Patient centricity—a health care approach that prioritizes patients’ needs, preferences, and experiences, ensuring they are active participants in their own care and decision-making—has emerged as a core principle in contemporary health care practices [1-4], closely linked to the ideas of sustainability and responsiveness [4]. Empirical evidence supports the importance of an integrated supply chain in enhancing patient centricity [5]. Studies have shown a direct and positive relationship between integrated supply chain performance and patient centricity, as well as between knowledge exchange and integrated supply chain performance. In this context, information technologies play a crucial role [1,4,6], while achieving interoperability remains a challenge that must be addressed to improve supply chain integration.

Three main concepts are linked to patient centricity and advances in digital innovation: (1) service inclusivity, (2) shared responsibility or management, and (3) remote patient monitoring [4]. Moreover, these concepts are reflected in 3 prominent roles: facilitator, connector, and enabler [4]. The first provides patients with access to services and data, the second enables connection and information sharing across health care infrastructure, and the last pertains to remote health care [4]. In this context, interoperable health care systems—when viewed through the lens of the supply chain or value chain—play a critical role in advancing patient centricity. This perspective emphasizes the need for seamless integration and data exchange among stakeholders across the health care ecosystem, thereby supporting holistic patient care and promoting a more integrated approach to health delivery. The alignment of these systems with patient-centric principles highlights the importance of collaboration, efficiency, and improved health outcomes in the evolving landscape of health care delivery.

The relevance of interoperability for a sustainable, patient-centric health care value chain requires a holistic understanding of the domain—one that also addresses the ongoing debate surrounding the classification of interoperability. Some authors propose simple classifications comprising 2 elements—syntactic and semantic interoperability [7]—or 3 elements—technical, syntactic, and semantic [8]. Other studies suggest as many as 8 [9] or even 12 classes [10]. References can be found to various types of interoperability, including technical [9,11], syntactic or syntactical [9,11], pragmatic [9,11], dynamic [9], conceptual [9,11], structural [9], functional [9], semantic [9,11], platform [11], process [12], organizational [13], people [14], knowledge [10], services [10], social networks [10], electronic identity [10], ecosystem [10], and legal interoperability [15,16].

Previous reviews on interoperability have not focused on specific industries such as health care, nor have they consistently provided structured classifications. One review addressed user model interoperability but lacked a systematic approach [17]. Another study examined evaluation models of interoperability, categorizing 12 models across 4 granularity levels [10]. While this study mentioned interoperability attributes, these were not developed through a systematic procedure [10]. Additionally, an investigation explored interoperability in the context of the Internet of Things, presenting a comprehensive taxonomy [11]. Nevertheless, the proposed taxonomy is based on a predefined model rather than a specific methodology for taxonomy development [11]. Furthermore, it is not a facet-based taxonomy.

There are also reviews focused on interoperability in health care. Nevertheless, most of them lack a comprehensive scope within the domain. They are often limited to a specific health care service, such as emergency care [18]; a particular technology, such as blockchain [19]; a type of health information technology (HIT), such as electronic health records (EHRs) [20,21]; or a specific type of interoperability, such as semantic interoperability [16,21-24]. One of these surveys proposed a taxonomy for the semantic interoperability of health record systems; however, its development did not follow a specific methodology [21], and the taxonomy was not facet based. Similarly, we found reviews focused on specific and relevant health interoperability standards or technologies [25-29], yet none of them included a taxonomy.

Three studies addressed the topic of interoperability in health care using a systematic and broader approach compared with previous studies [9,30]. The first was a literature review that assessed 61 records [30], focusing on identifying functional requirements for data integration. In line with its scope, this study did not include a quantitative field assessment. The second study surveyed 36 articles to present the interoperability requirements for health information systems [9]. This survey presented word clouds and independently classified and summarized interoperability standards and architecture components, all grounded in the literature review [9]. However, it did not offer a comprehensive organization of the knowledge domain, as the classifications were not integrated. Notably, the study did not propose a taxonomy, nor did it follow a specific methodology for taxonomy development or identify the facets of the topic. The third study conducted a bibliometric analysis of system interoperability and data linkage in health care, including both cluster and network analyses of the topic [31]. The authors selected a sample of 63 journal articles based on their average citations per year, among other criteria [31]. While this criterion is relevant, it may exclude more recent articles that have not yet accumulated sufficient citations, as well as gray literature—only 1 conference paper was included in the sample [31]. In line with its approach, the study did not propose a taxonomy grounded in a literature review. Furthermore, no studies were found that examined interoperability in health care from a supply chain or value chain perspective.

Following a review of the relevant studies, we identified a clear scarcity of systematic reviews that address interoperability in health care as a comprehensive knowledge domain. There is a need for a thematic structure and a higher-order, facet-based taxonomy of knowledge, particularly from the perspective of the supply or value chain. This study aims to fill that gap.

A systematic literature review has been defined as a type of research synthesis [32] that employs predefined methods aimed at reducing bias, thereby producing more reliable results [32]. These reviews incorporate various methods of data analysis [33], including quantitative approaches such as bibliometrics [34], as well as qualitative, in-depth document analysis. Additionally, the literature suggests systematic methods for taxonomy development [35,36]. Consequently, a systematic approach is the most appropriate for our inquiry.

A holistic perspective—currently lacking in the literature and pursued in this study—helps eliminate knowledge silos and can be achieved through the integration of both quantitative and qualitative methodologies. The use of machine learning tools enables the handling of large volumes of information, offering critical insights into the knowledge structure of this comprehensive domain. Additionally, a thorough review of a selected sample [31,37] supported the development of a facet-based taxonomy that captures the higher-order facets of the topic. Previous studies in other subject areas have used systematic approaches, combining quantitative bibliometric assessments with taxonomies derived from in-depth literature reviews [38].

This research pursues 2 fundamental objectives. First, it seeks to vividly map the intricate knowledge structure of the expansive field of health care interoperability, viewed through the compelling lens of the value chain. This exploration highlights not only the critical topics and emerging hot spots that shape the discourse, but also the complex interconnections that bind key themes together. Second, it aims to develop a robust, higher-order framework for a taxonomy of interoperability in health care, offering a nuanced perspective that underscores its significance within the evolving landscape of the health care value chain.

This review aims to be both comprehensive and grounded in high-quality data. To achieve this, the initial search was designed to capture a broad spectrum of information, including both peer-reviewed articles and gray literature. However, to ensure consistency and minimize subjective judgment regarding data quality and selection bias, we chose to rely on specialized databases that index only peer-reviewed documents.

Data were retrieved from 6 databases on April 1, 2024, in accordance with PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) and PRISMA-S (Preferred Reporting Items for Systematic reviews and Meta-Analyses—Search extension) guidelines (see Multimedia Appendix 1) [39,40]. The databases consulted were (1) Scopus, (2) Web of Science (WoS), (3) IEEE Xplore, (4) Embase, (5) Cochrane, and (6) PubMed. In Scopus, the search was limited to the Title-Abstract-Keywords fields; in WoS, it was restricted to the Topic field; and in Cochrane, it was confined to the Title, Abstract, and Keyword fields. These limitations were applied to avoid the inclusion of irrelevant records, as broader fields—such as References in Scopus—could yield inaccurate results. For IEEE Xplore, Embase, and PubMed, no field restrictions were applied, ensuring the most comprehensive coverage of relevant literature.

The search terms focused on key themes such as interoperability, value chains, and supply chains within the context of health care. Notably, multiword terms were searched as exact phrases to prevent inaccuracies that might result from search systems interpreting the words separately (Textbox 1).

The eligibility criteria for document types (articles or conference papers) were aligned with the structure of each database, aiming to maintain consistency across sources as much as possible. The criteria were as follows: (1) conference paper and article for Scopus; (2) article and proceeding paper for WoS; (3) conferences and journals for IEEE Xplore; (4) article and conference paper for Embase; and (5) classical article, clinical conference, consensus development conference, consensus development conference (NIH), and corrected and republished articles for PubMed. Additionally, the review was limited to English-language materials to reduce potential bias in the quantitative analysis. These restrictions were applied using the databases’ built-in filters where available.

The period for the review was set from 2011 to 2024, considering that one of the most significant standards in health care—Fast Healthcare Interoperability Resources (FHIR)—was introduced in 2011 [27]. The eligibility criterion regarding topical relevance—whether the document focused on the subject under study—required subjective evaluation by the researchers. In line with recommendations from the literature, which emphasize the importance of a peer-review stage to reduce selection bias [39], this evaluation was conducted in 2 stages and involved both researchers. Following the approach adopted in previous studies [41], one of the researchers (CPD) conducted a thorough review of titles and abstracts to perform the initial selection. This selection was then peer reviewed by the second researcher (CAM). Any doubts or discrepancies were discussed and resolved collaboratively. As this process relied on the review of titles and abstracts, it was decided in advance to exclude any articles lacking abstracts. Microsoft Excel was used to manage and document the screening process. The eligibility criteria are summarized in Textbox 2.

The data retrieved from the various databases were first merged, cleaned, and prepared for analysis. The data cleaning process involved removing duplicates [42], correcting formatting issues [43], and addressing other errors such as spelling inconsistencies [44]. The literature outlines a variety of approaches for this stage. Some studies have conducted manual data reviews [45], acknowledging the effort required [46]. Others have utilized tools such as Microsoft Excel for data cleaning [45,47], while additional research has developed or used specific software tools—including custom packages [46,48]—or dedicated procedures [49]. Notably, some authors have reported that even after using software tools, additional manual data cleaning was necessary and subsequently performed [48,50]. Based on this, we opted to carry out a manual process using Microsoft Excel for data merging and cleaning. Furthermore, the literature presents differing views on the criteria for identifying duplicate records. Some studies rely solely on digital object identifiers (DOIs) [46], while others combine DOIs with additional attributes [46], or use alternative attributes entirely [51]. Considering these options, we chose to identify duplicates based on DOIs, as this is regarded as the most reliable method.

The procedure we followed began with retrieving and exporting bibliometric data from multiple databases. The data were then merged according to the structure of the Scopus database. Subsequently, we conducted data cleaning based on the 2-step approach recommended in the literature [44]. The first step involved detecting and removing duplicates, while the second focused on identifying and correcting errors and inconsistencies in preparation for bibliometric analysis [44]. Initially, records without a DOI were excluded. Duplicate records were identified based on duplicated DOIs, with only 1 instance of each retained. In the second step, error correction was carried out. This phase can be conducted using either Microsoft Excel or a thesaurus-based approach [44]; we applied both methods to ensure accuracy and consistency. First, we used Microsoft Excel to correct special characters in the columns identified as relevant for our study, as flagged by Biblioshiny (K-Synth Srl). Second, we applied a thesaurus-based approach during the keyword co-occurrence analysis to standardize terminology. Specifically, we treated the following terms as synonyms: eHealth and e-health; electronic health record, electronic health records, and EHR; IoT and Internet of Things; IoMT and Internet of Medical Things.

Missing data were also identified and addressed during the data preparation process. Certain statistical techniques commonly used in bibliometric studies—such as multiple correspondence analysis and network analysis—require complete data sets to avoid potential bias [52,53]. Therefore, it was essential to assess the extent of missing data [54]. According to the literature, missing data rates of 10% or less are generally considered low [54], and in such cases, any imputation strategy may be applied. One commonly used approach is complete-case analysis, which involves deleting records with missing data.

The data synthesis process consists of 2 stages: quantitative and qualitative. The literature supports the use of both quantitative and qualitative techniques for the development of classifications [36]. A key component of bibliometric analysis is the identification of the most relevant themes or topics within the domain under study [55,56]. Accordingly, the first stage involves bibliometric evaluation techniques, including the application of unsupervised machine learning methods. These are used to conduct network analysis, clustering, and multiple correspondence analysis to explore influential topics, emerging hot spots, and the interrelationships among key issues and thematic groups. This approach aims to reveal the knowledge structure of the field from a holistic perspective.

The primary quantitative technique used to reveal the knowledge structure of the domain was a co-occurrence analysis based on keyword terms. This method is particularly well-suited for identifying hot topics and thematic areas [57], and it provides insights into the most frequently discussed subjects within the field [38]. As previously mentioned, the study drew data from several databases, some of which include enriched keywords generated by proprietary editorial algorithms. To minimize potential bias introduced by varying editorial criteria or algorithms, we chose to exclude these enriched keywords and relied solely on those provided by the authors. The data assessment was conducted using Biblioshiny, the web interface of the Bibliometrix R package (version 4.1; R Foundation), which provided the tools necessary for performing the bibliometric analysis.

Cluster and factor analyses are techniques commonly used to classify elements within a data set. In accordance with the literature on classification methods [35], which defines the objective as “grouping objects of interest in a domain based on common characteristics,” clusters consisting of only a single element were excluded, as they do not meet this criterion. The cluster analysis was conducted using the Walktrap algorithm. We selected this algorithm because it is one of the most utilized in community detection [58] and “it is often effective at determining the correct number of communities and assigning items to their proper community” [58]. Other studies also acknowledge Walktrap as a well-performed algorithm [59]. In addition, we decided to perform the analysis based on other algorithms to avoid possible algorithmic bias. The Leading Eigenvalues algorithm, which also has a good performance in community detection [60], was selected.

After the cluster analysis, we performed the factor analysis. It is relevant to mention that the literature emphasizes that the distinction between clustering techniques and factor analysis lies in the way variance is partitioned [61]. In factor analysis, variance is distributed among factors, and the elements have loadings on the different factors within the solution [61]. Thus, we considered it pertinent to present and compare both solutions—those from the cluster and factor analyses.

The COVID-19 pandemic posed a significant challenge to the health care sector. Consequently, we examined the thematic evolution of this field, identifying the year 2020 as a critical turning point and using it as the final stage of the quantitative analysis.

The second stage was qualitative and involved a thorough analysis of the most relevant documents to develop a facet-based taxonomy. A key characteristic of such taxonomies is that they incorporate multiple perspectives that describe a topic [62] and, once completed, enable the use of compound terms—drawn from the same facet or across multiple facets—to define the object of interest [62]. As the meta-characteristic of this taxonomy [35,36], we established that aspects of interoperability would be integrated under the perspective of the health care value chain. We adopted an inductive [35], empirical-to-conceptual approach [36,61], based on an in-depth review of a selected sample of documents. This approach is most suitable when the objective is descriptive, as is the case in our study [36]. The process aimed to formulate a label or concept representing the type [61] and, in line with the literature, followed 3 stages [36] applied to the selected sample. First, each document was read in detail and summarized. Basic and recurrent topics—or codes—were identified and listed. Second, the common characteristics of these topics were identified, and the topics were grouped into categories based on their shared attributes. It was noted that some categories needed to be included within others at a higher level. This second step enabled the identification of the third- and second-level categories, or subfacets. Third, the second-level subfacets were reviewed and organized into higher-level dimensions or facets, representing the first-level facets. One of the researchers (CPD) carried out these steps and formulated an initial proposal. This proposal was then peer reviewed and evaluated by the other researcher (CAM). Any disagreements were discussed and resolved jointly by both researchers (CPD and CAM) until a final version was reached.

In addition, we followed the iterative process suggested by the literature [35,36] and evaluated the fulfillment of ending conditions [35]. The ending conditions we established were (1) having a representative and balanced sample composed of both journal and conference articles, with journal articles forming the majority to ensure the quality of the taxonomy; and (2) achieving category saturation—meaning no additional insights were found [63] regarding the categories under study—given the qualitative nature of this analysis.

Previous studies that conducted bibliometric evaluations selected a sample of studies for an in-depth review that follows and complements the quantitative analysis [37]. To reduce potential bias stemming from researcher subjectivity in the sample selection, this study based the sample selection primarily on the contribution of articles to the factorial solution obtained in the quantitative assessment.

The search yielded 1912 documents (Scopus: 610; WoS: 161; IEEE Xplore: 303; PubMed: 191; Embase: 643; and Cochrane: 4). Among these, 540 were excluded because they were not articles or conference papers or were not written in English. Additionally, 212 documents lacked a DOI, and 318 were duplicates; these were also removed. At the end of the identification stage, 842 documents remained for screening. It was found that 112 records were published before 2011, 7 did not include abstracts, and 319 were not focused on the topic. These documents were also excluded. Additionally, 34 out of 404 (8.4%) documents lacked authors’ keywords and were withdrawn, as we applied the complete data imputation strategy. Ultimately, 370 records remained. The list of these 370 studies is available on GitHub (see Multimedia Appendix 2). The stages of document identification, screening, and inclusion are illustrated in Figure 1 (also see Multimedia Appendix 1).

The data spans 370 documents published between 2011 and 2024, distributed across 223 sources. The most relevant source is Studies in Health Technology and Informatics, and 48 sources published 2 or more documents. Figure 2 presents the core sources based on Bradford’s Law. Bradford’s Law is a bibliometric distribution measure [64] that reflects journal productivity [64,65], based on the number of accumulated articles [65]. In this regard, it is worth noting that 17 sources comprise the nucleus or core sources. Nevertheless, Studies in Health Technology and Informatics is clearly the most productive one.

In addition, 358 out of the 370 (96.8%) documents were coauthored, with an average of 5.38 researchers per study. A total of 1008 authors’ keywords were identified. The trend topic analysis encompassed 19 concepts. Among them, ontology, telemedicine, and cloud computing had the oldest mean year of use, while blockchain, FHIR, and access control had the most recent (see Figure 3). This pattern reflects the growing relevance of new technologies, such as distributed ledger technology (blockchain), in this field. It is not surprising that access control also gained relevance in recent years, given that blockchain, due to its immutability and traceability, promises to enhance privacy and security in access management. Additionally, it is noteworthy that the FHIR standard gained prominence only after 2018, despite being issued in 2011 [27]. This may reflect the time required for a new standard to be widely disseminated. The average annual scientific production, excluding the year 2024, is 27.76.

We performed the cluster analysis using 2 algorithms to avoid algorithmic bias: (1) Walktrap—as the main one—and (2) Leading Eigenvalues—as the confirmatory one. Both solutions resulted in a 2-cluster configuration based on 23 nodes. Figure 4 depicts the solution generated by the Walktrap algorithm. The number of members and the specific elements within each cluster were almost identical in both algorithmic solutions. The only difference was the placement of the term EHR, which appeared in different clusters depending on the algorithm used.

The first cluster encompassed the terms eHealth, interoperability, EHR, ontology, semantic interoperability, telemedicine, FHIR, Health Level Seven (HL7), OpenEHR, standards, health information, system, semantics, and web services. The second cluster included the terms IoT, health care, blockchain, cloud computing, privacy, security, middleware, access control, and medical services. The network analysis revealed that 3 terms had the highest influence on node communication within the entire network. These are interoperability, EHR, and eHealth, with betweenness centrality scores of 70.971, 59.460, and 12.000, respectively. These terms also contribute to the speed of information dissemination within the network, showing the highest closeness centrality scores: 0.047, 0.043, and 0.034, respectively. All 3 terms are grouped within the first cluster. In the second cluster, the terms IoT, blockchain, and health care exhibited the highest levels of betweenness centrality—6.765, 2.581, and 1.283, respectively—as well as the highest closeness centrality scores: 0.034, 0.030, and 0.030, respectively.

In this section, we analyze the conceptual classification and structure of this domain using factor analytic techniques, specifically multiple correspondence analysis. This evaluation yielded a 2-factor solution, encompassing 2 dimensions that account for 59.46% of the total variance (inertia). The first dimension contributed 36.78% to the inertia, while the second contributed 22.68% (see Figure 5). This 2-factor solution was corroborated by the topic dendrogram (see Figure 6). Most themes were associated with the first factor. The second factor encompassed only 3 elements: sensors, monitoring, and medical services.

The thematic evolution of this field, with a critical turning point in the year 2020, is presented in Figure 7. In the initial phase, 6 themes characterized the domain: eHealth, digital imaging and communication medicine (DICOM), health care, semantic interoperability, EHRs, and ontology. By 2020, ontology persisted as a key term, while the other themes were replaced by interoperability, openEHR, and blockchain.

The literature review followed a systematic process [35,36] that encompassed 3 iterations. The initial iteration focused on documents with contributions of 1.0 or higher, resulting in a total of 28 articles, including both journal articles and gray literature. Notably, 15 out of 28 (54%) samples consisted of conference articles, indicating that the aim of achieving a balanced and representative sample—with journal articles as the majority—was not fulfilled. Therefore, additional iterations were deemed necessary to ensure category saturation. As a result, it was decided to incorporate 2 more iterations.

The second iteration included only journal articles, selected based on their contribution to the factorial solution, adding 35 journal articles to the sample. The third iteration comprised a final selection of 23 conference articles, primarily chosen for their contribution to the factorial solution and their availability. Unfortunately, we encountered access issues with 7 articles—1 from the first iteration, 2 from the second, and 4 from the third.

Ultimately, the review included a total sample of 79 out of the 370 (21.4) articles. Among these, 46 of the 79 (58%) documents were journal articles, while 33 of the 79 (42%) were conference papers. The list of reviewed articles can be found in Multimedia Appendix 3, and Figure 8 (see also [7,8,12-15,66-145]) illustrates the outlined systematic procedure.

Following our analysis, we confirmed that saturation was achieved. All facets were identified within the first and second iterations, with no new facets introduced exclusively in the third iteration (see Figure 9). Additionally, all other established conditions were met. Therefore, we concluded that the objective was fulfilled after completing the 3 iterations.

In developing the facets, we reviewed the themes identified in the network and cluster analyses. Three facets—Content, Object, and Source—represented these themes and were designated as core concerns. The content involved the technology primarily intervened (encompassing, for example, the themes EHR and IoT). Object referred to what is shared (reflecting themes such as semantic interoperability and semantics). Source pertained to the mechanisms used to achieve interoperability (involving themes such as blockchain and standards, among others). Consequently, these 3 facets were considered core concerns. However, an additional facet, named Ambit, emerged from the literature review and was regarded as an implicit concern. Figure 9 illustrates the 4 higher-order facets and their subfacets, extending to the third degree, and Textbox 3 summarizes the key findings.

The intricate landscape of knowledge in the field of comprehensive health care interoperability is clearly illustrated by 2 distinct clusters of terms that frequently converge. The results first highlight the importance of eHealth and interoperability in shaping the field. This finding aligns with previous studies on the concept of eHealth, which emphasize understanding “health” as an ongoing process rather than a fixed outcome, and stress the role of technology as a tool to support—rather than replace—humans [147]. The 2 clusters also reflected the focus on 2 HITs associated with interoperability concerns: (1) EHRs and (2) IoT. Additionally, standards development—as a response to interoperability limitations, including those related to semantic interoperability—was grouped within the first cluster, along with EHR. This highlights the challenges surrounding EHR interoperability and underscores the importance of semantic interoperability and standard-based solutions for this HIT. By contrast, emerging technologies—such as blockchain, middleware, and cloud computing, which can also contribute to interoperability—along with privacy and security, formed the second cluster, centered around IoT. This grouping reflects the complexity of interoperability in the IoT domain, which encompasses concerns related to privacy and security, as well as a strong interest in leveraging cutting-edge technologies to address these challenges.

Factorial analysis has revealed 2 fundamental factors that shape this dynamic field. Central to these are key concepts such as EHRs, FHIR, and semantic interoperability, which collectively form the backbone of innovation and collaboration in health care data exchange. Notably, the factorial solution aligned with the cluster analysis in identifying 2 distinct conceptual groupings that characterize the field. However, the specific membership of elements within each group differed. This discrepancy can be attributed to the fact that cluster and factor analysis partition variance through different methodological approaches, potentially leading to variation in how elements are grouped.

In addition, the thematic arrangement observed in the post–COVID-19 context was expected, as health care systems were compelled to prioritize interoperability to enable pervasive and responsive services [148]. The urgency to contain the pandemic underscored the critical need to manage a wide range of population health data generated by diverse technologies and systems [148]. Consequently, interoperability emerged as a central concern. Within this framework, EHRs gained heightened significance due to their role in systematically capturing individual health information. Notably, openEHR—a standard that facilitates multiple levels of interoperability, including semantic interoperability, and integrates domain knowledge from medical experts—maintained and even strengthened its relevance during this period [149]. The literature also underscored the recognized importance of the IoT during the COVID-19 pandemic. However, the widespread implementation of IoT solutions raised significant concerns related to security and privacy. In this context, blockchain emerged as a promising technology, offering features such as decentralization, traceability, transparency, and trustworthiness to address these challenges effectively [150]. Despite the considerable disruptions caused by the pandemic, the concept of ontology retained its relevance, continuing to attract scholarly attention and exploration.

The highest-order facets of this multifaceted domain—namely, object, source, content, and ambit—form a foundational framework that weaves a comprehensive understanding essential for advancing health care interoperability and ultimately enhancing patient care.

We adopted a systematic approach to data selection and assessment to minimize bias in our research. Nonetheless, several limitations must be acknowledged. First, in our effort to achieve a comprehensive analysis, we consulted and integrated data from multiple databases. Despite our best efforts to make well-informed decisions regarding data merging, cleaning, and preparation, these processes inherently involved data wrangling and manipulation, which may have introduced some bias. Second, our study is temporally constrained, as it focuses on articles published from 2011 onward. While this scope enabled us to capture the most recent and relevant developments, it also limited the inclusion of earlier foundational work. Third, although our literature review was conducted systematically and informed by the outcomes of factor assessments, we recognize that the risk of selection bias may still be present. Fourth, we developed a taxonomy through a systematic literature review process, making our best efforts to avoid bias; however, this process was grounded in category saturation based on a sample. Thus, this taxonomy has yet to undergo evaluation. Conducting such an evaluation is crucial for its final application and, potentially, for propounding policies to improve levels of interoperability in health care. Additionally, there is a significant opportunity for future research to integrate the higher-order dimensions we have identified with existing specialized frameworks. By leveraging faceted taxonomies, we can provide richer, more nuanced descriptions of this complex topic, enabling the development of terminology that facilitates effective integration. Fifth, the creation of ontologies that enrich the semantic context of these identified facets holds great promise for advancing our understanding and application of interoperability in health care. Sixth, technologies used to provide interoperability from the perspective of the supply chain, such as blockchains [151-154], present limitations and barriers that require further study—in terms of both their capability to provide interoperability and their own interoperability. Additional studies are also suggested regarding regulation, which could be seen as a barrier to the standardization of interoperability [155].

A previously conducted bibliometric analysis on interoperability in health care [31] identified 3 clusters based on abstracts: eHealth information stakeholder needs—which included terms such as “electronic health record” and “standard”—eHealth information quality assessment, and eHealth information technological governance trends—which incorporated terms such as “blockchain,” “privacy,” and “security.” By contrast, our study identified 2 clusters, with the term IoT acquiring high relevance. One potential reason for this discrepancy may stem from the selection criteria and the volume of data used—the earlier analysis assessed 63 articles with the highest citation counts [31]. Nonetheless, we concur that one of the clusters identified in our research pertains to EHR and standards, and the other to the use of advanced technologies, such as blockchain, to provide solutions, as well as to concerns such as privacy and security.

In alignment with the survey on interoperability requirements [9], our study acknowledges the significance of standards as a crucial attribute of the topic. The classification of standards presented in the aforementioned survey could enrich the subfacet related to the thematic scope of standards.

Ultimately, we believe that the contributions of our study—specifically, the classification of topics through cluster and network analyses, along with the classification of attributes based on a comprehensive review of selected studies—can offer valuable insights for reevaluating the ongoing debate on the classifications and levels of interoperability in health care.

Our research compellingly illustrates the vast domain of interoperability in health care, analyzed through the lens of supply chain management. eHealth emerged as a pivotal topic within this domain of knowledge, and the interoperability of EHR and IoT represents 2 key thematic categories encompassing several efforts. Finally, 4 critical attributes of interoperability were identified: (1) source, (2) content, (3) ambit, and (4) object.