Theories Informing eHealth Implementation: Systematic Review and Typology Classification

Introduction

In recent years, technological innovation in health care has developed exponentially, and eHealth is now widely viewed as a significant potential contributor to improved quality of care [1,2]. However, despite much policy-level and scholarly discussions of triggering a revolution in health service delivery, problems of implementation and uptake of eHealth among both patients and service providers persist [1,3,4].

Poor uptake of eHealth (a term with contested definitions [5] but, broadly speaking, “health services and information delivered or enhanced through the internet and related technologies” [6]) is often explained in terms of barriers and facilitators [1]. In a recent study, Schreiweis et al [7] identified 77 barriers and 292 facilitators in implementing eHealth services. Similarly, a systematic review by Granja et al [8] identified 27 factors that determine the success or failure of eHealth interventions. Although studies about barriers and facilitators are important, they tend to fall short of capturing the complexity of the implementation process and the multiple interrelated factors that determine the translation and uptake of eHealth [1,9].

Evidence suggests that theory-informed approaches to implementation science can enhance the translation and acceptance of eHealth into clinical care [1,10-18]. Theories offer explanatory frameworks and formal heuristic devices that have the potential to move beyond the basic listing of individual facilitators and barriers to implementation, to capture the dynamic interaction between them [1]. As Damschroder [19] notes, theory “enables knowledge to emerge out of seeming chaos,” facilitating exploration of complex relationships and interdependencies between variables that unfold in diverse and changing contexts [20]. This is of paramount importance in eHealth settings [1,18], which are characterized by a complicated interplay between patients, clinicians, the health care system, and the eHealth technology.

Many theories and models have been articulated to inform and explain eHealth implementation [15]. Despite this abundance, findings from several reviews show that only a small number of select theories have been used repeatedly across multiple publications and by several authors [21-24]. For example, a recent review by Harst et al [23] of 24 studies of end-user acceptance of telemedicine found that 2 theories accounted for 20 instances of theory use: the technology acceptance model (TAM) and the unified theory of acceptance and use of technology (UTAUT). Similarly, a review on the use of theory in eHealth weight management interventions by Willmott [24] identified 18 studies referencing a theory, of which 16 mentioned either the social cognitive theory or the transtheoretical model.

Theories most commonly used in the literature tend to emphasize individual factors, such as motivation, attitudes, and behavior, rather than the broader social and environmental factors impacting implementation [21,22,25]. This is despite evidence highlighting the multilevel nature of technology implementation in health care and the importance of targeting variables at different levels [1,26]. As Glanz and Bishop [22] noted, social and environmental factors may constrain individuals’ behavior even when they are highly motivated. Therefore, the authors recommend complementing individually oriented theories with theories of social, policy, or organizational change [22].

One hindrance to this is that the current eHealth implementation literature is fragmented across multiple specialty areas and disciplines, making it difficult to locate the range of theories available [27]. To improve the selection and application of theory, it is necessary to identify an array of theories, across diverse disciplines, that have the potential to inform eHealth implementation. A further issue is that many theories contain overlapping constructs but use different terms to describe them [26]. Synthesizing theories according to their similarities would facilitate their selection and application at different levels [27].

To address these issues, we conducted a systematic review and classification of eHealth implementation theories. The review aims to address the following question: “What theories exist across disciplines that have been used to inform or explain eHealth implementation?” Theories identified by our review were classified using the typology by Sovacool and Hess [28] for theories of sociotechnical change. This typology provides an accessible and useful framework for organizing and selecting diverse theoretical options that target variables at different levels. Its use also allows the identification of areas where further theoretical development is required.

Methods

Overview

This systematic review was conducted by members of the review team in accordance with the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines [29]. A PRISMA checklist is available in Multimedia Appendix 1 [29]. The authors adopted a flexible approach by continuing to apply the core principles of systematic review methodology but tailoring the PRISMA guidelines to the needs of this review [30]. As such, formal quality assessment was not conducted for this review, as the perceived validity or trustworthiness of the included studies did not address the overall research question, which sought to identify the existence of theories across a broad and varied body of literature.

Search Strategy

Electronic searches of PsycINFO, MEDLINE, Embase, CINAHL, Scopus, Sociological Source Ultimate, Web of Science, ABI/INOFORM, EBSCO, and ProQuest Databases were conducted by the review team in June 2019 to identify studies that applied one or more theoretical frameworks to inform eHealth implementation. For this review, implementation was defined as “the scientific study of methods to promote the systematic uptake of research findings and other EBPs (evidence-based practices) into routine practice, and, hence, to improve the quality and effectiveness of health services.” [31]. These databases were chosen because they were deemed to be likely to catalog studies and disciplines relevant to the eHealth context and the specific research question. The search was limited to studies published in the last 10 years (from 2009 to June 2019) and yielded 21,704 abstracts for initial consideration. A full list of key search terms used can be found in Multimedia Appendix 2. All records were converted into an EndNote library and reduced to 13,101 following deduplication. Papers were then title-checked for relevance to the topic and research questions and aims before further screening by 2 independent reviewers (MH and HW) in accordance with the detailed inclusion and exclusion criteria outlined below.

Eligibility Criteria

Individual studies were included in the review if they were (1) published in the last 10 years (from 2009 to June 2019), (2) published in English, (3) outputs of empirical research or theoretical papers reporting on one or more theories in the context of eHealth implementation (this included all study types and populations), or (4) published in a peer-reviewed journal. Studies were excluded if they were (1) published before 2009, (2) not written in English, (3) studies that did not report on one or more theories applied in the context of eHealth implementation, (4) gray literature not published in a peer-reviewed journal, (5) dissertations, theses, conference proceedings, or abstracts, or (6) any form of literature review. The full eligibility criteria for this review are provided in Textbox 1.

Identification and Selection of Studies

A total of 2 reviewers (MH and HW) independently applied the predefined inclusion and exclusion criteria to screen for relevant studies from those obtained through database searching. To ensure accuracy, record titles and abstracts were screened manually in EndNote, and documents that did not meet the selection criteria outlined above were excluded by the reviewers. Following 2 rigorous rounds of title and abstract screening, full texts of all potentially eligible studies were examined and further screened by the 2 independent reviewers (MH and HW) using the Covidence web-based software (Veritas Health Innovation Ltd), an effective tool for assisting research teams when performing systematic reviews or meta-analyses [32]. Articles that failed to meet the selection criteria were excluded and then cross-checked to ensure transparency and accuracy surrounding the reasons given for exclusion. Any conflicts in decision making during the screening phase were resolved via discussion between reviewers or, if needed, with the research coordinator (FKL) until consensus was reached.

Data Extraction and Presentation

As the standardized extraction tool in Covidence did not meet the specific needs of this review, a modified extraction form was developed and piloted by the 2 reviewers (MH and HW) with 10 included studies tabulated and refined accordingly. The modified extraction form was tailored to include characteristics relevant to the research question. The characteristics extracted by the reviewers included (1) name of theory, (2) description, (3) instances of theory use, (4) examples of theory application, and (5) theory type. Instances of theory use refer to the number of occurrences in which a theory was used. As several studies used more than one theory, the total number of theory instances exceeded the number of papers included in the review. Examples of theory application were drawn from the literature to specify how each theory informed eHealth implementation. The reviewers then determined each theory type by drawing on the typology by Sovacool and Hess [28] for theories of sociotechnical change. This typology categorizes theories according to where they tend to center their analysis. The term center is intended to convey that a theory may involve elements of multiple types but that it approximates one ideal type above all. This typology was considered a useful tool for ordering and analyzing the diverse theoretical approaches identified, as a basis for future theory building [33].

The typology includes 5 categories: agency, structure, relations, meaning, and norms. Agency-centered theories relate to people’s individual actions, beliefs, and attitudes, and assume that these can be explained without deeper consideration of broader social and systemic elements [28,34,35]. In contrast, structural theories propose that people are influenced largely by external forces beyond their control, such as their organizational, political, or macrosocial environments [28,35]. Relational theories attempt to interpret the interactional processes that influence the circulation of knowledge throughout different social networks. They view technology and society as coproduced and coconstructed, with no single dimension creating change by itself [28,36]. Meaning-centered theories focus on language, ideas, symbolism, narratives, rhetorical visions, and other cognitive dimensions that both orient action and are changed by it. Normative theories offer criteria by which to assess the positive or negative impact of technology on society or on a specific group. A sixth category, combined theories, was added to these 5 categories. This included meta-theories that explored a combination of individual, structural, or relational frameworks. All authors (MH, JW, JC, HW, CT, and FKL) reviewed and agreed upon the classification of theories using this typology.

Results

Search Results

The electronic search of key databases resulted in 21,704 potentially eligible articles (Figure 1). This number was reduced to 13,101, following deduplication. Of these, 12,001 papers were excluded based on title screening and application of the eligibility criteria previously outlined. Key reasons for exclusion of papers at title screening included eliminating those that were in non-English language or those that reported on an irrelevant topic to the research question, for example, non-eHealth or theory-related papers. The abstracts of the remaining 1100 papers were then independently screened by reapplying the inclusion and exclusion criteria, and a further 935 papers were excluded. Key reasons for exclusion at abstract screening included nonempirical or gray literature and papers that reported abstracts or protocols only. Following a full-text review of the remaining 165 articles, an additional 46 articles were excluded because of insufficient reporting on or mention of theories related to eHealth implementation. In total, 119 articles met the full, predefined eligibility criteria and were included for data extraction and synthesis of findings. The PRISMA flowchart in Figure 1 details the process of eligibility and study selection.

‎View this figure

Theory Summary and Classification

The summary (including theory name, description, instances of theory use, and examples of application to implementation) and classification of all theories used to inform and explain eHealth implementation is provided in Table 1. In total, 36 distinct theories were identified. Classification of these theories using the typology by Sovacool and Hess [28] showed that the theories used in the literature were predominantly agency centered (19/36, 53%), followed by relational (7/36, 20%), structural (6/36, 16%), meaning (3/36, 8%), and combined theory types (1/36, 4%). No normative theories were identified.

In total, 53% (19/36) of theories were classified as agency centered. Individual theories that occurred most frequently in the literature were the TAM by Davis and Venkatesh [76] (33 instances), UTAUT by Venkatesh [77] (32 instances), and Diffusion of Innovations Theory by Rogers [78] (16 instances). These theories were found to be primarily concerned with the individual and interpersonal elements of eHealth implementation. Although they did, to some extent, appear to consider the influence of organizational and social factors on eHealth adoption, individual attitudes, behaviors, and motivations remained the core focus of theoretical analysis. Theories classified as individual examined the adoption of eHealth either before or soon after the implementation of an intervention. However, they did not emphasize any form of user involvement in the development of an intervention. These theories tended to depict adoption as a temporally discrete and relatively immediate event, rather than as one stage in a larger multistage process. They often focused on what people were going to do soon, a decision they are about to make, or a behavior they need to alter. The diffusion of innovations theory provides an exception, as this theory considers time to be an essential factor influencing adoption [79].

A total of 20% (7/36) of theories identified in the literature were classified as relational. Of these, the normalization process theory (NPT) by May et al [80] occurred most frequently in the literature (17 instances), followed by structuration theory (ST) [81] (6 instances) and actor-network theory (ANT) [82,83] (4 instances). Sociotechnical systems theory, social information processing theory, social worlds theory, and boundary object theory occurred only once each in the literature. Relational theories emphasize social relations and interactions at the human-technology interface. They highlighted the complex networks of social structure and meaning in which people are embedded, proposing that the translation of knowledge is facilitated by processes of circulation both within and across different social worlds. Some relational theories, such as ANT and ST, emphasized the role of nonhuman actors, such as computer software or programs, in transforming and mediating social relationships. These theories tended to view technology and society as coconstructed or coproduced, with no single dimension dictating change by itself. Within these theories, coproduction and implementation were often described as continuous processes, in which eHealth interventions were adapted to better accommodate different end-user settings and needs.

A total of 16% (6/36) of theories were classified as structural. The most common structural theory was institutional theory (IT) [84] (3 instances). Resource dependence theory, theory of middle managers’ role, contingency theory, task technology fit theory, and technology organization environment theory occurred only once each in the literature. These theories conceptualized structure as including institutional or organizational systems as well as political, cultural, and other macrosocial environments. They often assumed that people are constrained or influenced by external forces frequently beyond their comprehension or control. For example, IT posits that organizational structures and cultural norms drive eHealth implementation, despite strong political influence.

A total of 8% (3/36) of theories were classified as meaning centered: expectation confirmation theory, personal construct theory, and cultural dimension theory. Each of these theories occurred only once in the literature. These theories tended to focus on the cognitive dimensions (expectations, perceptions, and beliefs) that explain people’s willingness to accept the use of new health technologies. Although some meaning-centered theories, such as cultural dimension theory, have considered the influence of cultural values on the adoption and use of eHealth, these theories nonetheless centered their analysis at the individual level and were often used in combination with agency-level theories.

The Consolidated Framework for Implementation Research (CFIR) [85] was the only theory to be classified as a combined theory type. This theory is a meta-theoretical framework that provides a comprehensive listing of individual, social, and organizational constructs thought to influence eHealth implementation. However, it does not consider how these factors might be interrelated or how changes occur.

Discussion

Principal Findings

Evidence from a range of disciplines suggests that theory-informed approaches to implementation science are integral to the translation and implementation of eHealth into clinical care [1,10-18]. Analysis of the 119 studies included in this review identified 36 distinct theories that inform or explain eHealth implementation. However, only a few selected theories (UTAUT and TAM) were dominant, which is consistent with the findings from previous reviews [21-24]. Although these theories have been empirically proven to explain or predict certain aspects of implementation, Willmott et al [24] and Davis et al [21] caution that overreliance on common or favorite theories without direct questioning of their underlying assumptions limits progress in the field.

The typology by Sovacool and Hess [28] facilitated a closer examination of the assumptions underlying eHealth implementation theories. The findings revealed that the majority of theories were agency centered, emphasizing individual factors rather than the broader social and environmental factors impacting implementation. Although these findings were consistent with previous reviews [21,22,24,86], the wider net cast for this review provided the needed validation that this trend can be observed across multiple specialty areas and disciplines [27]. This calls into question whether theories currently being used to inform and explain the eHealth implementation adequately address the multiple and complex factors that influence the implementation process, and highlights the need for more dynamic, multilevel models of eHealth implementation [21,23,87].

This review identified a number of theories classified as relational or structural, which, to varying degrees, capture the complexity and multilevel nature of eHealth implementation. The most commonly cited relational theories were NPT, ST, and ANT. These theories recognize the important role of actors, relationships, and networks in mobilizing knowledge and embedding interventions into everyday practice. For ANT, networks are made up of both human and nonhuman actors, and technologies are understood to have agency and the potential to transform human interactions [88,89]. From this perspective, it may be a particularly useful theory for examining the implementation of eHealth technologies and the impacts these technologies have on human behavior. A criticism of ANT is that it has a flat ontology and refuses to consider institutional sources of power and inequality. Here, NPT and ST offer a possible extension, as both theories recognize the inseparable intersection between individual agents and wider social and organizational structures and norms. Structural theories also consider the influence of external forces on individual behavior and decision making. For example, IT, the most commonly used structural theory in this review, posits that an organization’s environment is capable of strongly influencing the development, acceptance, and use of eHealth interventions. This theory is considered particularly relevant for application in eHealth environments, which are highly institutionalized and subject to multiple regulatory forces, high levels of professionalism, and growing network externalities that can influence adoption decisions [50].

Of particular interest was the lacuna of normative theories identified in this review. Normative theories attempt to answer whether a technology is a net positive or negative for society and individuals [28]. To do so, they often rely on evaluative criteria determined by ethics, moral studies, political ecology, or social justice. Social justice theory and sustainable development are 2 common examples of normative theories. The absence of normative theories in eHealth implementation studies is emblematic of the broader tendency of implementation science to overlook the importance of contextual factors, such as economic, social, historical, and political forces, that perpetuate inequalities in the delivery of health care services [90]. This omission is concerning in the context of eHealth, as digital technologies have been found to exacerbate inequalities associated with older age, lower level of educational attainment, and lower socioeconomic status [91]. Future research should not shy away from normative questions of equity, justice, and sustainability and should find ways to incorporate theoretical approaches that enable exactly that.

When incorporating or combining theories, Sovacool and Hess [28] highlight the need for careful consideration of the epistemological baggage of different approaches. Combining multiple theoretical approaches may offer a more complete understanding or explanation, yet such combinations may mask contrasting assumptions regarding key issues [92]. For instance, are people driven primarily by their individual attitudes and motivation or do pervasive organizational cultures and social systems impose norms and values that shape people’s behavior, making individual characteristics relatively unimportant? These challenges may account for the tendency of theories to target variables at the same level. One exception was the CFIR framework, which was the sole theory that provided a menu of constructs at different levels for researchers to choose from. However, although CFIR recognizes the multilevel nature of eHealth implementation, it does not consider the relationship between constructs or how change takes place, leading Nilsen [92] to contend that it should not be considered a theory at all. Further research is needed to explore how diverse theoretical perspectives can be brought together in ways that capture the dynamic interaction between constructs [1], while avoiding disconnects and incompatibilities [28].

Limitations

This study has several limitations. First, papers not published in English were excluded, which may indicate a selection bias. The decision to keep the research question and inclusion criteria for this review broad resulted in a high yield of papers and, to some extent, reduced the specificity of search results. This decision was made to ensure the identification of the full spectrum of theories being used to inform and explain eHealth implementation. Restriction of inclusion criteria in previous systematic reviews [24] led to the omission of a number of key theories that provide a more comprehensive explanation of the various constituents of the implementation processes. A further limitation is that the protocol for this systematic review was not registered. However, every care was taken to ensure compliance with the core principles of the systematic review methodology. As Mallett [30] noted, systematic reviews do not constitute a homogenous approach, and researchers may adopt a more flexible approach that better suits their research purpose while continuing to comply with the principles for conducting a systematic review. Finally, the literature search for this review was conducted in June 2019. Given the rapid rate of publication in the field of eHealth, it is likely that recent relevant articles have not been included. As completing an updated search was not feasible for the research team, we suggest that future studies must continue to identify theories used to inform and explain the implementation of eHealth interventions.

Conclusions

This systematic review identified 36 theories that are being used to inform and explain eHealth implementation and classified these theories using the categories adapted from the typology by Sovacool and Hess [28] for theories of sociotechnical change. The results highlight the dominance of theories that focus mainly on individual readiness to accept health technologies rather than the various disorderly social processes or systemic dimensions of implementation. This calls into question whether theories currently being used to inform and explain eHealth implementation adequately address the multiple and multilevel factors that influence the implementation process. Nonetheless, this review identified a number of theories classified as relational, structural, or combined, which, to varying degrees, capture the complex interactions within a wider organization and policy system. Although less prominent in the literature, these theories may be particularly applicable to the implementation of eHealth in health settings and services.