The unprecedented growth of telemedicine services during the COVID-19 pandemic is acknowledged to have played a vital role in saving the global health system from collapse [1,2]. The impetus created by this crisis provided a compelling evidence base for the value added that telemedicine services contribute in terms of strengthening health service delivery. There is now considerable evidence of the promise that high-quality telemedicine services hold for supporting the multitude of challenges faced by health services around the world, contributing to improved health outcomes, quality, and health care provider and patient satisfaction [3-5], and to reducing costs [6], improving health service access for rural and remote communities [7,8], and protecting the environment [9]. However, the challenges faced by countries around the world to consolidate, sustain, and expand national telemedicine services [3,10,11] point to the multilevel and multifaceted complexity of this innovation.

Telemedicine is recognized as a fundamental pillar in the digital transformation of health systems, with a strong global push for scale-up [12-15]. However, defining “telemedicine” is becoming increasingly abstruse, with no universally agreed definition [16,17], and many related terms, such as “telehealth,” are used interchangeably by some and distinctly by others [18]. Moreover, the boundaries between telemedicine and other digital health interventions are becoming increasingly blurred, with telemedicine progressively being implemented alongside other “digital health interventions,” such as wearable devices and artificial intelligence. This study adopts the World Health Organization (WHO) definition of telemedicine, subsumed under “digital health,” encompassing remote clinical synchronous or asynchronous communication between those separated by distance, either client-to–service provider or service provider–to–service provider [19]. This definition includes the internet-based transfer of medical data, such as images, patient records, and clinical information, from remote monitoring devices. In this study, telemedicine is examined as an integrated part of the health ecosystem rather than as a stand-alone, siloed intervention.

There is now considerable evidence of the barriers to, facilitators of, and lessons learnt from the implementation of telemedicine across the world [3,20,21]. There is also an increasing number of knowledge tools (KTs) [19,22-28], defined as resources that help translate evidence into practice [29], often found in the form of conceptual frameworks, models, and theories [30]. However, there is considerable heterogeneity across these KTs, as well as a lack of consensus about the building blocks, taxonomy, and definitional boundaries of the constructs of telemedicine implementation along the life cycle [31,32]. This challenges telemedicine stakeholders, who need user-friendly knowledge mobilization resources to support them in their urgent task of designing, developing, and optimizing telemedicine services [32-34].

This implementation research, rapid umbrella review, and translational study contributes to an understanding of the requirements for the development of telemedicine systems. In this study, “implementation research” as opposed to “implementation science” is understood to focus on approaches for promoting the integration of evidence into the design and development of telemedicine interventions [35]. The study aimed to identify the evidence base for the constructs of telemedicine service implementation at scale by conducting an overview of systematic reviews of telemedicine implementation theories, models, and frameworks to improve telemedicine service development and outcomes, answering the research question: What is the state of the art of the implementation constructs of telemedicine architecture? Furthermore, this study aimed to assess the evidence base supporting telemedicine implementation KTs; identify shared conceptual constructs and outliers; and formulate recommendations to guide telemedicine services’ design, development, and optimization.

A rapid umbrella review was undertaken to consolidate knowledge on the building blocks or constructs of telemedicine implementation. Its adoption was prompted by urgent requests for technical guidance received by the WHO Collaborating Centre for eHealth at the Universitat Oberta de Catalunya, in response to needs expressed by national telemedicine leaders in lower-resourced countries. These focused on strategic advice regarding the core constructs of a telemedicine ecosystem to support national scale-up efforts. In addressing these demands, a lack of expert consensus and significant knowledge gaps were identified [3,10,11,31,32]. The combination of urgency, limited resources, and the need for timely, actionable outputs justified the selection of a rapid umbrella review as the most appropriate methodological approach [5,12-15]. On the other hand, the umbrella review approach was well suited to the substantial body of secondary evidence from systematic reviews focused on telemedicine implementation constructs. Umbrella reviews have grown in popularity in public health [36,37], and by modifying, streamlining, and accelerating the research process, the “rapid” umbrella review approach provides a robust, resource-efficient advantage [38,39], notwithstanding some limitations [40]. This accelerated research approach has responded to ongoing national telemedicine implementation efforts by promptly integrating state-of-the-art evidence—understood as the synthesis of current knowledge [41,42]—while also contributing to future telemedicine knowledge translation.

Our study was primarily guided by the WHO Rapid Review to Strengthen Health Policy and Systems [38] and was further informed by additional methodological guidance relevant to rapid and umbrella reviews, complex interventions, qualitative and mixed methods analysis, and implementation science [36,37,40,43,44]. The study followed the reporting requirements of the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) guidelines (2020) [45] along with the extension for complex interventions [46]. The study was registered in PROSPERO on February 13, 2024 (2024 CRD42024512516; Multimedia Appendix 1).

Search term keywords were defined, informed by existing evidence synthesis, Medical Subject Headings (MeSH), and the technical support of a specialist librarian. On May 20, 2024, a multifield search was adapted to the indexing language of 3 databases (PubMed, Web of Science, and Scopus; Multimedia Appendix 2). The database search was supplemented by recommendations from telemedicine experts and a search of systematic review repositories (including Joanna Briggs Institute Evidence Synthesis, Cochrane Database of Systematic Reviews, PROSPERO, Database of Abstracts of Reviews of Effects, Epistemonikos, and Campbell Collaboration). The primary researcher (CK) performed the search, and all the articles were imported into the Covidence (Veritas Health Innovation) software and deduplicated.

Screening was undertaken in a 2-stage process from May 20, 2024, to June 1, 2024, based on predefined inclusion criteria (Textbox 1). After pilot-testing the criteria for mutual understanding, 2 researchers (CK and JMR) independently screened titles and abstracts against the criteria, and conflicts were resolved by the third researcher (FSR). Shortlisted reviews were then assessed based on their full text by the same review team at the second stage of selection. During the selection process, several consensus meetings were held to ensure rigor, quality, and consistency in that process.

To integrate the full spectrum of evidence relevant to the research objectives, the search terms struck a balance between specificity and broader inclusivity, as recommended by the umbrella review methodological guidance [36]. The “study design” criteria included research methodologies that demonstrated systematic and robust secondary evidence synthesis, specifically “systematic reviews” and “scoping reviews” (referred to as SRs), with all other study designs excluded. The search on “telemedicine” included related terms, such as “telecare” and “telehealth,” as well as broader terms like “remote consultation,” “eHealth,” and “digital health.” These broader terms were further screened and only considered for inclusion when primary research on telemedicine was present. Studies that explored telemedicine implementation constructs at scale, reaching multiple populations (irrespective of setting, specialty, group of patients, condition, or disease), were included, while studies reaching a single population were excluded. Given the increasing integration of telemedicine into the wider digital health ecosystem, digital health studies that explored implementation constructs and incorporated telemedicine were included, whereas those with no mention of telemedicine were excluded. Studies before and during implementation were included, but those focusing on postimplementation outcomes were excluded as these were being synthesized in a concurrent meta-analysis undertaken by the same organization. Finally, to meet the state-of-the-art research objectives, the search was limited to studies published between January 1, 2018, and May 20, 2024. No additional filters were applied. Three main factors informed the selection of this timeframe. First, this period saw an unprecedented surge in evidence generation and global policy directives promoting the development and scale up of national telemedicine initiatives [5,19,47-49]. Second, there was a marked acceleration in technological advancements, including the widespread availability of high-speed internet, the ubiquity of smartphones, the expansion of remote monitoring devices, and the emergence of artificial intelligence. These innovations have significantly enhanced the scalability of telemedicine while also introducing new implementation challenges and opportunities [13]. Third, the volume of telemedicine research increased substantially during this period, reflecting growing scientific interest and activity in the field [50-53].

Critical appraisal, although not mandatory for rapid reviews [38], was undertaken as a good practice measure. On completion of the second screening, the final sourced SRs were appraised in duplicate by the same review team as the screening, using the Joanna Briggs Institute Critical Appraisal Checklist for Systematic Reviews and Research Synthesis [36]. These were scored on a 4-point scale of “very good,” “good,” “fair,” and “poor.” Those scoring “poor” were rejected due to inadequate methodological quality.

Data extraction was performed by the principal researcher (CK), with key variables extracted in duplicate by JMR, using a predefined template in Excel (Microsoft Corporation). A third researcher (FSR) verified the data. Multimedia Appendix 3 provides the extraction table template, which includes the following: citation, geographic characteristics, methodology, objectives, search strategy, review findings, critical appraisal score, and a summary of the characteristics of the primary research of each SR (including the number of studies, years of publication, methodologies, country, settings, type of telemedicine interventions, clinical specialties, and theoretical underpinnings). Where possible, data were extracted verbatim to remain faithful to the authors’ original intent.

The analysis was primarily qualitative, combining a meta-aggregative approach with narrative, thematic, and framework synthesis [36,37] and summarizing and explaining the findings in tables and text. Analysis followed the principles of reflexivity, acknowledging the inherent subjectivity of the researcher, through a 6-step iterative process of familiarization, coding, thematic categorization, theme review, further definition, and reporting [54,55]. This included exploring similarities, variations, and exceptions across reviews. The review findings were coded in ATLAS.ti (version 25; Scientific Software Development) by the principal reviewer (CK) using thematic inductive analysis. The research team had regular discussions over the course of the analysis to scrutinize assumptions in the interpretation of the data. Consideration was also given to the impact of primary study duplication across reviews. Due to the broad nature of the research question and the heterogeneity of the reviews, a meta-analysis was not possible, nor relevant, although characteristics of the studies were summarized in descriptive statistics and frequency tables. The findings of the SR authors were synthesized, without attempting to reinterpret findings, drawing out generalizable conclusions to guide practice and policy [36].

In the analysis, attention was paid to the methodological considerations of conducting evidence synthesis of complex adaptive health systems [37] comprising multilevel components and “fuzzy” boundaries [56], with a diverse range of stakeholders that interact in nonlinear dynamic ways. Such systems are known to be unpredictable in the way they self-organize and respond to change. Indeed, understanding “context” and how to integrate it into the analysis was particularly important [57].

Our study fully adhered to the allowable methodological adaptations outlined in rapid review guidance, ensuring both methodological rigor and the timely delivery of findings [38,43]. As required, all modifications have been transparently reported in this section. First, the search was scoped, applying a year of publication filter (2018 to 2024) to ensure state-of-the-art evidence in recognition of the exponential growth of telemedicine research since the COVID-19 pandemic [51], and the search strategy was limited to 3 databases, with a supplementary search of review repositories. Second, coding was undertaken by one researcher, acknowledging the inherent interpretive and subjective nature of thematic analysis [55], and the research team enriched the analysis through iterative collaborative reflection. Third, the research protocol objectives were further strengthened to more explicitly consider “knowledge translation” tools and “context.” Alternations were documented in a “decision log.” All deviations from standard umbrella review procedures were consistent with accepted rapid review methodologies and enabled the generation of timely, high-quality findings to address an urgent knowledge gap. Nonetheless, as is inherent in the rapid review approach, the interpretation of findings should be approached with appropriate caution.

From the database search, we identified 1224 articles, 565 (46.16%) from Scopus, 373 (30.47%) from PubMed, 278 (22.71%) from Web of Science, and 8 from other supplementary searches. Of these, 372 (30.39%) duplicates were removed, leaving 852 (69.61%) articles. After being screened by title and abstract, 821 (96.4%) of those studies were excluded. The remaining 31 (3.6%) articles were screened for full text, and a further 13 (42%) studies were excluded: 8 (61%) for reasons of wrong intervention, 3 (24%) for wrong study design, and 2 (15%) for wrong outcomes (Figure 1). Finally, 18 (58%) SRs were selected and critically appraised, all meeting the minimum quality standard for inclusion. On the 4-point scale, 15 (83%) of the included reviews were assessed as good and 3 (17%) as fair, with none of the reviews meeting the highest or lowest quality standard (Multimedia Appendix 4 [58-75]).

The SRs explored telemedicine implementation dimensions through different conceptual lenses. A case study approach was taken by half of the studies (9/18, 50%), either exploring implementation from a regional [60,63] or a country perspective [68,74,75], or an intervention, such as remote patient monitoring [71], telerehabilitation [73], virtual care [61], and home-based care [62], with one focusing on the COVID-19 pandemic [59]. In total, 3 (17%) of the reviews examined telemedicine implementation from the perspective of implementation constructs, such as trust [58], change management [66], and telemedicine infrastructure [63]. The remaining studies explored implementation theories, drawing on various models relating to design [67], implementation readiness and assessment [69,72], policy [70], spread and scale [65], and other theoretical models [64].

Each SR synthesized the findings of its study of the constructs of telemedicine implementation into a “KT” (Multimedia Appendix 7 [58-75] provides a summary of the constructs of each KT). KTs can be broadly grouped into 4 types. First, diagrammatic schemas of implementation provide conceptual models of telemedicine implementation [64-67,70,71]. Second, some (2/18, 11%) SRs consolidated and mapped the dimensions of existing theories [69,72]. Third, several KTs took the form of an implementation matrix delineating enablers and impediments [58,59,62,63,68,74,75]. Finally, 2 (11%) reviews provided a set of thematic strategic recommendations to guide the implementation process [60,61]. Many of the SRs emphasized that their respective tools had been designed to support implementation in practice [60,63-67,70-72,74], although none indicated that the tool had been field tested.

Comparing the telemedicine implementation constructs of the KTs to each other (Tables 2 and 3), there was a high degree of heterogeneity in the conceptualization of implementation constructs across KTs. The taxonomies of KTs varied across SRs, with most being poorly defined and lacking in definitional boundaries. Even those that, on the surface, appeared to have a degree of similarity [67,70] were, on closer analysis, conceptualized, defined, and delineated differently. However, 2 broad dimensions of implementation were apparent across the KTs: those relating to the “process or lifecycle” of telemedicine implementation, and those that were more “thematic” elements of implementation. Overall, there was a predominance of the latter approach, with 14 (78%) of the reviews exploring telemedicine implementation in a “snapshot” or fixed point [58-64,68-70,72-75], while the remaining 4 (22%) took a process or life cycle approach, exploring how implementation had evolved longitudinally as the telemedicine system matured [65-67,70].

Considering the telemedicine implementation life cycle, most attention across the KTs was focused on the actual point of implementation of telemedicine services, rather than before the implementation or optimization. In relation to the thematic dimensions of implementation, the human and social, organizational, technology and infrastructure, and legal policy and financial aspects of telemedicine services received the most attention. Interestingly, across the KTs there were limited attempts to quantitatively rank the relative importance of implementation constructs beyond anecdotal efforts, except for one review which found, using latent semantic analysis, that the highest correlation was in the “human” dimension, across the constructs of “legal,” “financial,” and “privacy” [70].

By conducting evidence synthesis of implementation theories, models, and frameworks, this umbrella review provides a global perspective of the state-of-the-art of telemedicine implementation constructs. The findings represent 63 countries comprising low- to high-resource contexts and a range of settings from clinical care to self-care. Areas of convergence and discrepancies in the conceptualization of implementation constructs are identified, thereby contributing to the urgent need for consensus building on the dimensions of telemedicine implementation to support knowledge transfer and guide evidence-based high-quality telemedicine implementation. The findings of our evidence synthesis highlight the state-of-the-art essential requirements for a national telemedicine implementation ecosystem, structured into 2 overarching categories: process and thematic dimensions. The process dimensions include readiness and needs assessment, road map development and planning, change management, implementation of telemedicine services, and continuous improvement and performance measurement. The thematic dimensions cover human and sociocultural factors; organization, operations, management, and leadership; communication and coordination; policy, legal, and financial considerations; clinical conditions and quality of care; and the broader contextual environment.

Comparing our findings to those obtained from similar studies, none of the latter studies were found to be directly comparable in undertaking an umbrella review of telemedicine implementation constructs. The closest study is an umbrella review conducted by Mair et al [77] and later updated by Ross et al [78] focusing on digital health rather than on the unique characteristics of telemedicine architecture. However, similar to our study, those studies identified the multiple levels and factors to be considered in digital health implementation, including technological considerations, organizational issues, health professionals, and the processes of implementation. Some other relevant studies were either limited in scope, focusing on a single stage or aspect of telemedicine implementation, such as needs assessment, readiness and evaluation [79-83], and end-user experience [84], or were outdated given this rapidly evolving field [33,85,86]. Others [82,87] identified that implementation KTs varied in their degree of robust development and were often characterized by limitations such as a lack of evidence base and expert consultation; weak theoretical foundations; and, similar to our study, found limited field testing of KTs, an absence of practical guidance and evaluation, and excessive complexity regarding their application in practice.

A KT that incorporates most of the telemedicine implementation constructs identified in this study and attempts to address the limitations identified in other telemedicine implementation tools is the support tool to strengthen telemedicine: guidance for telemedicine assessment and strategy development, developed by the WHO [88]. This KT comprises a succinct evidence-based tool to support knowledge translation and guide practice. A unique strength of this tool is its conceptual architecture, which identifies telemedicine implementation cores (process indicators), domains (thematic dimensions of implementation), and items (implementation activities). The tool is informed by evidence synthesis [3] and a Delphi consensus study, the aim of which was to align expert opinion on telemedicine implementation constructs [89], and is currently being field-tested in various countries. This knowledge translation tool is not context, disease, user group, or setting specific; rather, it provides a generic resource that describes the building blocks required for the development of a successful telemedicine service along the life cycle of telemedicine maturity, which is adaptable to local needs and available resources.

A critical finding from our study is the importance of telemedicine implementation KTs that are flexible, enabling them to be adapted to different contexts and to support iterative, continuous learning. A recurring theme is the need for KTs to be adaptable to the realities of different cultural, regulatory, funding, capacity, and resourcing levels, particularly in the diverse contexts of lower- and middle-income countries. To support this, our study identified the substantial alignment of experts on the essential constructs of telemedicine implementation despite the considerable heterogeneity in the conceptualization of KTs. Finally, although boundaries of telemedicine implementation constructs are fuzzy and relationships across them are dynamic, this study emphasizes the operational imperative of delineating and defining constructs to help support the allocation of roles and responsibilities, resourcing, and coordination and collaboration for telemedicine implementation actors.

This study has both strengths and limitations. A strength of the study is its timely contribution to a complex, urgent gap in knowledge translation, which is hindering the global implementation and scale-up of sustainable, high-quality telemedicine systems. By synthesizing the state-of-the-art evidence of KTs, this study contributes to informing the development of meaningful KTs that support the pressing needs of practitioners who are designing, developing, and optimizing telemedicine services. The limitations of the study were mentioned earlier and mostly consist of modifications associated with accelerating the study methodology in accordance with “rapid review” guidance, implying a degree of caution in interpreting the results while enabling timeliness in sharing the findings.

This study contributes to an urgent real-world public health problem by synthesizing state-of-the-art evidence of telemedicine implementation to inform knowledge translation, thereby furthering the implementation of high-quality, sustainable, and scalable telemedicine systems. An important finding from this study is the need to continue building consensus on what the essential evidence-based telemedicine implementation constructs are, while providing flexibility to enable implementation to be adapted to the needs and realities of different contexts. Future implementation research needs to focus on field-testing evidence-based KTs for usability and adaptability to different telemedicine implementation settings.