Between data and touch: how nurses and patients negotiate the human boundary under data-driven surveillance in chronic care Authors: Qianli Zhu, Weizhen Hu, Laite Chen Published: 2026 ABSTRACT BackgroundThe integration of data-driven continuous monitoring into chronic disease management is often presented as a solution for clinical efficiency. However, this technological shift may substantially reshape care relations, power, and intimacy. Prevailing research, focused on technical validation, fails to capture how such systems operate as social forces. A critical, theory-informed examination of these lived transformations is needed. This study goes beyond documenting ethical dilemmas by critically investigating how data-driven surveillance restructures nurse-patient dynamics. We aim to uncover the mechanisms through which data-driven surveillance disciplines both caregivers and recipients, and to identify where and how the irreducible elements of humanistic care persist and resist.MethodsA critical qualitative design was employed, guided by Foucault’s concept of governmentality as a sensitising lens. Semi-structured interviews were conducted with 12 registered nurses and 15 patients with hypertension or type 2 diabetes in one Chinese tertiary hospital that had, for six months, implemented real-time automated physiological monitoring. Data were analysed using reflexive thematic analysis with a hybrid (deductive-inductive) approach.ResultsOur analysis reveals three core mechanisms: (1) systematic arbitration (nurses as constant negotiators between automated alerts and situated clinical judgment); (2) datafied self-surveillance, patients internalise the data-driven gaze, policing their own behaviour; (3) a shared convergence around the human boundary, both nurses and patients actively defend empathy, contextual interpretation, and emotional recognition as domains irreducible to data. This boundary is not a passive remnant but an active, shared accomplishment.ConclusionDigital surveillance in chronic care does not act as a neutral adjunct. Instead, it operates as a disciplinary technology that asymmetrically disciplines nurses and patients. However, the identified human boundary represents a critical site of resistance and re-humanisation. The future of humane digital health depends on deliberately designing socio-technical systems that strengthen, rather than erode, the relational and interpretive core of clinical practice.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12910-026-01562-9. Background The persistent global burden of chronic diseases has catalysed the integration of digital monitoring systems into healthcare, a promise of personalised, efficient, and proactive management [1]. In chronic care settings, data-driven surveillance systems, comprising continuous remote monitoring, predictive analytics, and automated alerting, are increasingly deployed to track patient physiology and behaviour in real time [2]. Proponents argue these technologies enhance clinical outcomes through early intervention and data-informed decision support [3]. However, chronic care comes with distinct challenges: long-term disease trajectories, sustained behavioural change, complex self-management, and a nurse-patient relationship built on repeated encounters and trust. The introduction of continuous, data-centric surveillance into this intimate context potentially disrupts the very foundations of care, embodied presence, empathetic communication, and shared decision-making [4]. Yet, prevailing research and ethical guidelines largely approach such technologies through a techno-centric or normative lens, focusing on technical accuracy, data privacy, and abstract principles such as autonomy and beneficence. This discourse tends to treat monitoring tools as neutral adjuncts, overlooking how they actively reshape the social fabric of care. To critically investigate this transformation, we turn to Michel Foucault’s analysis of disciplinary power, particularly his later work on governmentality [5]. Governmentality shifts the focus from overt repression to the subtle ways in which technologies, knowledge, and everyday practices guide individuals into self-regulation. Unlike a simple panoptic metaphor, governmentality allows us to examine how digital monitoring systems produce certain forms of subjectivity, e.g., the “data-compliant patient” and the “system-responsive nurse”, and how power operates through the very desire to be healthy, efficient, and monitored [6]. Explicitly, we operationalise three dimensions of Foucauldian theory [7]: (a) visibility, how continuous data capture creates a field of permanent, asymmetrical observation; (b) authority, how dashboard-generated alerts acquire a seemingly objective, non-negotiable status; and (c) subjectification, how nurses and patients internalise the logic of surveillance, reshaping their professional and embodied selves. This study moves beyond cataloguing ethical concerns to empirically examine the mechanisms through which digital surveillance reconfigures power and intimacy in chronic care. By conducting a comparative qualitative analysis of both nurses and patients within the same digitally-mediated ecosystem, we ask: How does systematic surveillance operate as a disciplinary technology? What specific mechanisms produce asymmetric power dynamics? And where, and how, do the irreducible elements of humanistic care persist and resist? Our contribution is threefold: (1) providing an empirical application of Foucauldian governmentality to digital clinical surveillance; (2) uncovering concrete mechanisms of systematic arbitration and datafied self-surveillance; and (3) identifying the strategic convergence of a “human boundary” as a site of resistance and re-humanisation, with direct implications for design and policy. Methods Study design This study adopted a critical qualitative design grounded in a constructivist-interpretivist paradigm. We used Foucault’s concept of governmentality as a sensitising theoretical lens, not as a rigid framework, but as a guide to attend to practices of visibility, authority, and subjectification. Data were collected through semi-structured, in-depth interviews and analysed using reflexive thematic analysis with a hybrid (deductive-inductive) approach. The study was conducted and reported following the SRQR checklist. Description of the monitoring system The participating tertiary hospital had been running a real-time automated physiological alert platform (brand anonymised for confidentiality) for six months before data collection. The system continuously collected vital signs (heart rate, blood pressure, oxygen saturation) via bedside monitors and wearable sensors. Data were transmitted every 5–10 min to a central dashboard located at the nursing station. Alerts were generated when a parameter exceeded a predefined threshold (e.g., systolic BP > 180 mmHg or heart rate > 120 bpm); alerts appeared as colour-coded flags (yellow for warning, red for critical) accompanied by an audible beep. Nurses were expected to review the dashboard at least every two hours and to document their response to each alert. The system did not replace clinical decision making but provided “decision support”, though in practice, nurses reported pressure to follow alerts. No video monitoring was used. Critically, the system operated on simple rule-based threshold logic, it is therefore characterised as a rule-based digital monitoring tool. Setting and sampling The study was conducted in the cardiology wards of one tertiary public hospital in urban China. All patients were hospitalised inpatients at the time of interview. A purposive, maximum-variation sampling strategy was used: Nurses (n = 12): registered nurses with ≥ 6 months of hands-on experience with the monitoring system. We intentionally recruited nurses with varying seniority (≤ 5, 6–15, ≥ 16 years) to capture different levels of embeddedness in traditional vs. digitally-mediated workflows. Clarification on exposure duration: Although the hospital-wide system was launched six months before data collection, some patients had prior exposure (ranging from 3 to 8 months) to the same brand’s wearable monitoring devices through outpatient or community pilot programmes before this admission. This accounts for the total exposure duration of up to 14 months reported in Table 1. We stratified our sampling by cumulative exposure time and considered this variation in our analysis. Table 1Participant Characteristics (Nurses n = 12, Patients n = 15)CharacteristicNurses (n = 12)Patients (n = 15)Age (years), mean ± SD (range)34.2 ± 8.1 (24–51)61.4 ± 12.7 (45–79)Sex, female / male11 (91.7%) / 1 (8.3%)8 (53.3%) / 7 (46.7%)Clinical experience (years), mean ± SD (range)10.5 ± 7.3 (2–22)N/AProfessional title, Junior (< 5 yrs) / Intermediate (5–15 yrs) / Senior (> 15 yrs)4 (33.3%) / 5 (41.7%) / 3 (25.0%)N/ADisease type, Hypertension / Type 2 diabetesN/A9 (60%) / 6 (40%)Disease duration (years), mean ± SD (range)N/A7.2 ± 4.5 (1–18)System use duration, 3–6 months / 7–11 months / 12–14 monthsN/A6 (40%) / 5 (33.3%) / 4 (26.7%)Education level, Primary or below / Secondary / College or aboveN/A3 (20%) / 7 (46.7%) / 5 (33.3%)Residence, Urban / RuralN/A10 (66.7%) / 5 (33.3%)N/A not applicable, SD standard deviation Table 1Participant Characteristics (Nurses n = 12, Patients n = 15)CharacteristicNurses (n = 12)Patients (n = 15)Age (years), mean ± SD (range)34.2 ± 8.1 (24–51)61.4 ± 12.7 (45–79)Sex, female / male11 (91.7%) / 1 (8.3%)8 (53.3%) / 7 (46.7%)Clinical experience (years), mean ± SD (range)10.5 ± 7.3 (2–22)N/AProfessional title, Junior (< 5 yrs) / Intermediate (5–15 yrs) / Senior (> 15 yrs)4 (33.3%) / 5 (41.7%) / 3 (25.0%)N/ADisease type, Hypertension / Type 2 diabetesN/A9 (60%) / 6 (40%)Disease duration (years), mean ± SD (range)N/A7.2 ± 4.5 (1–18)System use duration, 3–6 months / 7–11 months / 12–14 monthsN/A6 (40%) / 5 (33.3%) / 4 (26.7%)Education level, Primary or below / Secondary / College or aboveN/A3 (20%) / 7 (46.7%) / 5 (33.3%)Residence, Urban / RuralN/A10 (66.7%) / 5 (33.3%)N/A not applicable, SD standard deviation Participant Characteristics (Nurses n = 12, Patients n = 15) N/A not applicable, SD standard deviation Patients (n = 15): adults diagnosed with hypertension or type 2 diabetes, using the monitoring system for ≥ 3 months. We varied disease type, duration of system use (3–14 months), and age. Recruitment continued until data sufficiency, the point where new interviews largely repeated existing themes and no major new insights emerged regarding power dynamics, privacy negotiation, or relational transformation. Detailed participant characteristics are presented in Table 1. Data collection In-depth, semi-structured interviews took place separately with nurses and patients between June and December 2025. All interviews were carried out by a female nurse researcher with 5 years of clinical experience, trained in qualitative methods in a private room within the ward. Interviews lasted 45–75 min, were audio-recorded with written informed consent, and transcribed verbatim in Chinese by a professional transcription service. Transcripts were returned to participants for member checking (all confirmed accuracy). Selected quotations were translated into English by a bilingual researcher (native Chinese, fluent English); a second researcher back-translated a random 20% to ensure fidelity. The interview guide was developed iteratively. Questions were framed to elicit narratives about power, visibility, discipline, and relational change, without using abstract theoretical terms. Examples: “Can you describe a recent situation where the monitoring system alerted you? How did you respond? “; “How does being constantly monitored affect your sense of privacy or autonomy?“; “Has the way you communicate with nurses/patients changed since the system was introduced?” Data analysis We used Braun and Clarke’s reflexive thematic analysis with a hybrid coding approach [8]. Initially, deductive codes were derived from Foucauldian concepts: “visibility”, “authority”, “internalised gaze”, “arbitration” [9]. Simultaneously, inductive codes were generated from the data (“dashboard first”, “feeling like a report card”, “machine cannot hear grief”). Coding was performed using NVivo 14 software. After coding all transcripts, we clustered codes into candidate themes separately for nurses and patients, then compared across groups to identify shared, divergent, and relational themes. Themes were refined through team discussions and then theoretically integrated, i.e., we asked how each theme illustrated mechanisms of power, visibility, and subjectification. This process produced the three final analytical themes presented below. To enhance interpretive rigour, a second researcher independently coded 20% of transcripts. Rather than seeking convergence as a reliability check, we treated discrepancies as generative opportunities for reflexive dialogue, using them to surface alternative readings, challenge interpretive assumptions, and deepen our collective engagement with the data. This process enriched the analytical narrative without invoking a reliability framework. Confirmability was addressed through reflexive journals and explicit reporting of our positionality. Ethical considerations The study was approved by the Institutional Review Board of the First Affiliated Hospital of Wenzhou Medical University (YS20251416). All procedures adhered to the Declaration of Helsinki. Written informed consent was obtained, with explicit statements that non-participation or withdrawal would not affect care. Data were pseudonymised (codes N1–N12, P1–P15) and stored on a password-protected university server. For patient participants, special attention was paid to ensure understanding of data usage and the voluntary nature of participation. Given that the interviewer was a hospital-employed physician, we implemented additional safeguards for nurse participants: we explicitly stated in the information sheet and verbally that the study was “not” a performance evaluation, that individual responses would never be communicated to nursing managers or the hospital quality department, and that participation carried no implications for job security or professional advancement. We also offered nurses the option to withdraw their data up to two weeks after the interview without providing any reason. These measures were designed to reduce the perceived risk of reprisal and to encourage candid disclosure. Results Participant Characteristics The nurse cohort (N = 12) had a mean age of 34.2 years (SD = 8.1) and had been in clinical practice for 10.5 years (SD = 7.3). The patient cohort (N = 15) had a mean age of 61.4 years (SD = 12.7); Nine had hypertension, and six had type 2 diabetes. System use ranged from 3 to 14 months. Table 1 provides detailed sociodemographic and clinical information. Analytical themes: mechanisms of disciplinary power in digitally-mediated care Three interconnected themes were identified: (1) Systematic Arbitration, nurses as constant negotiators between data alerts and clinical judgment; (2) Datafied Self-Surveillance, patients internalising the data-driven gaze; and (3) Convergence around the Human Boundary, Mutual Defence and Shared Recognition. Table 2 provides a hierarchical overview of the three main themes and their constituent sub‑themes. Table 2Thematic overview: themes, sub-themes, and brief descriptionsTheme• Sub-themeTheme 1: Systematic Arbitration, Nurses between the Dashboard and the Bedside• Sub-theme 1.1: The dashboard as a “silent supervisor”, Nurses perceive the alert system as an authoritative presence that demands a documented response regardless of clinical context, effectively re-ordering their workflow from “patient first” to “dashboard first”.• Sub-theme 1.2: Negotiating between data logic and tacit knowledge, Nurses constantly justify their situated clinical judgment against data-driven prompts, often through defensive documentation that consumes time otherwise available for direct patient interaction.Theme 2: Datafied Self-Surveillance, Patients Internalising the Data-Driven Gaze• Sub-theme 2.1: The body as an auditable object, Patients feel that their moment-to-moment behaviours (eating, resting, emotional responses) are continuously evaluated by the monitoring system, turning their lived bodies into visible, gradable data streams.• Sub-theme 2.2: Variation by exposure and age, Patients with longer cumulative exposure show more fatigue and “gaming” behaviours (e.g., timing activities to avoid alerts); older patients express resignation, while younger patients adopt micro-resistance strategies.• Sub-theme 2.3: Mixed experiences, reassurance amid unease, Some patients derive security from continuous monitoring (“someone is watching over me”), yet this positive perception coexists with self-policing and anxiety, depending on individual disposition and social support.Theme 3: Convergence around the Human Boundary, Mutual Defence and Shared RecognitionSub-theme 3.1: What the system cannot see, emotion and life context, Patients consistently draw a line at emotional recognition, distinguishing automated alerts from nurses’ ability to ask about relational or financial stressors that affect physiology.• Sub-theme 3.2: Nurses actively reclaiming the interpretive space, Nurses deliberately protect time and space for empathetic, contextual communication, sometimes by intentionally delaying response to non-urgent alerts to finish a meaningful conversation.• Sub-theme 3.3: A shared vision for synergistic partnership, Both groups converge on a model where technology handles logistical surveillance (tracking, alerting), freeing human practitioners for interpretation, empathy, and motivational support. Table 2Thematic overview: themes, sub-themes, and brief descriptionsTheme• Sub-themeTheme 1: Systematic Arbitration, Nurses between the Dashboard and the Bedside• Sub-theme 1.1: The dashboard as a “silent supervisor”, Nurses perceive the alert system as an authoritative presence that demands a documented response regardless of clinical context, effectively re-ordering their workflow from “patient first” to “dashboard first”.• Sub-theme 1.2: Negotiating between data logic and tacit knowledge, Nurses constantly justify their situated clinical judgment against data-driven prompts, often through defensive documentation that consumes time otherwise available for direct patient interaction.Theme 2: Datafied Self-Surveillance, Patients Internalising the Data-Driven Gaze• Sub-theme 2.1: The body as an auditable object, Patients feel that their moment-to-moment behaviours (eating, resting, emotional responses) are continuously evaluated by the monitoring system, turning their lived bodies into visible, gradable data streams.• Sub-theme 2.2: Variation by exposure and age, Patients with longer cumulative exposure show more fatigue and “gaming” behaviours (e.g., timing activities to avoid alerts); older patients express resignation, while younger patients adopt micro-resistance strategies.• Sub-theme 2.3: Mixed experiences, reassurance amid unease, Some patients derive security from continuous monitoring (“someone is watching over me”), yet this positive perception coexists with self-policing and anxiety, depending on individual disposition and social support.Theme 3: Convergence around the Human Boundary, Mutual Defence and Shared RecognitionSub-theme 3.1: What the system cannot see, emotion and life context, Patients consistently draw a line at emotional recognition, distinguishing automated alerts from nurses’ ability to ask about relational or financial stressors that affect physiology.• Sub-theme 3.2: Nurses actively reclaiming the interpretive space, Nurses deliberately protect time and space for empathetic, contextual communication, sometimes by intentionally delaying response to non-urgent alerts to finish a meaningful conversation.• Sub-theme 3.3: A shared vision for synergistic partnership, Both groups converge on a model where technology handles logistical surveillance (tracking, alerting), freeing human practitioners for interpretation, empathy, and motivational support. Thematic overview: themes, sub-themes, and brief descriptions Theme 1: systematic arbitration, nurses between the dashboard and the bedside Sub-theme 1.1: the dashboard as a “silent supervisor” Nurses consistently described the monitoring system as an authority they could not ignore. N5 (staff nurse, 6 years experience): “The system flags a risk. I look at the patient, who is resting comfortably. But I still have to document a response. The algorithm is a silent, powerful colleague, or supervisor.” N10 (senior nurse, 22 years): “My first point of contact is now the dashboard. I assess the numbers before I assess the face.” Even nurses who doubted the alert’s clinical relevance felt compelled to act: N2 (junior nurse, 2 years): “If I don’t follow the red alert immediately, the head nurse will ask why. So I go, even when I know the patient is fine.” Sub-theme 1.2: negotiating between data logic and tacit knowledge Nurses described a constant arbitration process. This negotiation often placed nurses in a defensive position, having to justify human intuition against data-driven prompts. N7 (nurse, 8 years): “Sometimes I spend five minutes writing a note to explain why I didn’t follow the machine’s suggestion. That’s time I could have spent talking to the patient.” N9 (nurse, 15 years): “The dashboard is useful, but it doesn’t tell me if the patient is anxious or in pain. I’ve learned to trust my eyes more than the numbers. Still, I have to document why I overrode the alert. It feels like the system is always second-guessing me.” N4 (nurse, 12 years): “Sometimes the alert is real, sometimes it’s a false alarm. But you can’t just ignore it. I’ve developed a habit: glance at the dashboard, then walk into the room and look at the patient’s face. The face tells me more than the numbers – but I still have to click ‘reviewed’ on the screen.” Theme 2: datafied self-surveillance, patients internalising the data-driven gaze Sub-theme 2.1: The body as an auditable object Patients spoke of feeling constantly watched, not by a person, but by the system. P8 (male, 71 years, hypertension, System use 7 months): “I never paid attention to my heart rate before. Now the screen shows it every few minutes. I find myself holding my breath, waiting to see if it’s normal. It’s like the number decides how I feel.” P11 (male, 68 years, hypertension, System use 9 months): “I think twice about what I eat because I know the graph will show it later. It’s like someone is always watching.” P3 (female, 59 years, hypertension, System use 4 months): “My husband used to remind me to take my pills. Now my phone does it. But the thing is, the phone also records when I miss a dose. I feel like I’m being graded.” Sub-theme 2.2: variation by system experience and age Patients with longer system exposure (≥ 8 months) showed more acceptance but also more fatigue. P14 (female, 65 years, hypertension, System use 14 months): “At first I was nervous. Now I just feel tired of being measured. But I still check, the habit is there.” P1 (male, 52 years, hypertension, System use 11 months): “I figured out that if I sit still for five minutes before the automatic reading, my numbers look better. So that’s what I do. The machine doesn’t know the difference, and the nurse stops bothering me.” P13 (female, 73 years, hypertension, System use 6 months): “I don’t really understand all the numbers on the screen. My daughter helps me. But I know the machine is watching. Sometimes I just close my eyes and pretend it’s not there. What can I do? I’m too old to argue with a machine.” Younger patients (≤ 60 years) were more likely to “game” the system (e.g., timing eating to avoid alerts), whereas older patients expressed resignation. No patient reported complete indifference, the gaze was always felt. Sub-theme 2.3: mixed experiences, reassurance amid unease While anxiety and fatigue were prevalent, several patients also expressed a sense of safety derived from continuous monitoring. P6 (male, 58 years, hypertension, system use 5 months): “The beeping makes me feel secure. I know someone is watching over me even when the nurse is busy with another patient.” P10 (female, 62 years, diabetes, system use 6 months): “I was scared at first, but now I feel it is like a safety net. If something goes wrong, they will know immediately.” These positive perceptions co-existed alongside the disciplinary tensions described above, suggesting that the same technology can generate both reassurance and self-policing depending on individual disposition, social support, and illness trajectory. Theme 3: convergence around the human boundary, mutual defence and shared recognition Sub-theme 3.1: What the system cannot see, emotion and life context Patients consistently drew a boundary at emotional recognition. P12 (female, 69 years, hypertension, System use 10 months): “The machine sends an alert when my blood pressure is high. But it doesn’t know that I just had a fight with my daughter-in-law. My nurse called me later and said, ‘You sound upset, what happened?’ That’s the difference.” P5 (female, 55 years, hypertension, System use 8 months): “The alarm went off because my blood pressure was high. But I was just upset about a bill. The machine doesn’t know that. My nurse came in, sat down, and asked me what was really going on. That conversation helped more than any alert.” P9 (male, 62 years, hypertension, System use 5 months): “The nurse asked me why my blood pressure was high. I told her about my son’s divorce. She listened. The machine would never ask that question.” Sub-theme 3.2: Nurses actively reclaiming the interpretive space Nurses strategically emphasised empathy and contextual questioning as their unique value. N12 (nurse, 18 years): “The system is good for tracking, but I am the one who sees the fear in the eyes. I make sure that conversation still happens, even when the dashboard is beeping.” N3 (nurse, 4 years): “When I have time, I put the phone away and just sit. The patient needs to know that someone sees them as a person, not a set of numbers.” N1 (junior nurse, 2 years): “My preceptor taught me to always look at the patient first, then the screen. Sometimes the numbers look scary, but the patient is smiling and talking. Other times the numbers are fine, but the patient looks terrible. You have to put the two together. The machine alone will fool you.” This was not passive, nurses described actively protecting time for human interaction, sometimes by intentionally delaying their response to non-urgent alerts. Sub-theme 3.3: a shared vision for synergistic partnership Both groups converged on a model where digital systems handle logistical surveillance (tracking, alerting), freeing humans for interpretive, empathetic, and motivational work. P15 (female, 67 years, diabetes, System use 8 months): “I trust my nurse because she understands the machine but also sees me. I don’t want the machine to replace her, I want it to help her.” P7 (male, 64 years, hypertension, System use 9 months): “I like that the machine tracks my numbers, but I don’t want it to replace my nurse. She explains what the numbers mean and calms me down when I’m worried. The machine just beeps. I need both.” N6 (nurse, 10 years): “If the system took over the routine checks, I could spend that half hour really talking to the patient. That would be the best use of AI.” Summary of findings The results show that data-driven surveillance operates as a disciplinary technology through two primary mechanisms, systematic arbitration (for nurses) and datafied self-surveillance (for patients). However, both groups converge around a human boundary where empathy, context, and emotional meaning are preserved as irreducible domains. This boundary is not a passive limit but an active site of resistance and re-humanisation. Discussion This study provides a critical empirical examination of data-driven surveillance in chronic care, using Foucault’s governmentality lens. Our findings illuminate three interconnected mechanisms, systematic arbitration, datafied self-surveillance, and convergence around the human boundary, that asymmetrically reconfigure power, professional judgment, and intimacy. Below, we interpret these findings within the broader literature on digital health, nursing professionalism, and critical data studies. Systematic arbitration: nurses as negotiating agents Our finding that nurses become constant “arbitrators” between automated alerts and situated clinical judgment builds on prior work on human-in-the-loop systems[ 10]. Unlike earlier accounts that portray the human role as a simple override function, our data show that arbitration is a continuous, emotionally taxing, and professionally ambiguous activity. Nurses are not merely checking the dashboard; they are defending their own judgment against a system that carries an aura of objective, data-driven authority. This resonates with recent critiques of “data authority” in healthcare, but our study demonstrates how this plays out in the specific context of chronic care nursing, where the frequency of alerts is high and the relational stakes are intimate. Comparisons with international literature reveal both convergence and divergence. Studies from Western healthcare settings have reported similar tensions between protocol-driven alerts and clinical autonomy [11]. However, nurses in our Chinese sample expressed a stronger sense of hierarchical accountability, the system’s alerts were often perceived as directives from the hospital administration, not just a clinical suggestion [12]. This may reflect the more hierarchical organisational culture in Chinese tertiary hospitals, where deviation from standardised protocols carries higher personal risk. Future cross-cultural research should directly compare how organisational context moderates the experience of systematic arbitration. Datafied self-surveillance: the internalised gaze Patients in our study described a pervasive sense of being watched by an opaque system, leading to self-policing behaviours (e.g., altering eating habits to avoid alert triggers). This mechanism aligns with Foucault’s description of the panoptic effect [13], but our findings show how digital data streams, rather than a physical watchtower, create a distributed, ever-present gaze. Compared to previous qualitative work on patient-generated health data [14], our participants experienced institutional surveillance (hospital-mandated) as more anxiety-provoking than self-tracking for personal wellness. The “datafied self” in chronic care is not a voluntary project of self-improvement; it is an auditable body subjected to external evaluation. This distinction has important implications: designers should consider adding features that give patients more control over how and when their data are viewed by clinicians, reducing the sense of exposure. Cross-nationally, our findings echo concerns raised in European studies about “dataveillance” in chronic disease management [15]. However, the internalisation we observed was more pronounced among older patients (≥ 70 years), who felt less able to negotiate with the system. By contrast, younger patients occasionally “gamed” the system, a form of micro-resistance that has been reported in other digital health contexts [16]. Recognising these age-based variations can help tailor support strategies. Convergence around the human boundary: resistance and re-humanisation A key contribution of this study is the identification of a strategic, shared human boundary. Both nurses and patients actively defended empathy, contextual interpretation, and emotional recognition as domains irreducible to data. This finding moves beyond the common lament that technology erodes human contact [17]; instead, we show that the boundary is an accomplishment, something that requires deliberate effort, time, and mutual recognition. Comparable concepts appear in the literature on “relationship-centred care” [18] and “compassionate technology” [19], but these tend to be normative ideals rather than empirically grounded mechanisms. Our study provides concrete examples of how nurses protect this boundary (e.g., intentionally delaying non-urgent alerts to finish a conversation) and how patients recognise and value that effort. This suggests that the human boundary is not a passive remnant of pre-digital care; it is a site of active resistance against totalising datafication. A qualitative study on digital tools in primary care [20] highlighted persistent implementation barriers (e.g., workflow misalignment, trust concerns) and noted a lack of institutional support for clinicians’ relational roles, suggesting that any defence of empathetic time remains largely individualised and unsupported. In our Chinese setting, nurses described a collective, informal norm, experienced nurses taught junior nurses how to “manage the dashboard without losing the patient”. This informal peer socialisation may be a resource that other healthcare systems could intentionally cultivate. Ethical implications of datafied surveillance The datafication of health, as conceptualised by Lupton [21], provides an essential backdrop for understanding how patients’ bodies become auditable objects and how self‑tracking shifts from a voluntary project to an institutional mandate. Our findings raise several critical ethical issues beyond privacy and consent. First, “informed consent in continuous monitoring” requires re-examination: while patients signed generic consent forms, few understood that their moment-to-moment behaviours (eating, resting, emotional arousal) would be rendered visible and auditable. This echoes concerns about “digital vulnerability” [22], particularly among older adults who lacked digital literacy to negotiate system demands. Second, “the possibility of meaningful refusal” appeared largely theoretical, no patient in our sample felt able to decline monitoring without jeopardising their care relationship, a form of “coercive self-surveillance” that blurs the line between therapeutic benefit and institutional control. Third, “behavioural gaming” represents an unintended ethical consequence: patients learn to perform “good data” rather than engage in genuine self-management, potentially undermining clinical goals [23]. Fourth, “workload redistribution”, nurses spending time documenting override justifications, raises questions about opportunity costs: does surveillance accountability detract from direct care time? These ethical tensions suggest that future implementations must build in transparent data governance, patient control over data visibility, and organisational policies that explicitly protect relational care activities from being squeezed by documentation burdens. Implications for nursing practice and digital health design Three practical implications emerge. First, critical digital literacy should become a core nursing competency, not just how to use digital systems, but how to question, negotiate, and when appropriate, contest automated system outputs. Second, workflow design must protect space for human interaction; metrics that count only alert response times may inadvertently punish nurses who take extra minutes to address emotional needs. Third, re-humanisation by design points towards system features that support the human boundary. Although not directly tested in this study, one potential design direction suggested by our findings would be a “context‑aware alert hold” function that temporarily suppresses non‑urgent alerts during scheduled patient conversations (activated manually by the nurse, with a mandatory safety override for red‑critical alerts and an automatic time‑out after 5 min to ensure delayed alerts are later reviewed). These suggestions are intended as generative directions for future participatory design research, not as empirically validated solutions. Limitations This study has several limitations. First, it was conducted in a tertiary hospital in urban China; cultural and organisational factors may limit transferability to other settings (e.g., primary care, other countries). Second, the monitoring system studied was rule-based rather than AI-driven; our findings may not fully capture the distinct power dynamics introduced by self-learning predictive algorithms. We deliberately chose to study this system because it represents the majority of currently deployed “AI-labelled” tools in Chinese hospitals, yet our conclusions about datafication and surveillance are more directly applicable to rule-based digital monitoring than to advanced machine learning systems. Third, we relied on self-reported interviews rather than direct observation; future research using video-ethnography could capture moment-by-moment negotiation. Fourth, our sample did not include patients who refused the monitoring or nurses who left because of the system, their perspectives might reveal stronger resistance or different forms of agency. Fifth, the cross-sectional design cannot capture how power dynamics evolve as such monitoring systems become routinised. Sixth, despite reflexive efforts, the Foucauldian lens foregrounds power and discipline, potentially downplaying seamless, positive collaboration. We actively searched for disconfirming cases (e.g., nurses who praised the system without reservation) and report that they were rare but present, further research should explore conditions under which digital monitoring enhances rather than disciplines care. Seventh, our own biases may have shaped data interpretation. We note this not as a flaw, but as a typical feature of reflexive qualitative research: the analysis evolved from a more techno-optimistic stance to a more critical one as we confronted disconfirming evidence (e.g., the many patients who described anxiety rather than relief). Future researchers should be aware that their professional backgrounds may similarly colour their initial readings. Conclusion This study reframes data-driven surveillance in chronic care as a disciplinary technology that operates through two primary mechanisms: systematic arbitration (nurses constantly negotiating between data alerts and situated judgment) and datafied self‑surveillance (patients internalising the data-driven gaze). However, we also identified a critical counter‑movement, a shared convergence around a human boundary where empathy, context, and emotional meaning are actively defended by both nurses and patients, defining the limits of data-driven governance. A key priority for future ethical digital health should be a deliberate move towards “re‑humanisation by design”, deliberately architecting systems that empower nurses as critical practitioners and protect time for human connection. Our findings imply that digital systems should handle logistical surveillance, while humans remain irreplaceable for interpretation, empathy, and emotional recognition. What this paper adds to the field is an empirical application of Foucauldian governmentality to rule-based digital monitoring in nursing, concrete mechanisms of systematic arbitration and datafied self‑surveillance, and a grounded model of the human boundary as a site of resistance and re‑humanisation, offering actionable guidance for designers, managers, and policymakers. Supplementary Information Supplementary Material 1. Supplementary Material 1.