From Smartwatch to ICU: The Future of Care

How wearable devices and artificial intelligence are dismantling the wire-bound paradigm of intensive care—one sensor at a time.

By Michele D. Pierri, MD, PhD

Introduction: The Modern Monitoring Paradox

A substantial share of adults now wear a device capable of tracking heart rate (HR), blood oxygen saturation (SpO2), and sleep architecture. Yet the moment a critically ill patient is admitted to an Intensive Care Unit (ICU), they are tethered to a tangle of cables that restricts mobility, complicates nursing procedures, and interposes a physical barrier between clinician and patient and, most poignantly, between parent and newborn.

This paradox is not merely aesthetic. Wire-bound monitoring constrains early mobilization, a factor independently associated with shorter mechanical ventilation duration and reduced ICU length of stay (LOS). The objective of contemporary MedTech research is to resolve this contradiction: to transfer the unobtrusive, continuous sensing philosophy of consumer wearables into the most demanding clinical environments on earth, ensuring seamless monitoring that accompanies the patient from the ICU bed to home discharge.

A comprehensive 2025 review published in Intensive Care Medicine Experimental by Angelucci and colleagues from Politecnico di Milano and Humanitas University confirmed that wearable devices (WDs), defined as integrated systems comprising biological sensors, embedded microcontrollers, radiofrequency antennae, and power management circuitry, are now proving valuable across a broad spectrum of ICU applications: vital signs monitoring, delirium detection, glycemic surveillance, and pressure injury prevention [1]. Their review emphasized that validation of accuracy and integration into clinical decision-making workflows remain the foremost challenges for widespread adoption.

Key Takeaway: Consumer wearable technology has matured to the point where its underlying principles (miniaturized sensors, wireless data transmission, and algorithmic signal processing) are being reengineered for clinical-grade applications in critical care. Here are five breakthroughs that are leading this transformation.

1. From Fitness Tracker to Clinical-Grade System

A crucial technical distinction must be drawn at the outset. The term “sensor” refers narrowly to the transducer element that detects a biological signal: a photoplethysmographic (PPG) diode for pulse oximetry, a piezoelectric element for respiratory rate (RR), or a thermistor for core temperature. A wearable device, by contrast, is an integrated system that includes the sensor itself, embedded microcontrollers for on-board signal processing, radio-frequency antennae (typically Bluetooth Low Energy, or BLE, or proprietary low-power protocols), and power management circuitry. It is this systems-level architecture that enables the “wireless revolution”: transforming a heartbeat detected at the skin surface into a clinically actionable datum transmitted in real time to a centralized monitoring dashboard.

This transition toward non-invasive, cost-effective wireless systems is especially vital in low- and middle-income countries (LMICs), where staffing shortages render traditional one-nurse-per-patient monitoring unsustainable. The World Health Organization (WHO) has highlighted the potential of wearable monitoring for resource-limited settings.

The wearable medical device market reflects this momentum. Industry projections anticipate rapid growth through the next decade, driven by chronic disease burden, aging demographics, and the convergence of sensor miniaturization with 5G connectivity and edge computing. The United States Food and Drug Administration (FDA) has cleared 900+ artificial intelligence/machine learning (AI/ML)-enabled medical devices as of 2024, with ongoing additions each year across clinical specialties.

2. AI as a Time Machine: Predicting Sepsis Before Symptoms

Sepsis remains among the leading causes of in-hospital mortality. Its pathophysiology is governed by a cruel temporal equation: for every hour that antimicrobial therapy is delayed after the onset of septic shock, mortality increases substantially. Machine learning (ML) models that ingest continuous wearable data streams are now functioning as clinical early-warning systems—a form of “time machine” that identifies the trajectory toward sepsis well before clinical signs become manifest.

TREWS: Targeted Real-Time Early Warning System

Developed at Johns Hopkins University and later commercialized by Bayesian Health, the Targeted Real-time Early Warning System (TREWS) is among the most rigorously evaluated ML-based sepsis tools in the literature. In a landmark prospective, multi-site cohort study published in Nature Medicine, Adams and colleagues monitored 590,736 patient encounters across five hospitals. Among the 6,877 sepsis patients identified by the alert prior to antibiotic initiation, provider confirmation of the alert within three hours was associated with a 3.3% adjusted absolute reduction (18.7% relative reduction) in in-hospital mortality, along with decreased organ failure scores and shorter hospitalization [2].

Metric	Value
Patients monitored (TREWS prospective study)	590,736 across 5 hospitals
Median reduction in time to first antibiotic order	1.85 hours
Adjusted relative reduction in in-hospital mortality	18.7% (95% CI: 9.4–27.0%)
Provider adoption rate	89% sustained

A companion paper by Henry and colleagues, published in the same issue of Nature Medicine, further demonstrated that TREWS achieved 82% sensitivity in identifying retrospectively confirmed sepsis cases, with 89% of alerts evaluated by a physician or advanced practice provider [3].

COMPOSER and the Next Generation

At the University of California San Diego Health, the COMPOSER (COnformal Multidimensional Prediction Of SEpsis Risk) algorithm monitors over 150 clinical variables in real time within emergency departments (EDs). A quasi-experimental before-and-after study by Boussina and colleagues (2024, npj Digital Medicine) found that deployment of COMPOSER through a nurse-facing Best Practice Advisory (BPA) was associated with significant improvements in sepsis bundle compliance and survival [4]. Separately, Dascena’s InSight system—which holds FDA clearance—has demonstrated the capacity to predict sepsis onset up to 48 hours before clinical symptoms appear.

⚠ The Alarm Fatigue Problem: Even high-performing models like TREWS can generate false positives that desensitize clinicians over time. A 2025 review highlighted that TREWS, despite its high area under the receiver operating characteristic curve (AUC), exhibits a relatively low positive predictive value (PPV) in some settings, which can undermine real-world actionability and contribute to alarm fatigue—a well-documented patient safety concern [10]. Balancing sensitivity and specificity remains the core engineering challenge for any AI early-warning system deployed in critical care.

3. Reclaiming the Human Touch: Wireless Monitoring in Neonatology

Nowhere is the clinical impact of wireless monitoring more profound than in neonatology. In the Neonatal Intensive Care Unit (NICU), premature infants are surrounded by masses of wires that often exceed the infant’s own body weight in sheer bulk. These cables not only complicate routine nursing care—feeding, diaper changes, thermoregulation—but create a formidable physical barrier to skin-to-skin contact, commonly known as kangaroo care (KC), a practice supported by robust evidence for improving neurodevelopment, weight gain, thermal stability, and reduced infection rates.

A 2024 systematic review published in Frontiers in Pediatrics by Krbec and colleagues documented the iatrogenic consequences of conventional wired monitoring in neonates: adhesive-related epidermal stripping (approximately 4% of NICU graduates bear cosmetically or functionally significant scars), increased infection risk from frequent isolette openings to adjust sensors, and pain exposure linked to suboptimal brain growth [7]. These are not theoretical concerns—they represent measurable, evidence-based harms.

Researchers at Northwestern University and Ann & Robert H. Lurie Children’s Hospital have developed soft, battery-powered wireless epidermal sensors that provide measurement equivalency to bedside monitors for HR, RR, temperature, and SpO₂—while additionally capturing blood pressure (BP), infant crying, movement, and body position. Their work, published in Nature Medicine by Chung and colleagues (2020), demonstrated clinical validation in pilot studies involving 50 premature infants [6]. A key finding was the ability to quantify the physiological benefits of kangaroo care in real time—something impossible with tethered systems.

Beyond Monitoring — The Psychological Dimension: The removal of wires is not merely a technical convenience. Published literature consistently identifies the physical barrier of cables as a primary source of parental frustration and anxiety in the NICU [8]. Wireless systems enable unrestricted skin-to-skin holding, facilitate breastfeeding, and transform the NICU into a less traumatizing environment for families—a goal increasingly recognized as a clinical outcome in its own right.

4. Sensors Against “Invisible Wounds”: Preventing Pressure Injuries

Hospital-acquired pressure injuries (HAPIs) are among the most common, costly, and preventable complications in acute care. In the United States alone, over 2.5 million patients develop pressure injuries annually, with individual treatment costs ranging from $5,000 to over$ 100,000 depending on severity. National data suggests that conventional turn protocol adherence—using paper-based reminders—hovers around an alarmingly low 47–48%.

The LEAF Patient Monitoring System (Smith+Nephew) addresses this gap through a lightweight, single-use, disposable inertial measurement unit (IMU) sensor adhered to the patient’s chest. The system continuously monitors patient position, orientation, movement, and activity, wirelessly transmitting data to a nursing station interface. It provides visual alerts when repositioning is due according to an individualized protocol and introduces the Integrated Positioning Index™ (IPI), which combines turn frequency, turn angle, and tissue reperfusion time into a single actionable metric.

Metric	Value
HAPI reduction (RCT, Pickham et al., 2018; n = 1,312)	73% (per-protocol analysis)
Estimated annual cost savings per patient	$6,621
Peak turn protocol adherence at implementing facilities	Up to 98%
Cumulative monitoring data (as of late 2022)	>7 million hours, >60,000 patients

The clinical evidence base is substantial. A pragmatic randomized controlled trial (RCT) involving 1,312 ICU patients (Pickham et al., 2018, International Journal of Nursing Studies) demonstrated that sensor-guided repositioning reduced HAPI incidence from 2.7% to 0.7% on per-protocol analysis [5]. An economic analysis by Nherera and colleagues, published in the International Journal of Health Economics and Management (2021), confirmed the cost-effectiveness of the system from a payer perspective, estimating a $6,621 cost saving per patient annually and projecting that a cohort of 1,000 patients would avoid approximately 203 HAPIs per year [9]. One facility reported an 84.6% reduction in sacrococcygeal HAPIs from baseline (p < .00001), with estimated annual savings exceeding $1 million.

5. The Acid Test: When Laboratory Accuracy Meets Bedside Reality

As a clinician-scientist, it is essential to temper enthusiasm with methodological rigor. Validation studies conducted under controlled, often static, conditions can paint an overly optimistic picture. For example, devices such as the mWear wrist-worn monitor have demonstrated 94% concordance with standard bedside monitors (e.g., BeneVision N15) for HR and SpO₂ measurements in controlled settings. The 2025 Intensive Care Medicine Experimental review confirmed that HR measurements from wearable devices correlate strongly with both manual nursing observations and standard monitors, though this correlation weakens notably for RR [1].

The Signal Quality Bottleneck: A critical and often overlooked metric is signal usability. While bedside monitors typically lose only about 1% of recorded data, published studies have found that up to 76% of data from wrist-worn devices can be rendered unusable due to patient movement artifacts and electromagnetic interference. This gap between point-accuracy (how correct a reading is when obtained) and continuous reliability (how consistently usable data is generated) represents the primary engineering bottleneck for universal clinical adoption of wearable monitoring in acute care.

The same 2025 review underscored that most validation studies for ICU wearables have been conducted in community settings with healthy adults, raising questions about the reliability of these devices in hemodynamically unstable or critically ill populations [1]. Medical-grade wearable patches have fared better in surgical patient cohorts, suggesting that form factor, sensor placement, and adhesion technology all influence real-world performance.

However, the field is advancing rapidly. A 2025 study published in Nature Communications by Scheid and colleagues from Northwell Health demonstrated that a recurrent neural network (RNN) trained on continuous wearable vital sign data from 888 non-ICU patients could predict clinical deterioration alerts and adverse outcomes up to 17 hours in advance, with an area under the receiver operating characteristic curve (AUROC) of 0.89 [11]. This represents a paradigm shift: the combination of clinical-grade wearables with deep learning moves monitoring from passive observation to active, predictive surveillance.

Conclusion: Toward a Hospital Without Walls

The trajectory of these five innovations converges toward a single vision: an invisible yet omnipresent monitoring architecture that ensures continuity of care from the ICU bed to the patient’s home. In this paradigm, the hospital’s boundaries dissolve; monitoring becomes a persistent, unobtrusive service rather than a location-dependent infrastructure.

The clinical evidence reviewed here—from TREWS’s mortality reductions to the LEAF system’s pressure injury prevention, from wireless neonatal sensors enabling kangaroo care to wearable deep learning models predicting deterioration hours in advance—demonstrates that we have crossed the threshold from theoretical promise to measured clinical benefit.

Yet a fundamental question persists, one that touches the philosophical foundations of clinical practice: are we prepared to trust an algorithm with a life-or-death decision, or must technology remain a silent ally to human judgment? The evidence suggests a middle path. The highest-performing sepsis prediction systems do not replace the clinician; they amplify the clinician’s cognitive bandwidth, flagging deterioration that might otherwise be buried in the noise of a busy ICU. The TREWS data are instructive: outcomes improved only when providers confirmed the alert within three hours—a design that deliberately preserves the human in the loop.

Technology, at its best, does not supplant the art of medicine. It restores time and proximity—the two resources most eroded by the complexity of modern critical care—so that clinicians and families can be present where it matters most.

References

Angelucci A, Greco M, Cecconi M, Aliverti A. Wearable devices for patient monitoring in the intensive care unit. Intensive Care Med Exp. 2025;13:26. doi:10.1186/s40635-025-00738-8
Adams R, Henry KE, Sridharan A, et al. Prospective, multi-site study of patient outcomes after implementation of the TREWS machine learning-based early warning system for sepsis. Nat Med. 2022;28(7):1455-1460. doi:10.1038/s41591-022-01894-0
Henry KE, Adams R, Parent C, et al. Factors driving provider adoption of the TREWS machine learning-based early warning system and its effects on sepsis treatment timing. Nat Med. 2022;28(7):1447-1454. doi:10.1038/s41591-022-01895-z
Boussina A, Shashikumar SP, Malhotra A, et al. Impact of a deep learning sepsis prediction model on quality of care and survival. npj Digit Med. 2024;7:14. doi:10.1038/s41746-023-00986-6
Pickham D, Berte N, Pihulic M, Valdez A, Mayer B, Desai M. Effect of a wearable patient sensor on care delivery for preventing pressure injuries in acutely ill adults: a pragmatic randomized clinical trial (LS-HAPI study). Int J Nurs Stud. 2018;80:12-19. doi:10.1016/j.ijnurstu.2017.12.012
Chung HU, Rwei AY, Herber A, et al. Skin-interfaced biosensors for advanced wireless physiological monitoring in neonatal and pediatric intensive-care units. Nat Med. 2020;26(3):418-429. doi:10.1038/s41591-020-0792-9
Krbec M, Zhang X, Chityat I, et al. Emerging innovations in neonatal monitoring: a comprehensive review of progress and potential for non-contact technologies. Front Pediatr. 2024;12:1442753. doi:10.3389/fped.2024.1442753
Zhou L, Guess M, Kim KR, Yeo WH. Skin-interfacing wearable biosensors for smart health monitoring of infants and neonates. Commun Mater. 2024;5(1):72. doi: 10.1038/s43246-024-00511-6. Epub 2024 May 9. PMID: 38737724; PMCID: PMC11081930.
Nherera L, Larson B, Cooley A, Reinhard P. An economic analysis of a wearable patient sensor for preventing hospital-acquired pressure injuries among the acutely ill patients. Int J Health Econ Manag. 2021;21(4):457-471. doi:10.1007/s10754-021-09304-7
Papareddy P, Lobo TJ, Holub M, Bouma H, Maca J, Strodthoff N, Herwald H. Transforming sepsis management: AI-driven innovations in early detection and tailored therapies. Crit Care. 2025 Aug 19;29(1):366. doi: 10.1186/s13054-025-05588-0. PMID: 40830514; PMCID: PMC12366378.
Scheid MR, Friedmann B, Oppenheim M, et al. Development and validation of a clinical wearable deep learning based continuous inhospital deterioration prediction model. Nat Commun. 2025;16:9513. doi:10.1038/s41467-025-65219-8

Disclaimer: This article is intended for educational and informational purposes only. It does not constitute medical advice. Clinical decisions should always be made in consultation with qualified healthcare professionals and in accordance with institutional protocols. The author declares no conflicts of interest related to the technologies discussed.

From Smartwatch to ICU: 5 Breakthroughs Revolutionizing Hospital Monitoring

In This Article