A digital health ecosystem ontology from the perspective of Australian consumers: a mixed-method literature analysis

ABSTRACT This study presents an ontology that scopes the digital health ecosystem from a consumer–centered perspective. We used a mixed-method analysis on a set of papers collected for a comprehensive review to identify common themes, components, and patterns that repeatedly emerge within Australian-based digital health studies. Three major and four child themes were identified as the foundational aspects of the proposed ontology. The child themes have more precise concept definitions, inherited and distinguishing attributes. Out of 179 recognized concepts, 33 were related to the Healthcare theme; 23 concepts formed a cluster of employed devices under the Technology theme; 40 concepts were associated with Use and Usability factors. 60 other concepts formed the cluster of the consumer–user theme. The theme of Digital Health was seen as being connected to 2 independent clusters. The main cluster embodied 21 extracted concepts, semantically related to “data, information, and knowledge,” whilst the second cluster embodied concepts related to “healthcare.” Different stakeholders can utilize this ontology to define their landscape of digitally enabled healthcare. The novelty of this work resides in capturing a consumer–centered perspective and the method we used in deriving the ontology – formalizing the results of a systematic review based on data-driven analysis methods.


Introduction
Over the last decade, digital health (DH) platforms including applications and tools have enabled new approaches to rapidly deploying digital capabilities, with some promise in their ability to reform healthcare service delivery. 1-3 DH platforms have grown even more widespread during the recent COVID pandemic, 4 and are now regarded as indispensable in improving global health beyond the COVID-19 pandemic. [5][6][7] The creation and successful implementation of DH platforms, however, depends on a well-developed understanding of health service environments.
Understandably, different institutions and organizations prioritize their own requirements in relation to DH solutions, and in addition, they will often approach development or procurement, and implementation, in unique ways. For this reason, there has been an increasing interest in understanding the implementation issues of DH solutions. This challenge is observed through both organizational and stakeholder lenses. 8 At the organizational level, most health information studies seek operational viewpoints to explore how the transformation of healthcare occurs within a company or a country, how health products and services should be delivered in those settings, or what are the best organizational structures and optimized business models to offer DH services. 9 Literature suggests several vital factors for the successful delivery of the DH solutions including the processes and mechanisms of development, the necessity of administrative and logistic tasks, information exchange among users as well as ethical and legal aspects. 10,11 Some researchers have investigated how the involved stakeholders are engaged in developing a value-based DH marketplace. [12][13][14] They argue that emerging DH corporations adopt value propositions. Such corporations define very different business and service models compared with traditional healthcare organizations because they aim to deliver value that different customers can expect. These specific markets often focus on self-healthcare services via simple applications. They also often aim to leverage preventive DH technologies, including telemedicine, that connect healthcare service providers and customers in continuous remote service. 8 Health information that is collected, used, and distributed dynamically is also observed as having significant value from an organizational perspective. Cresswell et al. 15 recognized that health information requires strategic management to fill the gaps between planning, design, and implementation of DH. Organizational management of health information can further shape the existing capabilities of digital innovation to better provide delivery services within the health context of countries (See, for example, [16][17][18] Beyond the complexity of DH at an organizational level, research has also examined relevant digital transformation issues at the user level. There have been attempts to better understand the functioning of DH technologies; for example, how DH can satisfy users' requirements and preferences or how technologies aid and encourage users to engage in healthy lifestyles. 19,20 Others have explored the design and development of new technologies to improve users' health-related quality of life, or better understand the interaction process between users and technologies. 21,22 Also, factors like the adoption of health information technologies, and DH systems' usability have been highlighted in digital transformation studies in healthcare. 23 In thinking about the impact of digital transformation at the human level, policies made by authorities or the actions of providers, clinicians, patients, developers, and managers of health technologies, can have significant implications. However, in considering the human side of such transformation, it seems that there is a lack of human-value orientation in the evolution of the surrounding technical ecosystem. To the best of the authors' knowledge, little is known about how these two phenomena are related or may intersect. Therefore, in this study, we aim to discern how a digitally enabled healthcare ecosystem can be described when consumers are taken into account.
The following section describes the key concepts associated with ecosystem and ontology.

What is the digital health ecosystem?
For the purposes of this paper, the term Digital Health Ecosystem (DH ecosystem) is seen as a subdomain of Digital Ecosystem. 24 A digital ecosystem is "a self-organizing digital infrastructure aimed at creating a digital environment for networked organizations that supports the cooperation, the knowledge sharing, the development of open and adaptive technologies and evolutionary business models." 25 In other words, a digital ecosystem represents a constructed network of information systems and communication technologies that provide sustainable growth and productivity for society, industry, and government. 26 Applying this definition to the health domain, we consider a DH ecosystem as a specific information system network that aims to provide a variety of essential, digitally-enabled services to patients, clinicians, and providers, and thus to support health and wellbeing. Concepts around DH platforms and ecosystems have previously been examined in the literature. For example, in a study by Liyanage et al., 27 the authors explored such an ecosystem from the perspective of semantic interoperability in the setting of how health information is exchanged amongst stakeholders. In another work, Marcos-Pablos et al. 28 developed a general business model to exploit a health ecosystem. They proposed an ecosystem based on software platforms that centralize the actors' interaction. By focusing on caregivers, they depicted the linking collaboration between ecosystem actors who they classified as "primary (patients, relatives, and caregivers), secondary (care managers and health practitioners), and tertiary (members of public administrations and research networks, platform suppliers and operators, pharmaceuticals, medical hardware and software providers, as well as the ecosystem orchestrator)" (p.842). Finally, Iyawa et al. 17,29 attempted to outline components and strategies that facilitated the construction of Digital Health Innovation Ecosystems in Africa.

What is an ontology for the digital health ecosystem?
Ecosystems are modeled, visualized, or analyzed by various methods, such as computational models, conceptual frameworks, cluster analysis, taxonomy, or ontological approaches. 24,30 In this paper, we focus on constructing an ontology-based domain knowledge representation as an essential part of the modeling process for ecosystem analysis. [31][32][33] From the philosophical point of view, the term 'ontology' denotes a formal specification of a conceptualization for representing, sharing, and managing knowledge of a specific discipline or field. 34 In the context of information systems, this term has been frequently used to represent the structure of systems by understanding their constituent elements and the functions and behaviors that the systems may exhibit within their associated environment of actions. 35,36 Generally, ontologies play a significant role in defining a higher-level understanding of system components and their relationships, as they describe the domain of knowledge. In the setting of this research, an ontology could provide opportunities for gaining comprehensive knowledge and a better understanding of the consumer-oriented perspective of the DH ecosystem.
Several types of ontologies are introduced in the literature. A "domain ontology" represents a specific realm of interest (e.g., emergency health) and presents a particular perspective on some part of the reality of that realm. 37 An "upper-level ontology" describes generic knowledge that holds across many fields. This type of ontology provides a high level of abstraction and has a capacity to provide an ontologically sound and valid construct for defining standards, 38,39 and the uptake of general elements and basic concepts, and the relationships that exist between them.
In this research, we adopt a hybrid approach toward creating an ontology since it is a combination of the above two types and can better match the requirement for describing a DH ecosystem. 40 This approach allows new elements to be appended to the ontology without having to modify the earlier mapped shared terms and concepts. 41 The hybrid approach also allows the incorporation of other relevant ontologies already existing in the various domains of the health sector. 42 In other words, a hybrid approach enables the extraction of concepts that are employed or defined in other DH-related ontologies, which may even be independent of a particular problem or domain. 43 On this note, this approach provides a significant practical advantage, especially when there is no golden standard ontology to use for ontology construction, development, or comparison. However, the resulting structure of this work should not be confused with taxonomy, which aims to present the knowledge as classes and sub-classes. 44 Instead, in this work, when we use the term ontology, we refer to the formal specification of DH domain concepts and their relationships.

Objective
The general objective of this study is to shed some light on the domain of DH in Australia. More specifically, this research aims to improve our understanding of the DH ecosystem by defining its components from an Australian consumer perspective and expressing these in an appropriate ontology. This research further intends to depict an integration of fragmented sections within the DH industry of Australia.

Materials
Since designing any particular ontology requires domain knowledge, we included a set of papers that were collected for a comprehensive review of DH from the perspective of Australians. 45 The studies were identified through a systematic search method, including pre-defined selection criteria, via several online databases.
In addition to searching for studies on health information technologies, or health issues for specific cohorts, we aimed to establish a comprehensive review of a broader landscape of digitally enabled healthcare during the period 2014 to 2019 inclusive (i.e., pre-COVID). The review included exploratory, qualitative, and quantitative studies -as well as white papers (gray literature) -associated with the review's objectives.
The literature review sought to cover the topic of DH specifically while including the perspectives of Australian consumers, either directly or indirectly. In summary, the search of databases returned a total of 3811 articles, of which 98 papers were reviewed in this study (see Appendix A), including 14 review articles, 14 gray literature, and 70 experimental research papers published in quality outlets.
The search strategy, selection criteria, and the database were published in a separate and independent scoping review paper. 45

Analysis
This study presents a mixed-method analysis (qualitative and quantitative) to explore how the Australian DH ecosystem is portrayed in the literature. We used inductive and deductive approaches to identify common themes, components, and patterns that frequently emerge in Australian-based DH studies. These approaches were integrated and aligned with the four steps for developing and evaluating ontologies proposed by Delir Haghighi et al. 46 The first step in this process was to specify the objective and scope for the ontology. To progress to the second step, which is to acquire knowledge and derive key concepts, we identified a set of relevant papers for a systematic literature review process. For the third step, a qualitative analysis of the content of the identified documents was conducted by three reviewers to build a deeper understanding of the factors at play in the DH ecosystem and to generate the key concepts of the ontology. In parallel, we conducted a quantitative analysis of the text of the papers by using descriptive statistical and existing automated text analysis methods to generate the key concepts automatically. This analysis was employed further in the final step where we evaluated the ontology to determine if it is representative of the major sub-domains and concepts of the DH ecosystem within the Australian context. The main processes of the qualitative and quantitative analysis are described in the subsequent sections.

Quantitative analysis phase
To identify the DH ontology's core concepts, we initially analyzed the selected set of 98 papers in an inductive approach. Word frequency analysis was employed to determine if potential high-level concepts emerged from the content analysis. In this phase of the analysis, all papers were processed by NVivo 12 Plus software by automatic coding of word frequency. A word cloud was then generated to provide a visual representation of the word frequency tabulation of the concepts used in the papers.
The analysis revealed several repeated words and phrases. The authors discussed them in order to optimize the reliability of the analysis and select major keywords, as well as the classification of related terms. The chosen words and terms were arranged by a bottom-up strategy from concepts that could be grouped to build a general class. 47 The first author transformed the words into a schematic map to show the nodes and the connections between terms/phrases. An early draft of this ontology was shared with all researchers for the purposes of editing and updates during a deductive analysis. This was done alongside a review of the literature for preconceived themes. Each reviewer also identified candidate concepts to be included in the ontology. Then the review team determined the priority level of each concept, its definition, and relationships with the other concepts.
In the next step, we employed automated coding (i.e., using the existing coding patterns implemented within NVivo). The coding was run on the papers' content to generate the classifications of concepts and the possible correlations among them. Text frequency cluster analysis was used to outline similarities or variations on multiple dimensions of concepts. [e.g., data, clinical data, privacy data, etc.]. The cluster analysis of the frequently used words, in this case, was conducted by excluding jargon and proper names (e.g., author), numbers (e.g., years), and auxiliary verbs (e.g., could). Pearson's correlation coefficient (which ranges between −1 to +1 and the coefficient close to +1 indicates the strongest clusters correlation) was used to find similar words clustered more closely to each other; this, in turn, implied a closer thematic relationship between those words. Notably, the technique known as the "farthest neighbor" method was also applied to produce a hierarchical clustering of the concepts. The method takes the maximum distance between the clusters of words to find their proximity. This method assisted pattern identification and provided a dendrogram, which is widely used to represent a diagram of clusters' hierarchical trees in data mining and pattern recognition. 48,49 Figure 1. Word cloud analysis of the literature. Image "A" represent the world cloud analysis produced for merged experimental and all other papers. Image "B" represents the analysis conducted on only experimental papers. Image "C" represents the world cloud produced from review and gray literature.

Qualitative analysis phase
As part of the deductive analysis, the ontology development phase included studying the articles and coding them based on qualitative text analysis. Indexing of the articles was applied in this phase by three independent coders (AOA, FB, CB). They screened the content of the papers via weighted components or existing codes (e.g., digital health, technology, health, and data). In other words, each coder read the texts and applied their judgment on the presence of seminal concepts that had proximity to the codes. These concepts were mapped into the basic schematic that was developed in the quantitative phase. The next step was a collective discussion and comparisons of the independent analyses' results. The research team also negotiated disagreements about the extracted concepts and their relationships.
Any new conceptual items not listed by means of word frequency or cluster analysis were added to the ontological mapping. The research team discussed these new components and their relationships to the previously identified connections to enhance the level of validity and reliability of the ontology. To test if consensus existed in the identification of DH components, we assessed the inter-rater reliability on the components that were extracted and grouped under one specific cluster. The internal consistency among the team's derivation attitudes of DH components was measured as reliable, as seen in Table 1.

The evaluation phase of the ontology
Ontology evaluation can be achieved through several methods. 50 Drawing upon the methodology of Delir Haghighi et al. 46 to model and analyze ontologies, we employed a mixed criterion-based and data-driven evaluation method to ensure the ontology was constructed correctly. These methods are recognized as the most efficient approaches for gauging the precision of a complex ontology. [50][51][52] Depending on the domain, size, language, and construction procedures of an ontology, the verification and validation can be performed with respect to several specific criteria. 53 We assessed the quality of the ontology by the criteria listed in Table 2 as the most relevant to our case. It has been suggested that there could be subjectivity in selecting which criteria should be used for the evaluation of an ontology. Hlomani and Stacey 58 noted that cognitive biases could challenge good science that strives for the objectivity of evaluation. For example, cognitive biases such as actor-observer, anchoring, confirmation, and false consensus, could interfere with the reasoning process while developing the ontology by multiple researchers. To address this possibility, three experts in the DH domain conducted the evaluation, critically and constantly triangulating and inspecting each other's reasoning and decision-making pertaining to the structure of the ontology.

Results
Initially, the results of word frequency analysis are reported. We grouped the results by two different categories of papers: A) experimental papers; and B) review articles and gray literature. Each group of papers was analyzed separately. It was aimed to ensure a comprehensive understanding of their content and trace any discrepancies between academic and non-academic perspectives on DH presented in this material. As shown in the resultant word clouds (Figure 1), the content analysis run on all papers indicated the most frequent terms to emerge from the text mining. Health, Usefulness, Patients, Information, Data, Services, Participation, and Access were the most frequent terms. When compared, the experimental and gray literature groups of papers provided some common concepts (e.g., Health, patients, information), but they were ranked differently with respect to frequency. In addition, some unique concepts were found in each of the categories, as we have highlighted in Table 3. 1 However, the experimental studies appear to largely examine the usefulness of technologies in selfcare, aging, and also studying tasks related to managing diseases, such as diabetics. In comparison, the white or gray literature and review papers appeared to have focused more on concepts such as telehealth, medications, healthcare records, and digitalization in the Australian community.

Qualitative development of the ontology
Informed by the emergent key terms in the above word cloud, we conducted an in-depth qualitative analysis of the papers and synthesized the DH ecosystem ontology. We identified three major themes that embody the foundational aspect of the proposed DH ecosystem ontology. Theme 1 (Digital Health) operationalizes the concept of digitally enabled healthcare. Theme 2 (Consumer/User) denotes the network of users, including patients and macro-level customers of digital care (e.g., organizations). Theme 2 also covers concepts that define the needs of DH consumers. Theme 3 (Technology) identifies the technical elements and types of tools that are used to facilitate the delivery of healthcare services in a DH paradigm. By drilling down to the level of the components in the ontology, three of five associated components were interlinked to the main themes. Juxtaposed with the DH theme, data, information, and knowledge are the three major concepts that were consolidated and formed the first principal component. Consistent with the study by Dalrymple, 59 this component was branched into subjects, values, principles, processes, and functions to better describe how the digitalization of healthcare knowledge occurs and how health-related data is created, transformed, managed, and applied to the domain of DH interests. Refers to whether an ontology effectively communicates the intended meaning of its defined terms and contains objective definitions that are independent of a particular context Gruber 35 Accuracy States that the definitions, descriptions of classes, properties, and individual elements in an ontology are correct Gruber 35 Extendibility Demonstrate whether a user is able to define new terms for special uses based on the existing vocabulary of an ontology, in a way that does not require the revision of the existing definitions Gruber 35 Adaptability Measures how well an ontology anticipates its future uses and whether it provides a secure foundation that is easily extended and flexible enough to react predictably to small internal changes Vrandečić 54 Coverage/ reusability Refers to degree of reuse across domains or a different purpose or to build other ontologies Obrst et al. 55 and Duque-Ramos et al. 56 Expandability Refers to the ability of an ontology to be extended in order to describe specific application domains in a way that does not change its current definitions Gómez-Pérez 57

Preciseness
Refers to correct representation of aspects of the real world and the fraction of retrieved instances by the ontology that are relevant Vrandečić 54

Conciseness
Refers to the absence of redundancies including redundancies that could be inferred from its definitions and axioms Gómez-Pérez 57 1 Table 3 lists the top 21 words that are weighted about 30% of all 400 words identified by this analysis. The summary of the analysis can be found in the Appendix B.  "Healthcare" is the second component with a foundational association to the DH theme. This component defines the processes that deal with patients' medical conditions. As a node in the ontology, it ties several concepts to the DH theme, including the concepts of medical services and support, maintenance and improvement of wellbeing, as well as consumer healthcare issues. The healthcare component is also connected with clinical practice and consumer health education concepts.
Finally, "Health" was identified as the third key node in the ontology with the second level of connection to the DH theme through the healthcare component. Compared to "Healthcare" the "Health" node defines demands of a subject of care (patient) and their issues or problems that are identified by a healthcare provider (doctor or hospital). In other words, health is the component that relates to the state of being free from illness or injury. This component is connected with the concepts of illnesses, diagnosis of diseases, prevention methods/models, treatments, and medications.
The two remaining components have a direct relationship to the technology theme. "Use" and "Usability" describe factors for better formation and the best application of digital technologies. The concept of usability describes the experience of interaction of consumers with tools, systems, and tasks -this has enormous significance in the health and DH settings. 60,61 On the other hand, there is also a separate concept of "use" that encapsulates the broad sense of the employment of DH technologies in the healthcare industry. In the setting of this research, "use" is a concept that implies some level of evidence for a positive impact of DH on consumers or healthcare delivery. That said, proper design, accurate deliverables, empowerment of patients, and the safe management of consumers' conditions are also fundamental considerations in assessing the effective use of DH technologies in the DH ecosystem.
In a deductive analysis of papers, 179 concepts that shape the DH ecosystem ontology within the Australian context were identified. A total of 33 concepts established a relation to the healthcare component of the DH theme. The DH theme was also associated with 2 independent sub-concepts, i.e., DH strategy and funding. Regarding the technology theme, 23 sub-concepts grouped the technologies and apparatus employed by DH consumers. This theme was also linked to another cluster of 40 sub-concepts associated with use and usability. An additional 60 concepts make up the cluster connected to the consumer/user theme. Finally, the "data, information and knowledge" component was linked to the DH theme by a cluster of 21 concepts (See Appendix C for more details).
Last but not least, the ontology specifies the relationships between its components. However, we have not directed the relations among the associated nodes that stand for concepts within a domain and edges for connections between those concepts. 62 The undirected relationships have created a neutral ontology structure without any hierarchical ranking. An undirected structure ontology has allowed the parent concepts to be equivalent to the children's concepts. An undirected graph ontology also has avoided semantic confusion, which may be caused by defining a unilateral direction between two concepts where the property of one is only defined in a unique direction to another. So, for example, "the Phones includes Text-SMS" represents only an inclusion relationship between the components of Phones and Text. To illustrate the relationships that hold among the components of the ontology, 16 different relational statements were inserted via links; the defined relationships have role, dependency, inclusion and influence characteristics. The themes, components, and their relations are illustrated in Figure 2.

Automated analysis
To further evaluate the data patterns that emerged from our qualitative analysis, we employed an automated coding and clustering approach utilizing NVivo software (version 12 Plus). We developed a dendrogram as a diagrammatic representation of hierarchical clustering, as represented in Figure 3. The resultant dendrogram illustrates the clusters' arrangement and shows the distance between or any dissimilarity.  Health (F = 1490), 2 Information (F = 647), Services (F = 628), Use (F = 598), Patient (F = 560), Data (F = 484), Care (F = 432), Technology (F = 386), Online (F = 385), Healthcare (F = 365) and Systems (F = 311) were recorded as the frequently emerging themes. It can be seen that the nearest neighbor clustering couple were "technology-use" and "healthcare-systems." The "patient" was another prominent single cluster in close relation with the group of the technology-use dyad. "Data" and "Online" were other terms frequently grouped together but further distanced from the other clusters.
The current ontology reflects a good level of clarity and accuracy. Firstly, the ontology concepts were extracted and organized by exploiting the grouping possibilities. From a semantic point of view, the concepts that had meaningful relationships, such as "usability in technology" or "consumers" were organized in a single cluster. Secondly, the concepts, components, and classes that construct this ontology match correctly with the definitions, descriptions, and language that global experts and DH advocates employ. In other words, the ontology communicates the key concepts and their relationships clearly, and in a way that the reader can easily understand.
The ontology also has high precision. At the semantic level, for instance, the repetition of concepts having the same meaning (e.g., diseases and illness) was avoided, and concepts that may create confusion (e.g., Data and DH strategy or tech and health literacy) were differentiated. The abstraction level was also placed on the same hierarchy level. This enabled the conceptual modeling to transform an attribute from a concept to a higher class or vice versa. For instance, the concept of "wearables" ascribes a specific technological property to other concepts such as "telemedicine, AI or sensors" that are connected from upper parent levels.
The ontology also has a dynamic structure that can ensure its reuse/coverage across different healthcare and technology domains. It covers the broad scope of DH and has the ability to describe the specific areas within it from the perspective of different health consumers.
Consistency is also observed between the components of the ontology. At the outset, there was a consistent process among the coders to extract the concepts and generate the relations between them. Secondly, the automatic clustering results verified that the produced components and classes corresponded to the outputs of manual thematic qualitative construction of the DH ecosystem ontology.
Conciseness is the other criterion that is satisfied. The qualitative evaluation process eliminated the definitions or concepts from the lexicon that were unnecessary or inaccurate for the DH ecosystem. Rather, the set of concepts in this ontology aligns well with regard to consumers and the health and information technology domains.
We assert that even if not all terms, then the essential concepts of the DH ecosystem, for example, mHealth, telemedicine, data privacy, benefits, values, diseases, age and gender, and the relationships between them, are very well covered in the proposed ontology. Where terms looked like synonymsinter-author agreement was sought to represent the best to eliminate any potential ambiguity or redundancy in the ontology. For example, the concepts of "online" and "internet" are often used interchangeably, and it could be argued that there is no line that demarcates one term from the other. Both terms are a subset of the concept of "digital technologies" which in this setting, includes digital devices, channels, and platforms that are used to promote healthcare aims. The authors, however, segmented the terms, arguing that, in the case of online health services, a live internet connection is necessary for them to be practically useful. Online healthcare provides services to patients with healthrelated issues or is used by clinicians, hospitals, and clinics, and even by insurance companies, to support patients. In contrast, a broad range of health-based services, including education and 2 F stands for Frequency administration, are delivered using internet-based healthcare telecommunications; the services were provided synchronous or asynchronous. Hence the two concepts (i.e., online and internet) were utilized separately in the proposed ontology.
Additionally, the ontology offers the conceptual foundation for a range of tasks that can be important for DH implementation from the consumer's perspective. More importantly, this ontology is also a sound basis for extension, expansion, integration, and adaptation.
The proposed DH ecosystem ontology is flexible and can be expanded in different contexts. The ontology design allows any significant changes in the environment to be incorporated into it without the need to remove any themes or elements and without the need to alter or revise the already identified semantics. For example, when planning or evaluating innovative DH solutions for a specific case of a rare disease (e.g., Hashimoto's Disease), the concepts associated with the relevant technologies, data-related issues, clinical information of patients (including symptoms, medical history, laboratory and physical examination) can be added to the related clusters or themes of the ontology.
Finally, the current ontology can be adopted by various disciplines. It could be developed for the purposes of knowledge representation; for instance, in hospital and personalized healthcare services, healthcare policy, health economy, and biomedical informatics.

Discussion
To the best of our knowledge, this work is the first attempt to analyze and describe, in ontological terms, the structure and function of the DH ecosystem in Australia. We propose a formal representation of the most relevant concepts associated with the DH ecosystem based on our analysis of published research, specifically including a consumer perspective. The developed ontology allows readers to identify how the multiple actors, tasks, factors, and technologies that constitute the DH industry interact in providing healthcare and wellness services.
Moreover, the proposed DH ecosystem ontology describes and integrates the knowledge in this diverse domain, capturing the interactions between information technologies, complex healthcare systems, clinical processes, health data governance, and usability factors. The ontology also explicitly describes the "consumption" of DH, since the concepts addressed have been elicited from a literature review that took a consumer-centric focus from the outset.
Our review of the relevant literature identified social and human issues as some of the main factors driving the successful development and use of DH in the Australian context. Amongst these factors, the ontology describes the age, gender, and cultural-linguistic diversity of consumers as essential factors in examining suitable healthcare processes. 63 Consistent with the existing knowledge in implementing DH, 9,15 there are several other significant human factors in the DH Ecosystem. That is, the ontology describes the "success" of DH which is critically dependent on the attitudes, motivations, and concerns of the consumers, their expectations of DH services (e.g. cost savings via telehealth), and the engagement and contribution of consumers to the design and development of DH.
The DH ecosystem ontology also lists a set of acute and chronic diseases (plus various related applications relevant to these) that have been implemented and studied in DH in Australia. The listed diseases were discussed in the literature in relation to technology-based treatments or technologyoriented healthcare services.
It is important to note that technological innovations are often reflected in scientific articles with delays. 64 Therefore, it is not always possible or realistic to expect scientific papers to report emerging technologies or interventions employed in the DH domain soon after their creation. This is why it was important to ensure that the proposed ontology was expandable and robustly evidence-based up to the point of the publication of the sample studies. Consequently, the ontology is very useful as a starting point to represent the main concepts for consideration when the new developments in DH are planned or evolve in the future.
As in several other studies, 65,66 "data" is strongly represented in our proposed ontology. As one of the key concepts, it is associated with relevant content and policies around security and confidentiality and with various ethical considerations. "Health data" was mentioned, from the Australian consumers' perspective in particular, as a significant element. It is shown in the ontology that consumers require health data to be accessible to patients as they are the owners and co-creators of this data. However, the data must be handled in a reliable, secured, de-identified (where applicable) fashion and relevant consent protocols must exist. 67 These identified concepts and associated sub-elements can be useful to inform the development and use of specialized DH solutions by relevant parties (e.g., vendors and governments). The ontology also describes a clear relationship between "health analytics technologies" and "health knowledge." This connection highlights the essential role of accurate health data processing for supporting or improving decision-making, specifically in diagnosing diseases.
We evaluated the ontology against several criteria, including clarity, expandability, and feasibility, which were completed as part of a thorough review process by the team of authors. Based on the available evidence in the literature, it is difficult to describe, and hence, follow a "gold standard" evaluation approach. However, we know that the themes and concepts representing the DH ecosystem in Australia are comparable with other developed ecosystem ontologies. For instance, referring to the Digital Health Innovation Ecosystem described by Iyawa et al., 29 the essential components of their model are very similar to our proposed ontology. The DH theme is a single cluster in their ontology that relates to several subelements such as privacy, telemedicine, and gamification. Our proposed ontology not only included such factors but also branched various elements under other common subcomponents, which is an indication of the greater clarity provided by our ontology, as well as an indication of how it can be further expanded in time.

Limitation and future steps
The introduced ontology should not be seen as a universal or final product. Rather, we believe that the developed ontology is a starting point that is based on existing scientific and semi-scientific studies of introducing digital technologies into Australian healthcare. In fact, our ontology mirrors the information which was found in the literature. If there is a paucity of information to reflect a comprehensive view of the consumers, this is rooted in the lack of studies or empirical studies in those areas, which is acknowledged and should be addressed in future studies.
The current ontology is not simply based on the output of an automatic data-mining process. It represents the output of an additional thorough qualitative analysis and conceptualization effort produced by a team of DH specialists. Therefore, we were able to ensure that the developed ontology is closely connected to the current real-world DH domain in Australia and represents the taxonomic concepts that make up the DH industry and the relationships between them as defined in that literature.
In subsequent work, we plan to add another level of rigor to the DH ecosystem ontology through external evaluation by relevant industry professionals and DH experts. In addition to our manual assessment, we will also try to develop an approach that can automatically capture semantic coherence in the proposed DH ecosystem ontology. This process would also assist in better differentiating concepts and addressing any lack of balance of the taxonomic structure.
We derived an ontology that can help direct the attempt toward a gold standard. In the next steps, we plan to employ various techniques to define the ontology in a computer-understandable way, by generating graph (e.g., by Protégé) and domain documentation (e.g., by SKOS vocabulary)

Conclusion
This study introduced a DH ecosystem ontology from the perspective of Australians. The proposed structured knowledge representation is a solid starting point for developers and researchers to understand the current state-of-the-art of knowledge in the healthcare industry, particularly based on empirical evidence from the literature. Undoubtedly the dissemination and usage of this ontology could result in a better understanding of the DH ecosystem.
In summary, the key contributions of this study are twofold. Initially, we have identified several core concepts that illustrate how consumers, technologies, and health are represented as themes and dependent key elements in the DH Ecosystem from an Australian perspective. The suggested ontology provides a means to empirically evaluate other DH ontologies based on how many concepts and definitions can be automatically detected and subsequently applied through their use in a DH Ecosystem.
Secondly, the study demonstrated the applicability of the ontology creation method proposed by Delir Haghighi et al. 46 that was originally used in the domain of emergency management and adapted for our purposes. Our novel contribution to this methodology was: 1) conducting analysis over the content of the articles and generating the ontology through the systemic review process; and 2) utilizing a data-driven approach for ontology evaluation.
Last but not least, our defined ontology could be applied as an instrument to improve the requirements of existing projects and enhance the capabilities of the community of developers and providers working in DH healthcare as supported by Serbanati et al., 24 Specifically, Moreno-Conde et al. (2019) 68 recommend ontology as a cornerstone for future innovation. In other words, the adoption of ontology would assist in improving the future design and innovation of DH technology which are currently made with an imperfect understanding of the needs of consumers, core concepts, and relationships between them. Additionally, the adoption of the defined ontology by healthcare management centers will allow them to coordinate data collection that may contribute to better decision-making over the movement, use, and health behavior change of DH consumers.

Disclosure statement
No potential conflict of interest was reported by the author(s).

Funding
The author(s) reported there is no funding associated with the work featured in this article.

Author contributions
FB and AOA conceived of the presented idea and that approved by CB. AOA and FB developed the theory and AOA performed the computations and analyses. FB verified the analytical methods. AOA drafted the manuscript, and it was revised critically for important intellectual content by all authors. All authors approved the manuscript for publication and agree to be accountable for all aspects of the work.
Abraham He is passionate about multidisciplinary studies where cognitive and information systems meet each other and his current focus is on consumer health informatics. He aims to understand how digital healthcare technologies are perceived by the public, how they should be designed to accord to consumers' health requirements, and how innovations should transform the activities of the healthcare industries in the best ways to deliver excellent customer service to citizens. Abraham also works in Bayesian Networks (BNs) that are applied to support decisionmaking under uncertainty; especially, health-related problems that are targets for medical decision support systems.
Chris Bain is a Professor of Practice in Digital Health in the Faculty of Information Technology at Monash University in Australia. Chris' position is the first of its kind in the faculty. He has more than 30 years' experience in the health industry, including 12 in clinical medicine. He's led numerous software development and implementation projects in the health industry and works with many faculties and Institutes across the University, as well as with a range of health industry partners, in leading the Monash efforts in Digital Health.
Frada Burstein is a Professor (Adjunct) at the Faculty of Information Technology, Monash University, Melbourne, Australia. At Monash University, Prof. Burstein led multiple projects on the use of technology in the healthcare context. Her current research interests include business intelligence, clinical decision support, and health informatics. Her work appears in leading journals such as Decision Support Systems, Journal of the American Society for Information Science and Technology, Journal of Medical Internet Research, and others. Prof. Burstein has been a guest editor of a few special issues of journals and collections of research papers. She is a Fellow of the Australian Computer Society and Distinguished Member of the Association for Information Systems. Full research profile available at: https://research. monash.edu/en/persons/frada-burstein