited English and multiple health needs) and also, importantly, illustrate the wide diversity of experiences within that group.

Third, based on the findings of this review as well as on the solutions to digital disparities proposed by the authors of primary studies and commentaries, we suggest adding a co-design component to research. Heard et al [40] proposed adapting lived-experience studies to inform the design of inclusive policies and interventions that can take account of the multiple social, cultural, and political contexts within which individual lives are lived and choices are contemplated. They cited the Intersectionality-Based Policy Analysis Framework as a tool to inform the design of such approaches [94]. The framework provides guidance and direction to address the challenges of health inequities across diverse populations in 3 ways. First, it provides an innovative structure for critical policy analysis. Second, it captures the different dimensions of policy contexts, including history, politics, everyday lived experiences, diverse knowledge, and intersecting social locations. Finally, it generates insights, knowledge, policy solutions, and actions that may not be fully captured through equity-focused policy frameworks. This systematic approach will help in designing policy responses that mitigate instead of increase the potential unequal effect of this phenomenon.

Health disparities are already wide and (in many countries) increasing. As society becomes increasingly digitized, the problem is likely to escalate if intersectional disparities are overlooked. This paper, which is intended as the starting point for a wider debate, has outlined a novel and ambitious research agenda. We invite others to help address and extend it.