This research ascertained the characteristics of the work of Australia’s Health Information Managers (HIMs). It also explored the challenges (disorder) that they face and how they manage these (create order) to produce the official facts (‘the truth’) of the patient. A qualitative approach involved in-depth interviews of 55 HIMs. The HIM-participants, employed throughout the healthcare system, had spent between seven and 40 years in the profession and were confident of their expertise. The domains of their work and professional knowledge were identified as: health information science and management, including data privacy; health classification; health informatics, and information and communication technologies; and health data analysis and epidemiological research. The HIMs’ articulation work involves complex, system- and data-based governance, data privacy, and technology-related functions intended to purify and establish the source of ‘truth’ of health data for their ultimate construction of official facts. In curating, classifying and analysing the scientific facts of the patient they produce population health data and calculate hospital performance and revenues. Their ordering work necessitates constant problem-solving, intermediation, and interfacing between stakeholders, technologies and technicians amid systemic pressures and expectations of their outputs. Paradoxically, their data ordering and truth-making are invisible to patients and the public.
In the last decades, nursing homes in Denmark, have become increasingly data-oriented, with the expectation that more digital data enable more person-centred and effective care practices. On basis of ethnographic fieldwork at three nursing homes in the capital of Denmark, we explore how data about care is produced and travel within the nursing homes and become powerful claims of evidence. We illustrate how nursing home data can be considered as socio-material constructions and highlight the professional data work it takes to produce data. Drawing on recent work in Critical Data Studies, and an ‘Ethics in Practice approach’ to professional care work, we show how data work and processes of datafication might interfere with ideals and values associated with what is considered ‘good care’ and demonstrate that information about situated care practices do not always travel well as data. The paper concludes by addressing the risk, that the partial nature of nursing home data, might render aspects of care invisible and hereby undermine the situated and creative care practices, that are necessary if we are to enable true person- centred care.
As information systems (IS), including data-driven solutions, become increasingly prevalent across sectors, the need for ethical data-driven decision-making (DDDM) is gaining recognition. This scoping literature review explores how agency and ethics are addressed in scientific discussions on DDDM. It examines the ethical terminology used, its definitions, and the forms of agency identified in 79 peer-reviewed articles from disciplines such as computer science, education, and engineering. Agency is central to ethical judgment, decision-making, and action. The review reveals a wide range of ethics-related concepts, many of which are loosely defined, incoherent, and problem-oriented. Drawing on a sociomaterial perspective, the study identifies three forms of agency relevant to DDDM: human, technological, and distributed. These forms reflect the complex interplay between individuals, technologies, and institutional contexts. The review concludes that ethics and agency must be considered together to clarify responsibility at different levels of decision-making. Although ethical concerns are increasingly discussed, the debate remains fragmented. In the broader context of IS and the rapid development of generative AI, cross-disciplinary research is essential to address ethical challenges comprehensively. This study contributes both practical insights and theoretical understanding to the evolving discourse on ethical DDDM.
Healthcare organizations strive to become increasingly data-driven. Achieving this vision requires significant efforts to make data accessible, intelligible, and integrated into healthcare practices. In this paper, we explore data work as infrastructuring, focusing on the ongoing practices required to make and maintain data infrastructures in municipal healthcare. Based on an interpretive case study of a Scandinavian health agency, we identify four distinct data infrastructuring work practices: accessing, structuring, reconfiguring, and visualizing data. Our findings show how these practices enable a sociotechnical network of health data embedded in local practices. Through this study, we make three contributions to the research on data work. First, our findings expand the empirical knowledge on data work in healthcare by showing data work in the backrooms of healthcare where data infrastructures are set up and developed. Second, we contribute to the conceptualization of data work by framing it as infrastructuring, meaning data work aimed at enabling a sociotechnical network of health data embedded in local practices. Third, we elaborate on this conceptualization by showing how it involves efforts at transforming data from objects of work to embedded enablers of work.
This paper explores current themes in research on healthcare data work and practices. The concept of ‘data work’ has received increased attention in connection with the datafication of healthcare through various clinical and administrative IT systems and subsequent strategies to make the sector data-driven. However, data is not a given or ready to be applied, but must be created, shaped, packaged, transported, interpreted, and put into action. The datafication of healthcare therefore implies that more people must work with data. In this paper, we provide an overview of the main themes and findings in the current state of the literature on data work in healthcare. We identified 90 papers through a scoping review and conducted a thematic analysis which identified our main themes in the literature: collaboration, power and politics, changes in tasks, roles and responsibilities, and data challenges. We present these four themes, and subsequently discuss our findings in relation to standards and classification, tensions in data work, and the future of work, identifying gaps in the literature and pointing to possible future directions for research.
The increased adoption of platform-based services in healthcare warrants new scrutiny of bias in patient care. Emerging data-driven electronic health records allow for future data extraction and eased access to data streams for purposes other than recording individual patient care, raising concerns about risks of bias and discrimination. This paper investigates nurses’ de-biasing practices in their daily data work. Our understanding of bias applies to nurses’ relational practice and perspective on data work, wherein de-biasing entails attention to sociodemographic factors. We ask: What do nurses consider as relevant data to document as part of their work caring for patients, and how do they account for risks of bias? We investigate this question ethnographically in two Danish hospital wards, via 67 total observation hours. We find that nurses may approach de-biasing the patient record with a tactic of omitting data related to sociodemographic factors (e.g., class, education, or cultural background), which could provide advantage or disadvantage for the patient’s treatment. As EHRs are increasingly designed for cross-patient data extraction and new forms of decision support, nurses’ de-biasing is important to consider in relation to the risks of discriminatory patient care.
Observability receives scant attention in Information Systems (IS) research, appearing primarily within the framework of Critical Realism (CR). CR draws a sharp line between what can be observed and what remains beyond observation. Its stratified ontology treats observed empirical events as mere effects of generative mechanisms that reside in the deepest, unobservable ‘real’ domain. A closer look reveals that CR holds two parallel views of its ‘real’ domain: (1) a fallible scientific ontology derived retroductively from empirical inquiry, and (2) a philosophical ontology presented as infallible, universal, and timeless. This research note contends that (a) the justification for CR’s philosophical ontology rests on an outdated conception of philosophy as an armchair discipline that provides foundations for empirical science, and (b) the motivations underlying this ontology are both political and theistic. Consequently, CR contributes little to our understanding of observation. Observation does not—and should not—depend on CR or any other philosophical foundation. Observation is best achieved when coupled with instrumentation, experimentation, and an explicit acknowledgment that most of our knowledge depends on testimony.
As healthcare is being reshaped by technology, researchers are increasingly relying on patients to generate data. However, the efforts patients dedicate to data collection for research and treatment purposes are seldom scrutinised. We examine the data work of 21 patients diagnosed with schizophrenia or bipolar disorder who participated in our digital mental health intervention, equipped with a smartwatch and a mobile application. We offer in-depth insights into their somatic experiences while living with technology over time by analysing the impact of data work when living with technology. Our paper offers insights into the data work of patients, highlighting its importance as a valuable resource for future digital mental health interventions. We outline what living with technology for individuals with serious mental illnesses entails by unravelling their data work into five themes. Moreover, we provide a model that others can use when incorporating somatic IS artifacts into practice.
Data is a critical resource in healthcare research, yet ensuring high data quality remains a persistent challenge. Errors introduced at the point of data entry can propagate through the data lifecycle, affecting usability, integrity, and research outcomes. In response to this challenge, we designed and developed a digital infrastructure over the course of four years capable of accommodating diverse structured health data while enforcing predefined data standards through a validation system. We use that digital infrastructure to examine the root causes of data quality issues at the input stage by analysing the type of data work needed for meaningful data curation. Through a combination of feasibility evaluation, surveys, and semi-structured interviews, we identified recurring mistakes, user behaviour patterns, and underlying reasons for poor data quality. We contribute six key elements of data work necessary for meaningful decision-making and research purposes, grouped under three categories: (i) the data work needed for reaching intrinsic data quality in data governance, (ii) the data work needed for reaching contextual data quality in data governance, and (iii) the data work needed for reaching representational and accessibility data quality in data governance.
In this Introduction, we provide an overview of the twelve papers that comprise this special issue on data work and information systems in the context of healthcare. This collection of papers presents much needed analyses of the practices through which data is produced. We consider data work a socio-technical process where data is shaped and constructed with care to make them integrous and thus ‘healthy.’ The papers make visible a multitude of people working with data to generate, shape, format, interpret, recontextualize, and visualize data. This is the human labour that underlies datafication processes, Big Data, and Artificial Intelligence. Work that is often not noticed but taken for granted. As these papers show, this skilled work should instead be acknowledged and granted resources if efforts to make use of data in healthcare are to succeed.
Artificial Intelligence (AI) is expected to enhance public services by automating tasks and improving efficiency. However, visions of future AI use often overlook the early stages of its integration. We investigate how public organizations prepare for AI in service innovation shedding light on pre-implementation phase before implementation and use. We ask: How do public organizations prepare for AI integration in service innovation? Information Systems (IS) research recognizes the phases preceding implementation as crucial for shaping technology use. Given AI’s evolving nature—constantly in the making, shaping, and being shaped by socio-organizational processes—questions arise about how organizations navigate uncertain technological futures. Drawing on a two-year qualitative case study examining the planning for AI in Norwegian public services, we find that public organizations engage in sense-making during the initial stages, gradually forming organizing visions that shape how future AI technologies and practices might look. These visions emerge through activities for exploring integration opportunities and possible use cases for AI; through data work for finding the right data and ensuring data quality; addressing legal constraints; and facilitating data sharing. We contribute to IS by further theorizing the crucial early phases in which visions about AI are formed, which can impact the broader integration of AI in public services.
Encouraged by the vast amounts of data produced in healthcare, political ambitions to make healthcare data-driven have led to significant investments in infrastructures. This paper explores how healthcare professionals work with data and infrastructure to pursue data-driven ambitions in regional hospital departments. Drawing on the concept of ‘data journeys’ (Leonelli, 2020), I examine how data are produced, processed, mobilized, and repurposed through Business Intelligence tools and a data warehouse. Based on qualitative interviews and fieldwork, I identify three central forms of data work: reconfiguring data infrastructures, managing data quality in practice, and visualizing and recontextualizing data. My findings show that data journeys are not linear, but cyclical, negotiated, and shaped by infrastructural constraints, professional judgment, and institutional tensions. These tensions, which are conceptualized as breakdowns, ontological ambiguity, and disputes, are not mere disruptions, but significant to how data becomes actionable. Lastly, I introduce the concept of the ‘productive paradox of error’ to capture how error management generates new epistemic and ontological work, and positions tensions as generative forces in healthcare professionals’ pursuit of becoming data-driven.
Multiple actors increasingly work with data across organizational boundaries. While prior research has emphasized formal roles and frameworks, we know little about how tensions around who gets to own, access, and control data are negotiated in practice. In this paper, we advance data diplomacy as a framing to examine how negotiations about data governance are carried out when diverse actors with heterogeneous interests must collaborate. Drawing on a phenomenon-focused interview study with 12 data professionals, we identify how data diplomacy unfolds either ‘at the table’ or in the ‘doorway’, through backchannel talks or with charter agreements. We develop a matrix of four modalities that explain how data professionals adopt relational strategies that vary in timing and formality to navigate tensions from multi-actor complexity in data governance. Our insights on data diplomacy contribute to IS research by showing how data governance frameworks and policies are enacted through negotiation and interpersonal engagement in everyday work.
This speculative position paper critically examines ground truthing in medical AI and challenges the dominant assumption in machine learning (ML) that ground truth is singular and objective. By treating disagreement as ‘noise’ instead of an epistemic signal, current ML pipelines obscure the interpretative complexity of medical expertise, reinforcing epistemic sclerosis—a phenomenon where AI systems calcify classification schema and discourage professional scrutiny. In response, we introduce frictional design to ground truthing as a conceptual framework that enables a deliberative, iterative, and multi-perspective process to the ground truthing workflow and use of AI systems. We first examine the literature on design interventions in ground truthing practices that attempt to leverage expert disagreement and uncertainty, before gathering and extending these interventions within a frictional design framework. We propose a structured set of design interventions, including multi-labelling, deliberative annotation workflows, reflexive documentation, and uncertainty-aware AI decision-support systems, to support more adaptive, accountable, and epistemically robust ML pipelines. As a design paradigm, frictional ground truthing can collect and inspire designs that preserve interpretative plurality, resist epistemic sclerosis, and foster more responsible AI decision-making.
