1 Introduction

The healthcare sector faces a multitude of challenges such as, among others, an aging population, increasing demand for personalised care, a shortage of health professionals, suboptimal task allocation, and rising operational costs. While not a panacea per se, digitisation does offer substantial opportunities to help in many of the challenges faced by the healthcare sector [1,2,3]. Personal Health System (PHS) Technologies, including health data monitoring, intelligent processing, evidence-based interventions, and active patient engagement [4], can mitigate clinical errors, aid healthcare professionals, improve information management, and enhance patient access to health services [5]. However, the integration of digital technologies and data sharing into existing healthcare practices often does not fit well with established institutional norms, with data sharing in healthcare being perhaps one of the most prominent "headache" topics [6]. Data sharing in healthcare often conflicts with deeply rooted values like privacy, trust, and control [7], and with ethical codes and GDPR regulations regarding doctor-patient confidentiality and healthcare data classification. Actors in the healthcare domain encounter various challenges related to data-driven healthcare innovations, including a lack of trust to share data, concerns about personal health data re-identification, strict or unclear privacy regulations, and limited governance frameworks for data access and security [8].

In both academia and industry, secure and privacy-friendly digital data sharing solutions have been developed that are tailored to institutional health practices in a ‘by design’ [9,10,11,12,13] approach. Privacy-Enhancing Technologies (PETs), referring to a group of different technologies, play a significant role in this effort, aiming to enhance both privacy and security. This, in turn, facilitates data sharing within existing data technologies where privacy and security serve as limiting factors. PETs contain various techniques, including anonymisation, pseudonymisation, and data attribute obfuscation, often relying on mathematical algorithms and cryptographic methods for cross-organisational analysis while safeguarding privacy [14, 15]. For a comprehensive understanding of PETs, we encourage the readers to explore the resources provided by the Information Commissioner's Office [16]. Over the last decade, PETs have become more accessible and commonplace [11], moving from lab-and experimental settings towards the market [14, 17]. Yet, despite progress made on PET techniques, many PET developments seem to remain at the proof-of-concept stage [14, 18]. Resolving the privacy, trust, and control issues described above through technology design alone may not suffice for the adoption of digital solutions in the healthcare sector. Given that PETs offer a broad range of solutions that mitigate or resolve privacy and trust-related concerns in healthcare, it raises the question of why their adoption is not faster and more widespread.

The widespread adoption of PETs remains elusive, as indicated by Klymenko et al. [19]. Well-known challenges in technology adoption and diffusion persist [2, 6, 11, 16]. Achieving widespread adoption requires research into methods that extend beyond mere technical constraints [25,26]. However, co-creation in a multidisciplinary setting also poses its own challenges, as different actors bring different perceptions of the problem, perspectives, interests and values that are not always congruent [27].

Transdisciplinary (TD) is an emerging field focused on co-creation in complex value-constellations. It stresses the need for interdisciplinary consciousness to enable change, which involves ‘a kind of know-how exhibited by individuals involved in a group process who can actively empathise with different disciplinary perspectives’ [28]. TD methods such as frame reflection and the dynamic learning agenda facilitate reflection-in-action, enhancing the ability to switch between different frames to solve unique problems [28]. Frame reflection helps to uncover patterns in first-order reflection (problem-solving) and second-order reflection (values and assumptions). The dynamic learning agenda promotes mutual learning by collectively identifying, analysing, and prioritising common challenges. It stimulates emerging designs, learning from experiences of others, participation of stakeholders, and facilitation of dialogue and reflection [29]. These methods prioritise collaborative framing and action, without the need for consensus, emphasising the collective deciphering of complex topics by connecting previously unrelated elements [30]. TD methods stand out from co-creation methodologies by including a wider range of disciplines, and integrating diverse knowledge and expertise beyond the academic field, thus addressing the challenges faced by PET adoption, particularly in bridging the academia-industry gap.

1.2 Related work and study aim

Most studies that have investigated the challenges of technology adoption in healthcare [3, 5, 6, 8, 19, 22, 31,32,33,34] primarily focused on identifying these obstacles. As two key examples, Witjas-Paalberends et al. [8] outlined key challenges faced by multidisciplinary teams adopting digital technologies in healthcare, including managing data variety, addressing data sharing and quality issues, implementing effective data management strategies, and overcoming a lack of big data skills. They proposed that adopting best practices, fostering trust, and reducing conservatism in the field could accelerate the integration of advanced big-data technologies and methods into collaborative healthcare innovations. Similarly, Klymenko et al. [19] emphasised the gap between academia and industry regarding PETs specifically, highlighting the need for tailored education on PETs to navigate privacy compliance complexities. Future steps involve creating learning materials and deploying e-learning platforms to bridge academia-industry gap. While these studies shed light on implementation hurdles and provide recommendations, the underlying value-driven reasons why current attempts to overcome these challenges fail remain unclear.

In this study we explore whether sustainable adoption of PETs in the health domain is impacted by conflicting logics and disciplinary disparities. By applying TD approaches to a case study around PET development in the Dutch health care sector, we explore how these approaches can facilitate collaboration and mutual learning to overcome these hurdles. The purpose of this study is to reveal, through the application of the aforementioned TD methods, the following objectives: a) underlying value misalignments and b) if such methods help in overcoming such misalignments to feed mutual learning and c) whether such mutual learning has an effect on alleviating barriers for adoption. It thereby aims to contribute to existing literature on barriers for PET adoption, as well as to literature of TD approaches for leading complex innovation processes. Diverging from previous studies on technology adoption in healthcare, this research adopts a collaborative perspective to delve into the persistence of these challenges and explore strategies for overcoming them.

2 Methods

We employed a qualitative case study methodology in line with our exploratory and empirical objectives. Via action-research that has taken place during project-in-development, several TD approaches (frame reflections and dynamic learning) were applied. While prior research has explored TD approaches for societal change [29, 35], there is a gap in empirical investigations specifically focusing on TD approaches for PET adoption in healthcare. This approach appears well-suited to uncover how different actors within the health system perceive PETs and create strategies to harmonise their diverse perspectives in the complex multi-actor ecosystem of PET design. See Annex 7.7.5. for a visual overview of the relations between methods, case study, data collection, analysis and the insights flowing from the analysis.

2.1 Case study

We opted for a case study at the start of the PET development process. TD approaches, including frame reflection and the dynamic learning agenda, are a means to stimulate action and reflection during the development stage. Introducing these methods at the start of the project provides the necessary time to evaluate their suitability for revealing diverse perspectives and promoting collaborative learning, meanwhile evaluating the hurdles the actors experience and foresee.

Building on the domains described by the NASSS framework, the selected case is a Dutch project within an international research program, containing 12 partners, including technical knowledge institutes, large companies, Small Medium Enterprises (SMEs), and University Medical Centers (UMCs). Over a three-year project, this consortium aims to develop PET infrastructure for enhanced cardiovascular risk management in personalised patient care. The case included a wide range of actors in the technology development chain which is valuable for understanding different perspectives within a multi-actor health system.

The focus on cardiovascular diseases (CVD) as the health condition (domain: condition or illness) where PETs (domain: technology) hold significant potential [31]. Historical CVD data often remains locked in data silos due to privacy concerns. PETs provide an opportunity to unlock and analyse this data collaboratively while ensuring privacy [13], potentially improving CVD management. Within the context of PETs, organisations (domain: organisation(s)) engaged in their development often overlap with those adopting (domain: adopter system) the technology. In a multi-actor health system, various entities assume roles as data stewards, data providers, or technology providers. PETs function as collaborative tools, necessitating governance rules and data sharing agreements.

The Dutch healthcare system (domain: wider context) operates within a decentralised, market-based model, with shifts in government involvement and distinct agenda-setting processes for health information technology [36]. The diversity within this system brings varied perspectives on problems, interests, and values, underlining the importance of interdisciplinary learning for sustainable change. The embedding and adaptation of PETs over time (domain: the interaction and mutual adaptation between all these domains over time) is expected to be challenging in the Netherlands as well as in other European countries, requiring a fundamental shift in assumptions regarding health data treatment [32]. PETs align with European GDPR principles [33], yet limited understanding of how organisations and adopters within the Dutch healthcare system perceive and value (domain: value proposition) PETs, remains.

2.2 Data collection and analysis

Our data collection approach comprised semi-structured interviews (n = 8), questionnaires (n = 20), and focus group discussions (n = 20) involving representatives from the organisations in our case study (see descriptives in Table 1, more background information in Annex 1), referred to as ‘actors’. We initiated the process with interviews guided by the NASSS framework, focusing on technology, value proposition, and organisations. However, we remained adaptable to include other domains emerging in discussions. We then conducted two rounds of questionnaires and focus group discussions to validate challenges and values and gather individual prioritisation before group discussions. The focus group discussions aimed to explore participants' unique perspectives on challenges and values, uncovering their underlying reasoning. In the final focus group discussion, participants collectively determined which challenges should be prioritised for the dynamic learning agenda, aligning with our TD approach's goal of stimulating action and reflection.

Table 1 Participant descriptions: organisational types and participation modes
Full size table

Conversations were transcribed verbatim and analysed using mixed inductive-deductive coding in ATLAS.ti. This hybrid method involves predefined categories from a specific theoretical framework, in this case the NASSS framework, while simultaneously allowing themes to emerge directly from the data. Combining these methods balances the structure provided by predetermined categories with the flexibility to identify new themes based on the data itself [37]. The inductive process involved breaking down the text into separate parts (open coding), connecting codes (axial coding), and selecting codes into overarching categories (selective coding). Throughout the coding process, the first, second, and last authors of this study discussed and agreed upon codes and patterns, considering their individual biases stemming from diverse backgrounds. This process led to the identification of first (perceived challenges) and second-order (related contextual values) reflections.

3 Results

The study reveals a set of eleven challenges that play an important role in the successful adoption of PETs according to the actors.

3.1 Key identified adoption challenges: Alignment of frames, project goals, approach and urgency among actors

The identified challenges are considered important for proceeding specific contextual values, including collaborative power, trust in technology, financial and business prospects, strengthening PETs, and enhancing cardiovascular prediction power. The findings are categorised according to how various actors align in framing the challenge, its urgency, and appropriate approach, reflecting the diversity in perspectives observed. Here, "framing" refers to which contextual values actors associate with particular identified challenges.

3.2 Different framing with partial alignment in terms of urgency and approach

In our study, diverse perspectives among stakeholders highlighted varying perceptions of the challenges, with some commonalities emerging. The most critical challenge according to the actors relates to achieving legal clarity (see Table 2 for the shared definition). Actors frequently mentioned their concern about how to ensure alignment with GDPR and other legal frameworks. This challenge was viewed differently by various actors, with knowledge institutes and large companies emphasising its importance for building trust in the technology. Notably, a UMC representative linked it to data quality for enhanced cardiovascular predictions, underlining the hesitance of hospitals to share individual data without legal clarity (see Table 2 for illustrative Q1). As a response to the plenary focus group discussion, in which the importance of this shared challenge among actors became clear by being prioritised as the most important challenge to be addressed; participants collectively decided to swiftly mobilise experts to address it. Following this, they arranged regular meetings to oversee the progress and implement subsequent actions.

Table 2 Shared Definition of Legal clarity challenge, related contextual values and illustrative quotes of the participants
Full size table

Another crucial challenge identified was the lack of organisational readiness, which is a two-sided challenge as demonstrated by the shared definition among actors (see Table 3 for the shared definition). They emphasised that trust cannot be established without the necessary expertise and explicit support (see Table 3 for illustrative Q2). For for-profit companies, this challenge had implications for their business prospects, potentially discouraging technology adoption by clients. However, it also presented an opportunity to offer support as a service. A representative from a knowledge institute highlighted the obstacle to collaborative power in PET development, emphasising the need for the engagement of the "right people", such as legal experts, IT professionals, and privacy and security officers. Participants recognized the multifaceted nature of this challenge and chose to collaborate on solutions.

Table 3 Shared Definition of Organisational readiness challenge, related contextual values and illustrative quotes of the participants
Full size table

User-friendliness (see Table 4 for the shared definition) was another concern, according to SME and UMC actors, who stressed the need for diverse guidance encompassing roles like Data Protection Officers, security experts, data scientists, and doctors. Such guidance was believed essential for a comprehensive understanding of PET and its value for customers. SMEs sought this guidance for securing their financial and business prospects, stressing that security should be balanced with ease of use without imposing excessive complexity or costs (see Table 4 for illustrative Q3). UMCs emphasised the need for tailored, clear guidance to instil trust among healthcare professionals (see Table 4 for illustrative Q4), critical for adoption. This approach was also seen as beneficial for addressing the challenge of organisational readiness.

Table 4 Shared Definition of User-friendliness challenge, related contextual values and illustrative quotes of the participants
Full size table

Unclear governance (see Table 5 for the shared definition) was identified as a noteworthy challenge for successful PET adoption by both SME actors and a large company. Clarity in governance is critical not only for collaborative power (see Table 5 for illustrative Q5) but also for establishing trust in the technology and achieving robust cardiovascular predictions. For the large company, this challenge raised concerns about ensuring data stewardship, given that PET influenced conventional data quality checks (see Table 5 for illustrative Q6). Addressing this challenge effectively may require alternative approaches, including (technologically enforced) agreements within the data sharing infrastructure.

Table 5 Shared Definition of Governance challenge, related contextual values and illustrative quotes of the participants
Full size table

3.3 Differences in framing with limited alignment in terms of project goal, approach and urgency

Among the identified challenges, disparities and divergent perspectives emerged among the actors. Initially, data semantics and data quality (see Table 6 for the shared definition) seemed to be points of alignment, but deeper examination revealed varying interpretations of their significance and urgency. UMCs and large companies considered these challenges critical due to the negative impact on predictive power, essential for cardiovascular applications and clinical utility (see Table 6 for illustrative Q7-8). The knowledge institute, focussing on the PET developments, saw these challenges as obstacles to technological advancement. SMEs, viewed it as an obstacle for a scalability product in the market, requiring a broader approach that includes compliance with (international) data standards. SMEs acknowledged the importance of these challenges but considered them as beyond the scope of the current project (see Table 6 for illustrative Q9). Given these differing viewpoints, stakeholders chose to concentrate on challenges where a consensus on the approach could be achieved, temporarily setting aside data semantics and quality concerns.

Table 6 Shared Definition of Data quality and data semantics challenge, related contextual values and illustrative quotes of the participants
Full size table

The concern about an unclear business model (see Table 7 for the shared definition) seemed, brought up by SMEs and large companies, was recognized by all actors, however the level of importance and urgency varied, especially at the start of the discussion. Private company representatives perceived the project as an opportunity to explore new business prospects, leveraging cutting-edge technology (see Table 7 for illustrative Q12). Within the collaborative framework, participants saw potential in each other as customers or partners, indirectly validating the PET market. SMEs found value in this collaboration for their product development and roadmaps. As the discussions progressed, other actors recognized the importance of this challenge, however, still disagreeing on the urgency. In the final focus group discussion, private company actors also agreed that business model development could be delayed to a later stage.

Table 7 Shared Definition of Business models challenge, related contextual values and illustrative quotes of the participants
Full size table

Furthermore, the challenge of technical interoperability (see Table 8 for the shared definition) was acknowledged by both the knowledge institute and SME actors, although with different value-based perspectives. SMEs emphasised its significance for market viability and business prospects (see Table 8 for illustrative Q13), while the knowledge institute's framing related to computational viability, reinforcing PETs from a technical standpoint (see Table 8 for illustrative Q14).

Table 8 Shared Definition of Technical interoperability challenge, related contextual values and illustrative quotes of the participants
Full size table

3.4 Similar framing with alignment in terms of project goal, approach and urgency

Besides disagreements, actors framed some of the challenges in a similar fashion, associating specific challenges with the same value. Regarding the nature of adoption (see Table 9 for the shared definition) of PETs, actors from knowledge institutes, SMEs, and UMCs concurred quickly in recognizing that the tangled adoption process acts as an obstacle to collaborative power (see Table 9 for illustrative Q15) within the current project and beyond.

Table 9 Shared Definition of Nature of adoption challenge, related contextual values and illustrative quotes of the participants
Full size table

Furthermore, consensus emerged among all actors regarding the challenge related to the current absence of proven value (see Table 10 for the shared definition) in technological innovation. This challenge is perceived as significant, given the assumption that the lack of established value could hinder trust in the technology (see Table 10 for illustrative Q16). This trust is essential to engage the 'right people,' often referred to as champions, in the adoption process. In the health domain, demonstrated value in the form of publications is of high importance.

Table 10 Shared Definition of Proven value challenge, related contextual values and illustrative quotes of the participants
Full size table

3.5 Single framing by one actor

The participant from the knowledge institute, who is also an expert in PETs, identified challenges related to technological maturity (see Table 11 for the shared definition). While the value is to strengthen PETs, expecting participants to surpass the existing state-of-the-art while aiming for a similar, already adopted product is deemed overly ambitious (see Table 11 for illustrative Q17). No objections or vocal consensus were expressed.

Table 11 Shared Definition of Technological maturity challenge, related contextual values and illustrative quotes of the participants
Full size table

Table 12 presents an overview of actor-identified topics of concern in relation to contextual values based on organisational type.

Table 12 There were notable differences between the topics of concern brought up by the different actors, while underlying contextual values align into five core values. The rows indicate the organisations that raised the topics of concern themselves during the interviews, irrespective of their recognition of topics brought up by other actors in focus group discussions. The columns demonstrate the contextual values that form the basis of the topics of concern according to the actors
Full size table

4 Discussion of findings

4.1 The role of TD in breaking adoption impasses by addressed underlying value (mis)alignments

Our research highlights three essential findings regarding the lack of PET uptake in healthcare. Firstly, we underscore the presence of conflicting institutional logics within the healthcare sector, with actors from diverse backgrounds prioritising distinct challenges based on their missions and value systems. This misalignment, if unaddressed during collaborative efforts, can lead to disagreements and hinder progress. Recognizing these differences and implementing effective communication and collaboration strategies are a necessity. Secondly, our study demonstrates that TD methods play a role in enhancing awareness of disciplinary disparities and fostering a more appreciative attitude toward alternative viewpoints. TD methods encourage non-defensive analyses of value judgments, facilitating interdisciplinary consciousness and mitigating internal conflicts. An active facilitator is instrumental in this process. Thirdly, mutual learning helps in promoting collaborative efforts necessary for overcoming barriers to PET adoption. Shared understanding and interdisciplinary awareness (actively empathise with different disciplinary perspectives) enable actors to identify collective actions, but tensions may arise when issues are framed differently. TD methods do not provide guidance for decision-making when actors favour different approaches. In multidisciplinary partnerships (including multiple different disciplinary perspectives), ethical and governance decision-making complexities prompt the need for further research on decision-making within these contexts and potential governmental interventions.

Firstly, the presence of conflicting institutional logics, often characterised as 'academic logic' versus 'commercial logic' in literature [38], leads to tension among actors. These logics mirror discrepancies in missions and value systems. Although the actors in this study jointly came to a list of challenges, commercial actors prioritised business-related challenges—crucial for SMEs that are heavily reliant on direct income. In contrast, academic medical centres focussed on enhancing clinical cardiovascular prediction power, while knowledge institutes aimed at technological improvements. These divergent values, if not acknowledged during co-creation, can lead to misunderstandings and disagreements, impacting project execution, scope, and success criteria. For instance, disputes over data harmonisation approaches led to the temporary suspension of the semantic interoperability challenge. Rather than leveraging diverse perspectives [24,25,26], the lack of sharing and communicating these diverging logics can lead to differences driving conflict and neglect [39]. Our findings emphasise the importance of acknowledging these differences and implementing effective communication and collaboration strategies.

Secondly, TD methods employed in our study increased awareness of disciplinary disparities by contrasting individuals' perspectives. Coupled with reflective consideration of the underlying assumptions, the discussions encouraged a more appreciative attitude toward alternative disciplinary viewpoints. For instance, the initial lack of emphasis by academic partners on the urgency of business models shifted after insights from SME actors highlighted its crucial role in widespread adoption, though preferably addressed at a later stage. This transformation is important in fostering interdisciplinary consciousness and managing internal conflicts, required for successful transdisciplinary research and practice [28]. While introducing students to diverse perspectives in their education is recommended as one key element to improve this transformation [24], our observations highlight the facilitator's critical role in professional settings to sustain and drive transdisciplinary awareness and mutual learning. With divergent interests, facilitators must maintain political sensitivity to foster constructive dialogue and sustain it, especially among conflicting perspectives. Our study illustrates that without finding common ground, participants may discontinue dialogue, hindering the learning process. An active facilitator, often termed a champion in literature [40], significantly influences the co-creation process's success.

Thirdly, our study identified mutual learning as a catalyst for collaborative efforts required to address barriers hindering PET adoption in multidisciplinary partnerships. Through the development of shared understanding and interdisciplinary awareness, project partners were able to identify collective actions needed to effect change, as illustrated by their joint efforts to address legal issues (see result section on legal clarity). However, divergent framings of certain issues led to tensions and unresolved matters, like postponing semantic interoperability decisions and action, vital for PET adoption. Privacy-by-design strategies alone are insufficient for integrating data sharing technologies into institutionalised health practices, as disagreements about challenges related to collaborative power, trust in technology, financial and business prospects, strengthening PETs, and prediction power, remain unresolved. Consequently, the feasibility of addressing PET adoption challenges in partnerships with conflicting values was questioned. Co-creation among actors with varied value systems complicates ethical and governance decision-making. It raises questions such as "which values should take precedence" and "which solution should we pursue?". These complexities are particularly pronounced in mixed-system environments lacking clear authorisation for ethical and governance decisions [41]. While TD methods help identify key decisions, they do not offer guidance when actors favour different approaches.

4.2 Methodological considerations and further research

This study utilised a mixed-methods case analysis to explore collaborative dynamics in multidisciplinary partnerships focusing on PETs in healthcare. Note that the findings are context-specific and may not apply to partnerships with distinct participants, contexts, or objectives. However, the study maintains credibility and confirmability within this specific context [42]. Credibility was maintained through extensive participant engagement, persistent observation, and member checks, verifying constructions with the contributors themselves. Moreover, confirmability was ensured by grounding our findings in data and conducting audits with fellow authors to assess result verifiability and coherence. This case study did not involve broader stakeholders like citizens with cardiovascular risk or policy makers responsible for digitisation in healthcare. The absence of these stakeholders aligns with current literature, where research and development projects frequently exclude citizens and data subjects [43].

Follow-up studies should delve into collaborative dynamics to comprehensively understand stakeholders' value systems in PET adoption. Moreover, our study primarily examined partnership dynamics in a decentralised, market-based setting where data is considered private. Variations in data governance exist, as seen in Nordic countries with public data perspectives, leading to different decision-making and social sustainability considerations. Investigating these differences provides valuable insights into PET adoption in diverse data contexts, warranting further research. For example, different data governance approaches in various regions or countries reflect unique societal values, legal frameworks, and cultural aspects. Understanding how PET adoption aligns or diverges in these contexts allows for tailoring strategies that align with local needs. Moreover, differences in data governance can impact collaborations among entities involved in data sharing. Investigating these variations helps identify collaboration patterns and opportunities, facilitating smoother implementation of PETs in collaborative settings.

In future health and technology research, a need to address the limitations of TD methods remains. This study demonstrates that TD methods help in identifying key decisions but struggle to offer guidance when diverse actor perspectives emerge. Furthermore, the challenge persists in channelling these efforts towards subsequent phases, such as spin-offs. Current policy instruments like technology transfer hubs have revealed deficiencies, highlighting a clear absence of institutional understanding. Therefore, additional methods for improving decision-making in multidisciplinary partnerships should be explored, as well clear ownership-and engagement and sustainability strategies for partnerships [44]. The latter includes strategies on ownership of ‘assets’, partnership business models and responsibilities for knowledge transfer from the partnership to the end-user of the technology. In our case study, the end-user involves hospitals and an insurance company adopting and employing PETs for new insights in cardiovascular management. A promising avenue for future investigation involves identifying and defining essential new roles in such partnerships that can prove vital for long-term sustainability within this domain [34].

5 Conclusion

Adopting PETs is key to accelerating digitisation of the healthcare sector while simultaneously upholding data protection rights and increasing cybersecurity. While such technologies are market-ready, the uptake of PETs in healthcare is lagging behind, or even stalling. In this paper, we conducted a mixed-methods case analysis to provide a deeper empirical understanding of the dynamics in multidisciplinary partnerships working on PETs adoption in healthcare. Utilising TD methods we make three key contributions that help multidisciplinary partnerships to facilitate effective collaboration and future uptake. To truly enhance sustainable adoption, further research should explore how to effectively address tensions, include end-users and patients as additional actors and to take into account variations in data governance across different regions to understand the decision-making process for change.