Advances in technology have led to the development of wearable, portable, and point-of-care devices, which are gaining attention in health care. In critical care, continuous monitoring enables early detection of deterioration, real-time recognition of life-threatening events, and supports decision-making. Monitoring of heart rate, blood pressure, respiratory rate, oxygen saturation, and ECG telemetry is routine in ICUs. Additional parameters include end-tidal CO ₂ , intracranial pressure, intracardiac pressures, and central venous pressure. Photoplethysmography is a non-invasive method for monitoring oxygen saturation using red and infrared light to detect changes in blood flow in tissues by relying on the principle that oxygenated hemoglobin absorbs more infrared light, while de-oxygenated hemoglobin absorbs more red light. Continuous glucose monitoring (CGM) via implantable sensors in dermal interstitial fluid has been available since 1999, and in 2020, the FDA authorized inpatient CGM during COVID-19, prompting its gradual ICU adoption. ^, End-tidal capnography measures CO ₂ in exhaled breath, indicating cardiac output, pulmonary flow, and ventilation. It is widely used for procedural sedation, confirming intubation, and during CPR to assess chest compressions and detect return of circulation. ^,

Wearable sensors are also used to prevent hospital-acquired pressure injuries. Sensors tracking patient turns increased repositioning compliance by over 50%. Wearable wrist/thigh accelerometers show promise in detecting delirium. A pilot study combined accelerometry, video, light, and sound data to distinguish delirious from non-delirious patients. Wrist accelerometer data showed clear differences in activity patterns. Despite promises, challenges remain: accuracy, lack of validation, algorithm explainability, and workflow integration.

Technological advances and device miniaturization brought ultrasound to the bedside, and its value for guiding procedures like central line placement, thoracentesis, and paracentesis became quickly clear. Portable and handheld devices have expanded point-of-care ultrasound (POCUS) from procedural use to a core element of critical care. Rapid adoption and integration into acute care training transformed bedside examinations into POCUS-guided assessments. Structured protocols like RUSH, FAST, and ACES emerged to guide early assessment in shock and trauma. POCUS is now key in evaluating shock, lung pathologies (eg, pneumothorax, edema), fluid responsiveness, and cardiac function. Training in POCUS is widespread in internal medicine, emergency medicine, and critical care programs. Critical Care Echocardiography board certification now reflects its growing importance.

POCUS continues to evolve with AI integration with potential to enhance both training and clinical care, including in resource-limited settings where it enables novice users to assess advanced parameters. ^,, New patch-style transducers can auto-capture and interpret images with minimal input, which may broaden access and improve consistency.

The advent of advanced organ support technologies has fundamentally transformed modern critical care. The early decades of the field were defined by the introduction of negative-pressure mechanical ventilators, which saved countless lives. A pivotal milestone came in the 1970s with the use of ECMO outside of the operating room, laying the groundwork for advanced respiratory and cardiovascular support. Over the past 2 decades, particularly since the 2008 influenza pandemic, there has been a marked increase in ECMO, specifically veno-venous (VV) ECMO, utilization for adults with severe ARDS VV-ECMO provides gas exchange by removing blood from the venous system, oxygenating it externally, and returning it to circulation. Landmark trials such as the CESAR and EOLIA shaped initial ECMO practices, which rapidly evolved during the COVID-19 pandemic as ICUs faced an unprecedented burden of severe respiratory failure. Veno-arterial ECMO (VA-ECMO), on the other hand, provides is used for cardiac support and is increasingly used as a bridge to recovery, decision-making, or for transition to ventricular assist devices (VADs) or cardiac transplant. A 2024 systematic review and meta-analysis found that VA ECMO significantly reduced all-cause mortality in cardiogenic shock (OR 0.31; 95% CI 0.11 to 0.86). Where isolated cardiac support is warranted, the use of mechanical circulatory support devices such as intra-aortic balloon pumps, Impella, and VADs has become the standard of care.

Renal support in the ICU is most commonly achieved with CRRT, which provides hemodynamic stability and precise control of solute and fluid balance in critically ill patients unable to tolerate conventional intermittent hemodialysis . CRRT includes modalities such as continuous veno-venous hemofiltration (CVVH), continuous veno-venous hemodialysis (CVVHD), and continuous veno-venous hemodiafiltration (CVVHDF), which differ based on the mechanism of solute clearance. CVVHD primarily uses diffusion and is most effective for removing small molecules, while CVVH relies on convection to clear both small and middle molecules. CVVHDF combines both diffusion and convection. Despite these mechanistic differences, clinical studies have not demonstrated a significant advantage of one CRRT modality over the others.

In critically ill patients with systemic inflammation (eg, sepsis, cytokine release syndrome), extracorporeal blood purification devices help remove cytokines and other mediators. Oxiris is a specialized hemofilter that combines renal replacement with cytokine and endotoxin removal via an adsorptive membrane. CytoSorb is a hemadsorption device using porous polymer beads to remove a wide range of cytokines. The Molecular Adsorbent Recirculating System is an artificial liver support using albumin dialysis to clear protein-bound toxins, bile acids, and cytokines. Other systems, like Prometheus and single-pass albumin dialysis, also aid in toxin removal. Their main role is bridging to recovery or transplantation and offer no mortality benefit.

Decision making in critical care is data-driven, time-sensitive, and complex. Modern monitoring and treatment technologies generate enormous volumes of real-time data, often too vast for human cognition alone. A 2008 retrospective cohort study found that the median number of documented clinical data items per patient in a single 24 hour period was 1,348, with even higher counts for those receiving advanced organ support such as hemodialysis and ECMO. With technological advances over the last 2 decades, this number is likely much higher. AI methods, particularly those using machine learning, can organize and analyze high-dimensional data in critical care. These approaches are capable of recognizing patterns in large, complex datasets, which may support early identification of clinical deterioration, prediction of complications, and prioritization of interventions.

While most current AI applications in critical care are predictive, focusing on forecasting adverse events, emerging research is shifting toward prescriptive analytics, where AI recommends specific interventions. A promising technique in this space is reinforcement learning (RL), in which algorithms learn optimal strategies by simulating the long-term impact of clinical decisions. For example, GLUCOSE, an RL model, was developed and externally validated to dynamically suggest personalized regular insulin dosing for patients on the first day after cardiac surgery. In a blinded evaluation, the GLUCOSE model was found to be at least as safe and effective as senior intensivists in recommending insulin doses. Similarly, Komorowski and colleagues developed a RL agent, the AI Clinician, that was trained to provide individualized and clinically interpretable treatment decisions for sepsis.

Despite its promise, most AI models in critical care fail to progress beyond the proof-of-concept stage. Several challenges impede the clinical adoption of AI tools. First, there are significant financial and operational costs associated with deploying AI systems in a real-world setting. Second, implementing models that function in real time within clinical workflows also requires significant technical infrastructure and rigorous oversight. Third, effective evaluation and maintenance of AI models necessitate structured environments, such as translational research laboratories, data science cores, and interdisciplinary clinical informatics teams to evaluate model performance, monitor for model drift, and update algorithms as clinical practices evolve. Finally, the regulatory and ethical landscape for AI in medicine is still evolving. Concerns around algorithmic bias, data privacy, explainability, and liability must be addressed through transparent development practices and robust governance frameworks. In a push for AI integration into health care systems, the United States federal government in July 2025 outlined a comprehensive 90-point AI Action Plan in July 2025, calling for increased funding, cross-agency collaboration, and ethical deployment of AI technologies across sectors while citing historical slow technological adoption in health care.

A substantial body of evidence demonstrates that tele-ICU programs are associated with significant reductions in both ICU and hospital mortality. For example, a large multicenter study found a 20% increase in patient survival following tele-ICU implementation, even after adjusting for patient severity and comorbidities. These findings are corroborated by meta-analyses, such as a systematic review by Wilcox and Adhikari (2012), which reported a 21% reduction in ICU mortality (Risk Ratio 0.79) and a 17% reduction in hospital mortality (Risk Ratio 0.83) across multiple before-and-after studies. These mortality benefits are most pronounced in hospitals with higher baseline mortality rates and are attributed to continuous remote intensivist oversight, rapid recognition and interventions for clinical deterioration, and improved adherence to evidence-based protocols.

Beyond mortality, tele-ICU implementation has also been associated with reductions in both ICU and hospital length of stay (LOS). Studies have shown average ICU LOS reductions, with some reporting decreases of up to 30% and others noting reductions of 1 to 2 days. These reductions are beneficial not only for patients, by decreasing their exposure to ICU-related complications, but also for the health care system, by improving hospital throughput and optimizing resource utilization. However, not all studies have found such benefits. The TELESCOPE study did not observe a significant difference in ICU LOS or other patient outcomes between tele-ICU and usual care, underscoring that the effectiveness of tele-ICU may be highly context-dependent and influenced by specific implementation models.

Tele-ICU programs were rapidly scaled during COVID-19 to manage surges, reduce exposure, and sustain high-quality care. It enabled remote consults, monitoring, and rounds; outcomes were comparable to in-person care. ^,

Patient safety is another critical domain where tele-ICU has demonstrated clear benefits. Continuous remote surveillance significantly reduces the risk of adverse events such as unplanned extubations, patient falls, and medication errors. The synergistic presence of both bedside and remote critical care teams creates an invaluable layer of redundancy in patient care, which effectively minimizes errors and reinforces adherence to vital safety protocols. Furthermore, tele-ICU systems facilitate rapid response to alarms and abnormal laboratory values, enabling earlier intervention and the prevention of escalation to critical events.

These sophisticated systems enforce higher adherence to evidence-based protocols for critical interventions, such as deep vein thrombosis prevention and cardiovascular protection. This adherence has been consistently associated with lower rates of preventable complications, including catheter-related bloodstream infections and ventilator-associated pneumonia Importantly, tele-ICU also plays a crucial role in promoting the standardization of care across diverse critical care units, thereby reducing variability in treatment and improving overall patient outcomes.