‘Drilling down’ and ‘breaking glass’ – Evolving notions of public service ‘quality’ in the Norwegian health data space

Abstract

The datafication of healthcare and its implications for patients and practitioners are a topic of ongoing study. We suggest that analytic benefit can be gained by (a) unpacking the sometimes ordinary origins of the promissory claims, investments, and shifts in organizations and practices that often precede large-scale health data integration projects, and (b) centring analyses on how such projects relate to institutional shifts in normative concepts such as quality in healthcare and the increasingly institutionalized vocabulary of benefit. We illustrate these features’ relevance through an in-depth empirical case study of ‘the Health Platform’ (HP, Epic) as exemplifying that of what Klaus Hoeyer terms a ‘data space’ – an interface intended to unlock potential for analytics and efficiency across the healthcare sector. To date, this process has been marked by promissory dynamics as well as delays, problems, and ultimately fierce protests. In this context, we ask how might the notion of data spaces help explain evolving notions of quality in healthcare? In response we attempt an early application of Hoeyer et al.’s framework for data-intensive health environments as an interplay of formative and experiential dimensions. Based on our findings, we propose a broadening of Hoeyer et al.'s notion of promises in relation to data spaces, to acknowledge how starting points of constraint, need, and limited choice can cascade into promissory territory.

Introduction

The rise of large formal organizations has gone hand in hand with the development of performance and quality measures and created new sites for information systems capable of classifying, representing and intervening in organizational activities (Power, 2004). In healthcare, digitalization and datafication accelerate these processes, illustrating how public services’ interdependence with commercial ICT platforms may shift parameters for trust, identity, authority and territory, as expressed via fears and promises for the future.

Examples of this were on display across Norwegian media in the winter 2023/24, following introduction of a digital platform that would unite large amounts of healthcare data from Norwegian hospitals and municipal health-care services in a single system: ‘Introducing the Health Platform: – I'm terrified’. ¹ ‘Very worried about the Health Platform: – People are going to die’. ² ‘Medical associations withdraw from the work on the Health Platform’. ³ ‘Hospital director criticizes the Health Platform’. ⁴ ‘Demonstrations all over Central Norway: We want to bury the Health Platform’. ⁵ A few months later, a scathing report from the Norwegian Office of the Auditor General would criticize the process for its cost overruns, inefficiencies, lack of oversight, and threats to patient safety (NRK, 2023a, 2023b, 2024a, 2024b, 2024c, 2024d).

The controversy around the performance of the ‘Health Platform’ (HP) contrasts with the preceding optimistic and hopeful discourses of Norwegian policymakers who had long expressed a motivation for a single, integrated, seamless and easily accessible digital system embodying the ideals of digital public administration (Rose et al., 2015). Controversies also challenged the promissory narratives of technological optimism on HP's company website:

It is a national goal to build a common health data record system that follows the health information in all meetings with the health services – with the GP, in the home service, at the health centre, the emergency room and with the specialist in the hospital. The different practitioners get different screens, but all information is in the same system. It will provide support for the many work processes healthcare personnel are responsible for, and the citizen will get a better overview of their treatment through the citizen portal HelsaMi. (‘My Health’) ⁶

To better understand these contradictory narratives, it is helpful to look to the growing body of literature about the digitalization of healthcare services and how sectoral data are being used by Nordic governments and national, regional and local administrations, with varying success (Esko and Koulu, 2023; Hoeyer and Wadmann, 2020; Kempeneer and Heylen, 2023; ; Wadman and Hoeyer, 2018). In the Nordic countries, data registration has a long-standing tradition, with data collection practices and related infrastructures developing alongside the creation of welfare state institutions and the establishment of public healthcare services (e.g. Åm et al., 2025; Hoeyer, 2023; Snell et al., 2023; Tøndel and Anthun, 2013). Data have ‘moved to the centre stage of healthcare politics’ (Hoeyer, 2023: 2), enabled by high public trust in governments’ health data collection (Andreassen et al., 2021; Sætnan et al., 2011; Snell et al., 2023). Tupasela et al. (2020) explain how the creation of ‘data ecosystems’ which ‘cater to an increased thirst for linkable and complete data’ (ibid: 10) presents both a goal and challenge across Nordic countries.

Data-intensive healthcare (Hoeyer, 2016) is often couched in ideals of collaboration and exchange, raising the expectation that quality in the healthcare services improve automatically in line with individualized healthcare and data management, such as apps and digital tools (Hoeyer, 2023; Wadman and Hoeyer, 2018). Data, when decontextualized and mobilized, give information (gathered and transferred) an aura of usability, precision, objectivity and rationality (Bartl et al., 2019). These features are then used to measure the quality of healthcare itself. In a long-term historical overview of changing notions of healthcare (service) quality, Tøssebro (2019) documents how the notion of quality has become disassociated from human interaction and judgement on the work floor, and instead increasingly quantified, measured, controlled, and benchmarked. As an object of governance, ‘quality’ appears to take on an infrastructural character, assembled through a distributed process involving different intermediaries and integration across time frames (Diaz-Bone et al., 2020).

Against this analytic backdrop, platforms such as HP can be approached as catalysts for change in ‘quality’. When HP meticulously tracks the actions of health personnel, entire work processes become ‘datafied’ and rendered visible, comparable and primed for evaluation. Infrastructures for quantified knowledge production about ‘quality’ are already established (Tøndel and Åm, 2023), which HP connects to. Thus, the datafication moulds their materialities and meanings, through a process which alters the very notion of quality in healthcare (ibid). This includes changes in ‘the metrological drama’ at organizational level (Power, 2004) in the healthcare services.

This article shows how HP has altered the very notion of ‘service quality’ through the introduction of a new vocabulary where ‘quality’ gives way to ‘benefits’ as parameter for care. Benefits realization is not a new phenomenon in the Norwegian public sector (ibid.), but there is a need for research into the nexus between transformation and benefits realization (Isik et al., 2024). Interestingly, the field of health is among the social areas in which the disruptive transformation linked to datafication is most advanced and observable in daily routines as well as in institutions and policies (Diaz-Bone et al., 2020: 315). It is therefore important to explore changes in the relation between a commercially developed system like HP, its vocabulary, and the norms and values of core tasks. The Norwegian welfare state model, and the traditional resistance in Norwegian healthcare services to market-based organizational principles offers a promising research context for such an endeavour, by illustrating contrasts with organizational principles in other health care systems. Our research thus expands on evolving notions of quality within other health data regimes in important ways, contextualizing the shift from ‘quality’ to ‘benefit’ enabled by the HP.

While the notion of quality refers to a standard or a degree of excellence in its own sense, the notion of benefit adds the component of ‘gain’ in relation to specific goals and targets that need to be fulfilled, such as cost reduction or efficiency. It thus introduces a new layer in the services’ value orientation, making ‘benefit realization’ a priority. Data-enabled work processes and vocabularies integrate technical processes with cultural change, shaping behaviour and institutional norms, and legitimizing the datafied welfare state through ‘the twin passage points of economic efficiency and good practice’ (Strathern, 2000). The goals that specific benefits should realize differ at times between normative and administrative policies and the realities on the work floor, and these changes in the practical doing and formal meaning of quality might even widen the gap between these two. Research on the dynamic of separation of data/audit/procedures from work processes (see, e.g., Hull, 2012; Isik et al., 2024; Strathern, 2000) are indebted to Meyer and Rowan's (1977) seminal work on decoupling – ways in which organizations create and sustain gaps between formal policies (and technology inscriptions) and practice to reconcile external legitimacy with internal flexibility (Isik et al., 2024).

HP's promises echo discourses around the European Health Data Space (EHDS), the proposed EU-wide technical-legal infrastructure and governance model for health data circulation (on EHDS, see, e.g., Hussein et al., 2022). The latter inspired Hoeyer et al.'s (2024) recent conceptualization of ‘data spaces’ and cautioning against decision-makers’ tendency to approach health data integration as a merely neutral technical–legal exercise. The authors instead describe data spaces as ‘sociotechnical enactments’ (ibid: 10) constituted by, as much as they are constituting for human activities, their experiences and understandings (ibid: 4).

Hoeyer et al. see data spaces as comprised of formative and experiential dimensions that influence how such spaces are enacted and inhabited. Formative dimensions that bring data spaces into being are promises, work, and users, and relationalities between these, which provide generative sources of direction and mobilization. Experiential dimensions influence how data spaces are felt, judged, and interacted with by people, as some combination of what is deemed to be right, true, present and valuable. Importantly then, Hoeyer et al. stress the need to understand dynamics of expectations and direction-setting (and their legacies) as part of evaluations of how data spaces are experienced and interacted with in practice. Their proposed formative and experiential dimensions offer heuristic guidance for analysts rather than a fixed theoretical structure, and tensions and contradictions may arise between dimensions. Broadly, the dynamic envisioned is one where formative dimensions (promises, work, users) are building blocks for programmes of change, standing in a dialectical relationship with ways in which data spaces may be inhabited and continuously assessed against normative registers (right, true, present, valuable). Here, promises are expressions of ‘expectations, potential, commitment, or anticipation’ (2024: 4). In the case of EHDS, they are optimistically framed ‘desires and ambitions’ (ibid), manifesting the European Commission's intentions for citizen empowerment and for establishing authority over data, and harmonizing markets for digital health products.

The current article offers a different reading of the dynamic between formative and experiential dimensions, by exploring the role of constraint vis-à-vis promise formulation. We contribute to the conceptualization of data spaces via the case of the HP, to show how data spaces are (1) constructed through both expectations/promises and actual material and social conditions while also, once enacted (2) re-constitute the realities and understandings of its constituents by changing not only the vocabulary but also the meaning of what quality is in healthcare (from quality to benefits). We thus enrich Hoeyer et al.'s (2024) implicitly optimistic view of promissory dynamics, by also accounting for promises as grounded in situations of constraint.

Our approach seeks to connects insights on health datafication and data spaces, to current tendencies towards increasing monopolization and dominance by a few key actors in the tech sector (Khanal et al., 2025; Leclercq-Vandelannoitte and Bertin, 2024). These dynamics limit choice and impose constraints on the ability to imagine or realize alternatives even among major software purchasers in the public sector. The creator of HP, US-based Epic, is among the 5 leading EHR (electronic health records) providers and the most adopted EHR system across the globe (Chistie et al., 2023). Although Nordic countries constitute large buyers of new systems and thus presumably also capable of wielding significant influence over platform developers, they nevertheless appear to face major challengers in acquiring data management solutions that align with deeply held and institutionalized expectations around health data infrastructures (cf. Hoeyer et al., 2024: 7). This has been demonstrated in prior research into Nordic experiences with Epic's EHR system implementation in Denmark (Winkler et al., 2020) and Finland (Hertzum et al., 2022) and is further reinforced by how Norway is now following suit despite those well-known histories, as shown by Åm et al. (2025) and the current article. What this highlights is the importance of understanding the role of constraints in connection with system changes, and the role of constraints in the dynamics between aspirational discourses and changed practice.

In our investigation, we consider the interplay of formative and experiential dimensions of data spaces to understand changes in understandings of ‘quality’ within Norwegian public healthcare services alongside the introduction of ICT project competences and vocabularies in relation to HP. Tracking the language of ‘benefits’, one quickly stumbles across issues of traditional ways of patient data reporting vs new, increasingly datafied ones. Metaphors for data collection such as ‘drilling down’ to get information, exemplifies substantial changes in both practices and understanding of quality work in the healthcare services. Combining notions of data spaces and quality our research question is: how might the notion of data spaces help explain evolving notions of quality in healthcare?

Nordic countries are eager to use digitalization and datafication to streamline the provision of welfare (Andreassen et al., 2021). However, there are inherent problems in the datafication of the Nordic welfare sector. Many researchers (e.g., Hoeyer, 2023; Snell et al., 2023; Wadman and Hoeyer, 2018) warn of the dangers and problems inherent in the reliance on data intensification which are often overlooked or downplayed at the start of an ambitious datafication project (cf. Hertzum et al., 2022, see also Røhl and Nielsen, 2019, on the Danish health platform). Against this backdrop, the implementation of HP offers a useful example of a data space enactment in motion.

Importantly, data spaces are concerned with cross-national integration processes in Europe which are not present in the history of HP as planned in use locally (but at research level, data is supposed to become harmonized and usable for analysis across borders (Chistie et al., 2023)). However, the cross-municipal and cross-organizational integration features of the Norwegian case allows the concept of data spaces to shed useful light on its evolution. The data spaces framework acknowledges that any notion of (knowledge-)systemic change conceals complex interplays between expectations, narratives, experiences, frustrations and resistances. Its dual focus on formative and experiential dimensions helps to organize the different elements that make up the unfinished story of the evolving prospects and reactions to HP. As we return to in our discussion, the origin of HP is in one sense rooted in a situation of previous systems’ obsolescence and shortcomings, which later morphed into promissory narratives around HP's potential to solve future challenges in the healthcare system and to ‘automatically’ improve the quality of the services. Accordingly, we argue that further conceptualization of health spaces would do well to acknowledge that such origins and demarcations from previous systems may not simply precede more lofty data space promises (perceptions and practices) but also serve as a continual and evolving source of examples of evidenced mismatches between system ambitions and practice constraints.

After outlining our methods, we will describe the creation of HP as a data space co-constitutive of human activities and understandings. We analyse HP's multiple roles, as a lofty promise offering more efficient, professional, service-oriented and user-friendly healthcare; a feature of daily work realities where extensive datafication widens the gap between the different understandings of quality (work floor vs system); and an avenue for the introduction of a new metrics in quality language and thinking within the many enrolled services and their management (users).

Materials and methods

This article is part of a four-year research project on quality management within Norwegian municipal healthcare services: Measuring quality (MASQ): Exceeding the limitations of quality management in municipal health and care services. The project sought to analyse the co-evolution of performance and quality indicator systems and Norwegian national-level efforts to codify these. To that end, the team conducted multi-level studies into definitions of quality and their relationship with reporting and quantification regimes in Norwegian healthcare institutions. Although the project did not have an explicit focus on HP from the beginning, the authors designed a series of supplementary interviews informing the case study reported on here, in response to concerns raised in stakeholder interviews and dialogue events.

To study the implementation and consequences of HP we conducted two conversations with hospital personnel and 13 semi-structured interviews with fifteen informants, mostly municipal employees from IT and/or healthcare, whose professional roles had been altered or redefined as part of the HP introduction process. We interviewed 4 regional and 3 municipal ‘implementation managers’ who were responsible for the implementation of HP in their region or municipality and thus served as mediators between municipalities and HP developers. We also interviewed 5 local-level and 1 top-level ‘benefit coordinators’ tasked with following up on development status in the work to achieve quality/benefit targets and coordinate across municipal service areas and institutions, and who were generally employed by municipalities. Finally, we interviewed two ‘super users’ who also worked at the level of participating municipalities and were offered early and thorough training in the use of HP; these actors had been tasked with motivating and supporting other colleagues locally. We chose this sample because we were interested in the wider picture of quality measurement and datafication of the healthcare services rather than individual experiences with the new system, and importantly: we were empirically interested in how the platform could lead to a standardization of what in Norwegian was referred to as gevinstrealisering (a term roughly equivalent to the already introduced term benefit-realization, which signifies the conversion from a potential to an actual benefit). Although such vocabulary was used sporadically in Norwegian municipal health and care service evaluation around innovation and ICT projects even before HP (see, e.g., Tøndel and Åm, 2023), its usage appeared to intensify more recently. Accordingly, interviewees were selected for their central role in implementing HP in specific municipalities and/or regions and for their benefit realization mandate.

A limitation inherent to our approach is that we focused on those responsible for the implementation of HP rather than end-users (although four informants also fell into that category), therefore our article is less capable of answering questions about the experience dimension of the data space. To compensate for this, we supplemented our sample by reviewing previous interviews to identify instances where HP had come up: out of a total of 17 previous interviews with 30 individuals (several were group interviews), we identified 8 interviews with 18 individuals that fell into this category, and these informants’ statements also inform our analysis below. We furthermore used secondary literature on HP ⁷ , and consulted publicly available HP educational materials ⁸ and training presentations. ⁹ This literature was used to enhance our understanding of how HP works and to gather technical-organizational information about for example formal timelines or the use of terminology in educational material and was not part of the thematic analysis process. In addition, we used media articles as anecdotal statements to further explore our argument and inform our choice of direction in the study. Still, further research focused on HP end-users would be useful to better understand experiences on the work floor.

All participants received written information about the study and have given their oral consent (recorded) during the interviews. The interviews were recorded and transcribed verbatim. Our analytic strategy has been an inductive approach to thematic analysis of interview data (Braun & Clarke, 2006; Guest et al., 2012). Thematic analysis as method was used as ‘a method for identifying, analysing and reporting patterns (themes) within data’ (Braun & Clarke, 2006: 79). The main themes we identified were expectations and promises; implementation process; change in language; changes in work processes; benefit indicators (BIs) and change in understanding of quality. Our article is structured around these themes and around Hoeyer et al.'s (2024) differentiation between formative and experiential dimensions of data spaces, which are presented in the next two sections.

Great expectations, lofty promises and constraints on the ground – the formative dimensions of HP

Data spaces are constructed through the formative dimensions of promises, work and users, but these in turn manifest in response to material and social preconditions, such as legacy systems and practices. In this section, we describe how the implementation of HP responded to technical and social needs, later articulated as lofty promises about extended datafication of healthcare services into one integrated data space. Hoeyer et al.'s (2024) formative dimensions illustrate how data spaces come into being by reconfiguring promises, work, and users. In this section we ask: what roles did these elements play in the early stages of HP?

Technological transitions are not linear but are the result of decisions and interactions among governments, administrators, firms, technologists and other actors (Borup et al., 2006). These actors decide what is ‘obsolete’ and what is ‘necessary’ for the future (Van Lente and Rip, 1998: 216) and enrol others into corresponding transition processes. In this way, expectations and technological promises also take on performative functions (Van Lente, 2012). The promise of making healthcare better and more efficient lies at the heart of data politics and digitalization (Hoeyer, 2023). Expectations about what will occur in the future become arguments for investing in the present (Petersen, 2019), as is clearly seen in the HP case.

Worries about health care futures predate HP and were at the core of the Norwegian white paper One Citizen- One Journal (Meld.St.9, 2012–2013), which describes many of the challenges that HP would eventually be seen as a response to. The white paper warned that the ICT systems currently used by healthcare personnel would prove insufficient in the face of expected demographic challenges. In response, the Norwegian Directorate for Digital Health (Direktoratet for e-helse, 2016) recommended the creation of a single digital health data record system. It was decided that when the time came for hospitals in central Norway to update their health data systems, they would need to do so in partnership with the municipality, enabling the merging of health data into a single system. ¹⁰ Thus, as the region's largest hospital began preparing for the ‘end-of-life’ of their digital data recording system Profil in 2026 they had to enrol the municipality into the system-building process:

The idea behind the whole thing was the hospital, which needed a new system. And of course there was the one citizen-one journal idea, right? This came from the government; it came from above […] and so the hospital invited the municipality to a dialogue. To have one system: GP – municipal (healthcare)- hospital. One journal, one system, that was the way of thinking. (30.04.2024)

Looking back, several informants recounted to us how the system at that time could not meet demands for reports to higher organizational levels. Instead, this work had to be done either manually, by middle managers using improvised excel sheets, or expensively, by hired consultants. As these data recording practices were ill-suited for growing quantities of shared health data, it became attractive to look to platform organization recently deemed a ‘new ideal of data integration’ (Hoeyer, 2023): the Sundhetsplatform (Health Platform), a shared, clinical and administrative IT solution introduced in Denmark in 2016/17 (Røhl and Nielsen, 2019).

The Sundhetsplatform had been developed and delivered by the American IT supplier, Epic. Expectations had been high in Denmark, but results were highly contested by users, media and by researchers and described as ‘the American limousine that was not made for public transportation’ (Hoeyer, 2023: 100). Despite the adaptation difficulties of its system in a Scandinavian welfare state setting, Epic was given the opportunity to try again in Norway. According to our informants this happened because Epic was ‘the only candidate’ qualified to deliver on the public tender. In a world dominated by tech-giants (Khanal et al., 2025), it seemed impossible to find an alternative solution tailored towards the needs of the Nordic welfare state. Midt-Norge (central Norway) ended up choosing the same supplier who had delivered doubtful results in Denmark. Constraint, and lack of real choice, thus appears to have been a critical factor at this early stage.

One might reasonably speculate about whether any hesitation or doubt prompted by experiences from Denmark, gradually became concealed by promissory and aspirational discourses, which served as necessarily credible motivations to secure public license to take any action on system replacements at all. At any rate, to ground the lofty promises surrounding the implementation of a digital platform system within the concrete needs of the welfare state the promissory narrative around this large investment were redirected from being about practical matters of constraint and technical need, to address broader societal shifts and their implications for future healthcare.

Among the expectations that were given weight was the view that HP would be necessary in the face of the demographic shift expected over coming decades. Many informants explained the need for a new system with reference to the coming ‘eldrebølge’ (wave of old people): ‘we need to have a new system in place before the wave comes to hit us’. Statements such as this indicate a shift in justifications for the HP, from matters of practical necessity (replace an outdated system) to addressing anticipated future social challenges (raising costs of healthcare for the elderly). As most citizens expect the welfare state to take care of the elderly while public financial and human resources are tight, HP was seen as the solution that would enhance the professionalism, effectiveness, service-orientation and citizen-engagement (Rose et al., 2015) of the strained Norwegian healthcare system.

When seen as not just a solution for replacing obsolete systems, but also as a strategy for coping with future social and demographic challenges, the broad scope and high investments of the HP were seen as justified by local policymakers and politicians. Over a short period more than half of the municipalities ¹¹ and some of the largest hospitals in the region signed contracts with HP to go live before 2026. By November 2024, 34 municipalities had started using HP, which at that point had integrated 220 previous data systems into a single data space used by more than 33,000 people. ¹² However, delivering on lofty promises meant that real work was needed, not just by Epic but also by the main users of the system: the municipal healthcare workers and the nurses and doctors in the hospitals. According to the ‘law of medical information’ (Berg and Goorman, 1999) substantial work is required to make information usable in different contexts beyond its production, and HP is no exception. Our informants described that the formative building and implementation process took much longer than planned, and the transfer of data from the old to the new system required great efforts since old data from visits and consultations could not be imported automatically.

Once HP became operational, the misalignments between its promises and its altered conditions for work and users began to manifest:

We can also be honest that we had much too high expectations of what HP could solve in this half a year in operation. […] We see that it is far too complicated, and it is too big. It is a bit of a shame that we were unable to see it in advance in relation to the expectations of employees. […] We thought that when we get the HP, everything will be resolved, and then everything will be fine. Starting to think back, there was quite a lot that didn't work before and, but we thought that now it will be fine, sort of. Yes, there we went wrong. (9 April 2024)

Informants shared that healthcare workers struggled with the new interface and with the new language and work procedures that HP introduced. This resulted in backlogs for patient consultations and administration, particularly at the region's largest hospital. These problems prompted not only much of the negative media attention and protest marches mentioned at the beginning of this article but also presented challenges for the quality improvement that HP had been intended to provide. During our research it became clear that most municipalities had not yet been able to work on quality measurement, instead using the system solely for day-to-day data recording without reviewing entered data or utilizing the system's new analytic capabilities. ‘We bought a Rolls Royce, but we are using it as a Lada’, was a phrase used regularly (echoing the description from Denmark of HP as an American limousine) ¹³ , acknowledging that only a small proportion of the opportunities of the system were being used. The quote aptly illustrates not just a view of HP as being extravagant in its functionality, but also of an institutional and cultural mismatch between the insurance-based healthcare regime in the US, where HP originated and facilitating billing between insurance companies and care providers, and the single-payer regime of Nordic welfare states, where the system was being implemented. ‘Daily operation trumps everything’ and ‘when you are swimming for survival you are not thinking about the colour of the curtains’ was how informants described the state of data analysis and quality measurement in their municipality.

This shows that data spaces are formed and operate in both digital and physical territories where data promises cannot be realized without data work of data users (Hoeyer et al., 2024). When looking at the formation of HP it becomes obvious that aside from the nationally defined objective of harmonizing regional/municipal health records into One citizen – One journal, the starting-point for HP was not hopeful promises formulated in a vacuum (as implied by Hoeyer et al.'s critique of the EHDS), but also rooted in a situation of constraints, system obsolescence, and limited available options. Even fear played a role, insofar as the anticipated demographic wave of the elderly that might overpower Norway's health information infrastructures, contributed to the framing of promissory narratives around HP.

Alongside HP's adoption, the system's promissory language became imbricated across institutional structures and voices. In the next section, we will show how the introduction of new metrics and language for quality influenced not only work processes but also the understanding of what quality means within the healthcare system – reminding us that quality is primarily a policy term.

Just a change of vocabulary? How HP is experienced in practice

Having outlined the formative aspects of HP, we now turn to the experiential dimensions of data spaces and how they re-constitute the realities and understandings of those who operate in it. Specifically, we investigate how system shifts influence professional language, practice and thinking, as evidenced in informants’ understanding of quality.

From the beginning, HP was accompanied by an imaginary of datafication as rendering quality both measurable, comparable and accountable, and in turn as enabling effective and professional services tailored to rising demands. As one informant described it: ‘we have to switch to HP because we must work better and smarter, we will be grilled and checked for better quality, next of kin are pickier, patients are more demanding. You cannot operate like you did ten years ago’. To adapt work to a new era (the makers of) HP introduced new metrics and metaphors which not only influenced everyday language, but which also intended to change how quality was pursued and practiced more measurable (‘slicer dicer’ and ‘dash board’), open to accountability (‘drilling down’ and ‘breaking glass’) and comparable (‘benchmarking’). In essence, we describe this as indicating a shift in how quality in healthcare is defined and experienced, moving from quality to benefit. This shift has anticipated consequences for healthcare services, patients and professionals in terms of quality being a driver for organizational and political change (Swinglehurst et al., 2014).

Slicing and dicing at the dashboard – making quality measurable and effective

At the beginning, HP was ineffective. All municipalities told us that it took longer time than expected to learn the system, and that recording data was more time-consuming than in previous systems. This was partly due to the extensive changes in work processes and the new recording procedures introduced by HP. Work procedures were sliced and diced into countable tasks, acts and units which made work more measurable. Small work units were then recorded within the system so they could be taken out and analysed via a tool called SlicerDicer, a:

reporting tool that provides physicians, department managers, and other users with intuitive and customizable data exploration abilities. […] SlicerDicer can show results using a variety of different measures, including totals, percentages, averages, variance, maximums, and minimums ¹⁴

An accompanying system of codes was intended to make the work not only more measurable but also more efficient, allowing employees simply to tick off tasks and actions from a list. Not everyone appreciated this new way of entering information, and many hospital staff dismissed it as expressions of what Almklov and Antonsen (2019) has termed the ‘tyranny of the drop-down menu’, requiring many steps and leading to a potential loss of information as previous free-text options for entering patient information were removed. Yet, what might seem as ‘meaningless data at the ward level (often) makes sense at the administrative level’ (Hoeyer, 2023: 106). For many informants tasked with collecting and measuring data, the replacement of narratives with codes made sense. They told us that previous long texts had ‘zero searchability’; and had forced analysts to ‘read a whole lot of paper’ which ‘is probably great for [those entering the information], but little use for others’; and that one has to ‘stop with that silly prose about how we ate our slice of bread. There is absolutely no obligation to provide documentation related to that’. We were told that one rather needs ‘structure and codes to find out how many home-visits, how long they last, and who did it’. In the logic of measurable and quantifiable work processes, then, HP delivers a more efficient system, and the contrasting informant responses above suggest diverging views on the experiential dimensions of what is present and valuable. The slicing and dicing of information into small, comparable and measurable units made the data more present and valuable in the practical conduct of administrative workers and management while detaching it from the people on the work floor who preferred to have more personalized text. The broader question for us is what the broader implications of these divergences are when it comes to care for patients.

The phrase slicing and dicing aptly describes a shift from qualitative to quantitative logics: work processes and patient narratives are translated into measurable and comparable units. The results of one's work can be found on the dashboard, where information is made present in terms of Hoeyer et al.'s experiential dimensions. According to informants, interface use is still in its infancy; information is entered, but not (yet) taken out by employees and superiors to reflect on work processes. We learned that HP is expected to enable to drill down to extract data to steer the course towards more effective services, but in the current ‘survival mode’ of the implementation phase not many unit leaders have done so.

Drilling down and breaking glass – making quality transparent and accountable

One of the main changes that HP introduced was an expansion in scope from (mostly) administrative patient and health data to organizational and health personnel's work processes, all of which were rendered visible outside the individual units. Before HP was introduced, each unit of the municipal health care system, hospital, or laboratory would use their own data recording system, which could be either digital, paper-based, or a combination of these. HP offered a way to make all data visible and transferable between the healthcare units across the region. General practitioners are not currently part of HP, but even without the GPs the amount of data that is shared is far greater than before.

As large quantities of health care data are shared between different services/units and service/organization levels, there is a potential for greater transparency, as patient and treatment information can be accessed by data analysts and healthcare personnel beyond the individual practitioner. This potential raises new concerns: ‘Not everybody likes that, that they can see so much you know’, one informant told us. Especially in the psychiatric services there seemed to exist a certain reluctance to ‘know too much’ and a concern for privacy of the patients, reflecting unease with how this new data space may be inhabited, and with changing demarcations around the experiential dimensions of what is right (cf. Hoeyer et al., 2024) with respect to the recording of health data in new systems where shareability and access differs from previous data recording practices. These concerns echo the concerns about EDHS where notions of what is ‘legal’ and what is ‘right’ might differ between lawmakers and professionals in the healthcare sector (Hoeyer et al., 2024: 9). Professionals might have concerns about patients’ anonymity and privacy (‘know too much’) even though from a legal and administrative perspective it might be right or even necessary to have these insights across the data platform.

Another new terminology introduced by HP is drilling down: depending on the hierarchical level one can drill down to get information from any level below one's own. The unit leader has access to all the activities within the unit, the department manager sees all that happens in the whole department, while the municipality director and others with appropriate authorization can access all data in the system. Our informants stressed that drilling down would only be used to gather data about units for enhancement of quality and effectivization of work processes, and that it would not be used for monitoring employees. Even though our informants were not concerned that HP would lead to a greater surveillance of their work, in expressing a strong focus on hierarchical structures the term drilling down also reflects power differences between those who can drill down and those who cannot. The enhanced visibility of employee's actions and the embedding of hierarchy into everyday work language does not only lead to greater accountability but conceivably also to more control, or the fear of such.

Drilling down is an integrated part of the system, but what if one needs information from ‘higher up’ or beyond one's own unit of responsibility? Then one needs to break glass as our informants called it:

The access is controlled. […] You can see information from other departments. But there is some security in the form of ‘breaking glass’. You must justify why you want to see this information, and this will be logged in the system. You have to ‘break into the journal’, and you may need to. But you have to justify it. (15 April 2024)

Drilling down and breaking glass are thus more than new terms within a digital system. They are promises of reconstituted practices, envisioned as leading to greater transparency and accountability within healthcare to benefit the quality of services for patients while also generating more knowledge. The actions are suggestive of new ways of identifying and enhancing that which is valuable and true (cf. Hoeyer et al., 2024), such as quality, as a meaningful practice alongside and isolatable from direct employee oversight. However, these expressions also inscribe hierarchical rules concerning access and oversight onto everyday work processes.

Benchmarking – standardization and making quality comparable

One of the great expectations for HP has been that increased transparency and data sharing would enable greater comparability between the different municipalities through benchmarking; another practice associated with the mapping of quality and its enhancement. Harmonizing data input practices could enable new comparisons that were not possible under municipalities’ previous data systems and practices:

You can compare much more easily. Compare the same units with each other. Of course there are differences, different users, different logistics and different houses […] but everyone learns from each other. And it is also a goal – being able to compare and learn between the municipalities […]. (30 April 2024)

Our informants generally saw benchmarking as enabling learning rather than as problematic: ‘If they do it better in municipality x we can learn from them’. However, one informant was concerned that there is a general ‘copy-paste’ attitude among Norwegian municipalities who tend to copy something that has worked (economically) well elsewhere without much concern for local differences, such as differences in demography and resources. HP could exacerbate this tendency.

Hoeyer et al. (2024) warn against a (political) inclination to see data spaces as standardized areas enabling governance via data, where people work according to ‘interoperable formats’ (ibid: 9). In some instances HP's proposed one-size fits all processes have the potential to deliver less quality than previous solutions: One of the factors both employees and municipalities will be measured on via the HP is case-management-time, or the time it takes from when an application is made for a particular service (for instance, home-care services) and when that service is administered. The goal is to reduce this time to a minimum. One smaller municipality told us that while this factor might be important in larger municipalities who have many cases and applications, the local procedure has been different – they used to send a home-care nurse immediately upon application and worried about the administrative paperwork later. This very service oriented ‘quasi solution’, as they called it, is no longer possible in a standardized system where the right work procedures must be ticked off. This example shows that one-size-fits all solutions and standardized and automated work processes sometimes constrain rather than enhance the quality and effectiveness of healthcare services, as such it shows how HP could go hand in hand with processes of centralization. Clearly, the story of how quality co-evolves with health system digitalization perturbs valuable local practices and accelerates the widening of the gap of what is relevant (present in Hoeyers terms, 2024) and valuable in the work processes for those providing and those measuring the care given.

Healthcare datafication: From quality to benefits

The promissory dynamics and uptake of HP can be viewed as a case of a shift from ‘quality’ to ‘benefit’ in the context of healthcare datafication. Against this backdrop we suggest that Hoeyer et al.'s (2024) call for a new vocabulary for a data-intensive health environment, should be attentive to elements of constraint, evident in our case but easily overlooked in studies of lofty technology promises. We have approached our case in the context of a broader tendency in several Nordic countries to adopt (and adapt to) platform solutions (see Røhl and Nielsen, 2019; Winkler et al., 2020 and Hoeyer, 2023 for Denmark and Hertzum et al., 2022 for Finland) that in many ways are ill suited to single-payer welfare state healthcare provision. In the light of this, the current analysis of HP's trajectory in Norway illustrates the conceptual need to cope with constraint as a feature of promise-making, both via the aspirational discourses it is couched in (formative dimensions), the normative registers against which it is assessed (experiential dimensions), and the ways that this dynamic becomes reflected in changed practice ontologies (quality).

The datafication and digitalization which HP exemplifies, offers to make the esoteric notion of quality in healthcare more measurable, accountable and comparable. The current vocabulary and definition of quality and the idea that ‘quality is operationalized into quantitative measures’ has long historical roots (e.g., Tøndel and Åm, 2023), as the old phrase goes: ‘What counts is what can be counted’. Yet, quantification also shapes the terminology and understanding of quality (Tøssebro, 2022). When describing quality (‘kvalitet’ in Norwegian) our informants tended to reframe the phenomenon to one of ‘gevinst’ (indicating benefit, or gain). This change in wording is widely used across HP's educational and tutorial materials and has been taken up by actors working in the municipalities we studied. When asked to explain the difference between quality and benefit, informants gave similar answers: benefit is more than quality, it replaces and includes certain elements of quality but also emphasizes economic benefits (and thus excludes other elements). This change in terminology implied a more overt emphasis on cost-reduction through effectiveness as integral to quality, even if savings are not necessarily realized in practice:

Municipality X said that here (with HP) we save 40 million a year. But I don't deliver kroner and øre [money]. ¹⁵ I deliver quality and time. Those are the two things I can deliver to you. And then there is a dynamic in the healthcare services that means that if you get time in one place, you use it in another. […] we spend less time on those who do not attend, ergo we treat more patients, but we cannot be less staff, we just treat more patients. (17 April 2024)

At the time of writing, eight BIs have been established within HP from which municipalities could choose a smaller sample for starting to measure quality, effectiveness and user involvement of their local healthcare services. ¹⁶ They include continuity of care, measured by the number of employees who has visited with a patient in home care over a certain time period. Another BI is the degree of alignment between medicines prescribed by general practitioners and the medications that have actually been administered in home care, together with deviations from the prescribed schedule for administrating medicine. Other measures are the number of no-shows at doctors’ appointments, the time delay between referrals and medical treatments, time spent by clinicians, and the number of individuals who have actively downloaded and used ‘HelsaMi’ (the citizen/end-user app that accompanies HP) to plan their medical appointments.

For all its lofty ambitions and promises, the current ability of HP to capture and offer improvements to quality appear somewhat modest and rudimentary. Therefore, the question remains of whether HP can meet its funders and users’ expectations for major improvements in quality and effectiveness. The gap between ambitions and promissory formative language around HP, and the modest, quantifiable indicators current serving as proxies for ‘quality’, is large and noticeable.

Organizational constraints, similarly, played a crucial role at one stage of the HP trajectory; a hospital patient records system was becoming obsolete, and only a single alternative data product existed which met the basic specifications and needs for a data management platform at the necessary scale. Amid lofty promises and ambitions for limitless analytical potential then, practical constraints and limited options also played important roles not just in motivating programmes of change, but also – by setting up the contrast between stated promissory ambitions and limitations experienced in practice – in evolving narratives of discontent and controversy about the system. Thus, while promises and visions are part of the picture as we make sense of health data spaces, Hoeyer et al.'s framework might be usefully extended to more fully take organizational constraint into account as both a background feature on the promissory spectrum, and as a reference-point against which promissory ambitions are continuously assessed by users and publics.

Yet, notions of quality are shifting, and they are doing so with reference to the promissory rather than constraint end of the spectrum of datafication possibilities. The element of (prospects for) quality imposed via HP, seems effective at reframing users’ practices and imaginations. Politicians, bureaucrats, health care providers and patients all agree that high-quality healthcare is a priority for the future, however, prioritization of benefits vary (low-cost vs. full-service) and so do conceptualizations of what constitutes quality. HP's focus on BIs that make quality measurable, accountable, and comparable supports a widening of the gap between quality as understood on the work floor (through human interaction) and quality at the system level (through data). Here the measurement of quality becomes an actor in its own right, steering work processes and employees’ actions towards effectiveness and (economic) benefits. However, for many of HP's users, quality is still related to the work floor:

There is not much talking (about benefits). I know that the managers talk about benefit, but we haven't gotten to the point yet that we use it in our department. We don't necessarily talk about benefits (as a concept), but as something we see. (24 April 2024)

I do not feel that I am dealing with benefits. It's more the management, maybe. I work a lot with quality. That what is documented is of good quality. (29 April 2024)

Our data suggest that the implementation of HP has accelerated a divergence between different concepts of quality within healthcare widening the gap between the administrative and management level and the work floor of municipal and clinical health care services. This resonates with the observation of Tøssebro (2019) that there is a close relationship between the understanding of quality in healthcare organization and policy and governance transformation. Several informants point to the responsibility placed on unit managers for on the one hand interpreting standardized codes contextually and appropriately, and on the other hand motivating employees to make proper use of these codes in their practice. The protests described at the beginning of this article were a result of these changes in conceptualization of quality work and that what is meaningful on the administrative level might not be so on the ward level.

While most of our informants on the administrative level were satisfied with the implementation of HP and the possibilities it offered for effectivization and data-based enhancement of quality, clinicians were less satisfied and claimed that the ‘meaningless work’ (Hoeyer and Wadmann, 2020) of data input into slicer dicers and dashboards reduced the quality of healthcare work substantially. As the negative media attention and concerns mentioned at the beginning of the article shows, this poses a danger for the future of trust to the welfare state which not only depends on financial but also human resources and public acceptance of the services. If the benefits of enhanced cost and time effectiveness (economic efficiency; Strathern, 2000) through datafication continues to detach itself from the understanding of quality healthcare on the work floor and among the general public (good practice) the legitimacy of public healthcare policy could come under threat. Further research should continue to monitor these divergences and their consequences for healthcare provision and the Nordic welfare state model.

Conclusion and further research: Not all is benefits in datafication

Despite negative press and a long and cumbersome implementation process, expectations for HP remain remarkably high among many of our informants: ‘it is all very difficult now, we cannot measure quality yet, but it will be all fine;’ ‘it will be a success story;’ ‘2024 will be already good, 2025 will be even better;’ ‘I think it will be a better world’. So, despite its contested history, hopeful expectations remain about future improvements brought on with and via more and better data. We have sought to connect these aspirations to broader changes in how public healthcare in Norway and its Nordic neighbours co-evolve with the actually available options for quality measurement and data management offered by commercial actors such as Epic.

HP emerged not just from promises, but also from a place of necessity (the impending obsolescence of a management system), or even from a place of worry or fear (such as worries that the coming demographic shift would overpower today's information systems). Thus, our story both supports Hoeyer et al.’s (2024) view of data spaces as more-than-neutral, -technical, -standardized or -static environments, while also suggesting further directions for their data spaces framework. With respect to notions of quality, HP in its current form offers a broad-brushed phenomenology, effectively reducing quality to eight quantifiable features. The mismatch between promissory language and piloted realities is notable. We propose a broadening of Hoeyer et al.'s notion of promises in relation to data spaces, to acknowledge how starting points of constraint, need, and limited choice can cascade into promissory territory. Elaborating on this dynamic would be a useful avenue for further research and conceptualization of data spaces.

While the implementation of HP offered greater accountability and comparability on an administrative level, workfloor-level dissatisfaction remains about the dehumanizing effects of reducing work processes and patient histories into codes and units and the time dedicated to ‘meaningless’ data work. HP feed into a widening gap between the conceptualization of quality among administrators, technocrats and politicians on the one hand, and the notions of quality held by health workers. Similarly, notions of the benefits of EHR-enabled changes within the healthcare system also differ; making work more effective does not necessarily mean more cost-effective since the hours will be used elsewhere in the sector. The expectations for HP's potential in solving both technical and societal problems were high and while HP might not yet have delivered what was expected in terms of effectivization or quality management, politicians, technocrats and administrators still maintain that ‘one day in the future’, it will be better as the belief in data-intensive healthcare keeps being firm. Among some users, faith in HP as ‘the right tool’ is fragmenting, but the promise of improvement remains just as strong. One question raised by our research, is whether this aspiration reflects genuine optimism, or curtails severe constraints. The high degree of confidence in future improvement may for instance reflect citizens’ high trust of the welfare state in the Nordics (e.g., Hoeyer et al., 2024), rather than confidence in the technology itself.

Norway's HP experience is a reminder of how, for health data spaces, one-size does not fit all. Perhaps instead of big data dreams, a turn in data valuation towards ‘small[er] data mindsets’ (Snell and Tarkkala, 2024) might better align with Nordic welfare state contexts. Here, ‘successful’ benefit-realization (in accordance with its technical scripts) relies on a decrease in practice variation (Isik et al., 2024, our italics). This is a difficult prerequisite for a complex public healthcare region that relies on variance to fulfil its mission. While Hoeyer et al. (2024) advocate for the importance of more socially robust data infrastructures, tensions between promises, beliefs and practices may also have certain benefits, keeping means separate from ends that we may be better off without. During times of institutional change, such separation may inspire new ideas about what Norwegian healthcare is and what it should be. Ontological changes around core responsibilities such as service quality extend beyond language and into the very fabric of the welfare state and its orientation. Perhaps Berg and Goorman's law of medical information could be amended to expanded with the following caution: in order to uphold service quality, a degree of information incompatibility should be maintained along the healthcare production chain.