PURPOSE OF REVIEW:Hemodynamic monitoring has undergone a profound transformation over the last 30 years. The field has transitioned from the "standard of care" invasive pulmonary artery catheterization (PAC) toward minimally invasive and noninvasive technologies. This evolution is characterized by a shift in clinical philosophy. RECENT FINDINGS:A very strong shift away from the measurement of static pressures and flow defining specific hemodynamic profiles to the assessment of functional physiology and fluid responsiveness characterize the main changes in hemodynamic monitoring over this time interval. This review summarizes the four main milestones, the evidence-based "PAC-wars" and the "Great debate" (1995-2005); the shift to functional hemodynamics and more precise resuscitation (2005-2015), through the rise of goal-directed therapy (GDT) and protocolization based on patient subgroups (2010-2020); to the present where the focus is on noninvasive monitoring, using artificial intelligence (AI) to plumb biosignatures from time series data and focusing on tissue perfusion and end-organ function (2020-2025). SUMMARY:The use of hemodynamic monitoring to plumb bedside cardiopulmonary status and monitor responses to therapy with the goal of precise and personalized resuscitation focusing on end-organ recovery using advanced minimally-invasive and on-invasive monitors coupled with AI-aided physiological pattern recognition represents the pathway we are on.

Ejection fraction (EF) is widely used to assess cardiac function in sepsis, yet it is inherently load-dependent and may not reflect intrinsic contractility. We quantified the relative contribution of hemodynamic determinants to left (LVEF) and right ventricular EF (RVEF) in a porcine model of endotoxic shock. Twelve female pigs were instrumented with biventricular conductance catheters and studied at four stages: baseline, endotoxic shock, fluid resuscitation, and norepinephrine infusion. Six candidate determinants [effective arterial elastance (Ea), end-systolic elastance (Ees), end-diastolic volume (EDV), heart rate (HR), the time constant of isovolumic relaxation (τ), and internal flow fraction (IFF)] were evaluated using centering within clusters linear mixed models (LV: 12 animals; RV: 9 animals). Ea, EDV, and Ees were the dominant determinants of LVEF, accounting for 31%, 27%, and 19% of explained variance (combined: 77%), whereas HR exerted a negative effect. For RVEF, predictor importance was more evenly distributed, with EDV and Ea as the leading contributors in dominance analysis; HR was positively associated with RVEF, opposite to its LV effect. This loading-dominant hierarchy persisted when Ees was replaced by alternative load-independent indices: preload-recruitable stroke work (PRSW) and Starling contractility index (SCI). Despite substantial stage-to-stage changes in preload, afterload, and contractility, EF varied within a narrow range (LVEF: 37%-45%; RVEF: 50%-60%), reflecting offsetting changes among determinants. Within this pressure-volume framework, loading conditions rather than contractility were the primary within-animal determinants of LVEF during endotoxic shock. The RV data were consistent with a similar hierarchy, although with differing determinant profiles and greater analytical uncertainty.NEW & NOTEWORTHY This study provides a simultaneous quantification of six hemodynamic determinants of both left and right ventricular ejection fraction during endotoxic shock and resuscitation using biventricular pressure-volume analysis. Loading conditions dominated over contractility, and the determinant hierarchy differed between ventricles.

Central to managing critically ill patients is the identification of the etiology of cardiorespiratory insufficiency (i.e., shock), early appropriate targeted therapies to support the cardiorespiratory system to sustain adequate blood flow and oxygen to the tissues, plus specific treatments to reverse the cause of shock. Over the past 40 years, numerous advances in our understanding of shock, its severity, and its response to therapies, along with more specific and insightful monitoring approaches, have been developed. This perspective summarizes some aspects of that progress. We have come a long way, but we need to understand three things. First, that once organ injury has occurred all that our treatments can do is mitigate further injury, not reverse it. If initial aggressive resurrection efforts cannot restore organ function, then their actions often cause only iatrogenic injury. Second, existing advanced monitoring devices, no matter how insightful their data, will not improve patient outcomes unless coupled to a treatment that itself improves outcomes. Finally, all our advances over these years have underscored the fundamental need for having a thoughtful and observant bedside clinician cognizant of the pathophysiologic underpinnings of disease and its care who titrates care based on the patient's individual response.

Currently used sepsis severity indices rely on fixed variables and weights established decades ago, which are coarsely discretized and calibrated to a cohort that no longer reflects contemporary critical care. No alternative learned directly from patient trajectories is in routine use. We conducted a retrospective two-cohort study on a total of 29,116 and 7,691 adult patients meeting Sepsis-3 criteria from two hospital systems in Massachusetts and Georgie, respectively. We developed a sepsis index using 43 routinely charted variables over a 72-hour treatment window. Unlike previous studies, we use mortality as a treatment-level ranking signal rather than a per-state target, allowing credit to be redistributed non-uniformly across timesteps. Evaluation was done on a permanent 20 Our index demonstrated hourly prognostic information that meaningfully separates patient outcomes and is consistent with clinical expectation, indicating potential as a decision support tool complementing clinical judgement.

Tissue autoregulation to match local blood flow to metabolic demands requires both a high enough upstream mean arterial pressure (MAP) and perfusion pressure difference between this MAP and the downstream arteriolar critical closing pressure to drive blood flow into the capillaries whose downstream capillary pressure is approximated as mean systemic filling pressure. The closing pressure to mean systemic pressure difference represents a vascular waterfall such that increases or decreases in mean systemic filling pressure below closing pressure do not alter tissue flow. Tissues autoregulate their blood flow by altering upstream local vasomotor tone to increase or decrease closing pressure. In vasoplegia, like septic shock vascular tone is decreased, often decreasing closing pressure to approximating mean systemic filling pressure. Such conditions abolish autoregulation even if MAP and cardiac output are not decreased. Effective resuscitation from septic shock requires restoration of local vascular waterfalls. Often initial fluid resuscitation and vasopressor infusion restore vascular waterfalls and tissue blood flow. But often it does not. There are few readily available real-time quantitative estimates of tissue perfusion. Presently only capillary refill time (CRT) minors tissue blood flow changes. If initial resuscitation to target MAP values does not restore tissue perfusion and CRT is > 3 seconds, then potentially a vasopressor test to increase MAP to > 75 mmHg may be studied. If vasopressor-induced increases in MAP decreases CRT to < 3 seconds, then it can be continued. If not, then return to prior levels to minimize iatrogenic vasopressor risk. This paper lists unanswered questions that need studying.

Sepsis heterogeneity reflects diverse etiologies and patient-specific physiological responses, motivating phenotype identification to enable precision therapeutics. However, most phenotyping approaches rely on intermittently sampled clinical variables, whereas continuously recorded physiological waveforms remain underutilized. We developed a deep-learning framework to derive physiological phenotypes from five-minute pre-onset electrocardiogram, photoplethysmogram and respiratory-impedance waveforms in 2,174 ICU patients meeting Sepsis-3 criteria. From these signals, 192 cardiorespiratory physiomarkers were extracted and embedded using a Feature Tokenizer Transformer encoder, which outperformed alternative representation methods. Consensus clustering identified four stable sepsis physio-phenotypes (SP-1-SP-4) associated with distinct autonomic and peripheral vascular signatures. Despite similar baseline severity and demographics, phenotypes differed significantly in mortality (19-29%), septic shock, vasopressor use and mechanical ventilation, with divergent 28-day survival trajectories (P<0.01). Explainable AI provided clinically interpretable characterizations, and a trained classifier enabled real-time bedside phenotyping. This framework establishes waveform-based phenotyping as a foundation for precision medicine in sepsis care.

Bedside physiologic monitor alarms negatively contribute to clinician cognitive burden and patient wellbeing in hospital intermediate (IMCU) and intensive (ICU) care units. This was a cross-sectional study to characterize the types, sources, annunciation modality and prevalence of alarms generated by bedside patient monitors, as well as their distributions and hourly alarm load across 3 ICUs and 1 IMCU. Analyzing > 2 million monitoring hours from 17,442 patient encounters (59 ± 18 years, 14% Black, 43% female) over 5 years, we identified 65.6 million alarms. Of all alarms, 74% were silent while 26% were audible. Most alarms (88%) were technical, unrelated to patient condition, while 12% were physiologic, suggesting clinical instability. Among audible alarms, 68% arose from technical issues. The IMCU averaged 5 audible alarms per patient-hour, of which 67% were technical. ICUs experienced up to 10 audible alarms per patient-hour, with 54% of them technical. This comprehensive study of monitoring alerts shows that alarms in IMCUs and ICUs are predominantly technical, highlighting opportunities to refine monitoring technologies and management strategies. Such improvements would help clinicians focus on alarms signaling true changes in patient physiologic status and reduce the cognitive burden of frequent alarms, which can contribute to alarm fatigue and worsen patient outcomes.

The vascular waterfall (VW) concept describes a key mechanism of blood flow autoregulation, linking arterial critical closing pressure (Pcrit) to mean systemic filling pressure (Pmsf). This pressure gradient defines tissue perfusion pressure (Pa - Pcrit) and supports a framework for interpreting circulatory dynamics and blood flow autoregulation in health and disease. This review explores the historical development, physiological foundations, clinical implications, and future directions of the VW model in critical care. Originating from Starling's resistor model, the VW arises when surrounding pressure, whether from vasomotor tone or interstitial forces, exceeds intraluminal arterial pressure, defining a Pcrit, causing flow to become independent of downstream pressure. In such conditions, Pcrit functions as the effective backpressure. The highly compliant capillary bed accommodates low inflow pressures, often 5-10 mmHg above Pmsf. Bedside techniques, such as inspiratory hold maneuvers, enable estimation of Pcrit and Pmsf as global circulatory parameters. In vasoplegia, such as sepsis, reduced vascular tone lowers Pcrit, often approaching Pmsf, which abolishes the VW and impairs autoregulation. Vasopressors may restore perfusion only if Pcrit increases more than Pmsf. Similarly, in pulmonary circulation, VW physiology helps explain how elevated alveolar and pleural pressures during mechanical ventilation, particularly with high PEEP, affect venous return and right ventricular load, with implications for ARDS management. Looking forward, non-invasive measurement of Pcrit, integration with continuous hemodynamic monitoring, and predictive modeling may enable real-time VW-guided therapy. By bridging macro- and microcirculatory physiology, the VW concept offers a promising basis for precision hemodynamic interventions in critically ill patients.

Twenty-five years of septic shock resuscitation trials have progressively transformed the physiological understanding of shock and the way bedside interventions are conceptualized. Early strategies focused on correcting isolated global hemodynamic or metabolic targets through protocolized interventions, whereas more recent approaches increasingly recognize septic shock as a heterogeneous, dynamic, and context-dependent condition requiring individualized physiological assessment. This perspective discusses how hemodynamic resuscitation trials during this quarter of a century have reshaped both the physiological understanding and methodological foundations of early shock management. Initial strategies, exemplified by early goal-directed therapy, aimed to correct global oxygen-derived variables through protocolized interventions designed to optimize oxygen delivery. However, the lack of reproducibility of this approach in subsequent multicenter trials, together with the recognition of the limitations of several other variables as potential resuscitation targets, revealed that septic shock cannot be adequately addressed through rigid algorithms based on isolated endpoints. Subsequent progress incorporated peripheral perfusion assessment, systematic evaluation of fluid responsiveness, critical care echocardiography, and hemodynamic phenotyping. Within this trajectory, ANDROMEDA-SHOCK shifted attention toward capillary refill time as a rapidly responsive clinical perfusion signal, contributing to improve some outcomes, and potentially limiting over-resuscitation. ANDROMEDA-SHOCK-2 further operationalized a personalized strategy based on capillary refill time, sequential phenotyping, reversible hemodynamic tests, and serial reassessment. A quarter of a century of septic shock resuscitation trials reveal a consistent pattern: strategies centered on fixed and isolated physiological targets have repeatedly failed to achieve reproducible improvements in patient-centered outcomes. Recent data support a shift from isolated hemodynamic or metabolic targets, toward a physiology-guided, phenotype-driven resuscitation strategy, in which interventions are applied as reversible tests within short decision cycles and guided by rapidly responsive perfusion signals. Trials such as ANDROMEDA-SHOCK-2 represent important steps in this ongoing transition, but further refinement, validation, and implementation across diverse settings remain necessary.

RATIONALE:Resource-limited or austere environments represent a direct threat to the likelihood of survival of patients in need of emergent care. Medical autonomous care (MAC) offers the promise to provide solutions to these challenges, change delivery of care, and improve patient outcomes. OBJECTIVE:To summarize the discussions surrounding the role of medical autonomous care in healthcare delivery held during the Defense Health Agency (DHA) funded Medical Autonomous Care State of the Science meeting (MAC-SOS): Opportunities and Obstacles in September 16-17 of 2025 in Pittsburgh. METHODS:The MAC-SOS meeting was designed by a select scientific committee including stakeholders in all aspects of the development, ethics, regulatory pathways, marketing, clinical implementation, and surveillance of medical autonomous care technology, and was composed of seven plenary sessions and six breakout sessions for focused discussion. MEASUREMENTS AND MAIN RESULTS:230 attendees and 44 experts from diverse backgrounds including academia, the United States Department of War (DoW), industry and regulatory agencies were in attendance. Experts and attendees engaged in two days of discussions on the opportunities and obstacles framing the research, development, ethics, regulatory approval pathways, marketing, implementation, and surveillance related to medical autonomous technologies. CONCLUSIONS:The MAC-SOS meeting provided a forum for dialogue, sharing of ideas, and discussions on key challenges, obstacles, and opportunities that medical autonomous care technology faces from inception to clinical deployment and post-implementation surveillance. It also served as a primer for the creation of a sustainable alliance between all stakeholders, and a springboard for novel ideas and future collaborations.

OBJECTIVE:This European Society of Intensive Care Medicine (ESICM) guideline provides evidence-based recommendations on the volume of early resuscitation fluid for adult critically ill patients. METHODS:An international panel of experts developed the guideline, focusing on fluid resuscitation volume in adult critically ill patients with circulatory failure. Using the PICO format, questions were formulated, and the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach was applied to assess evidence and formulate recommendations. RESULTS:In adults with sepsis or septic shock, the guideline suggests administering up to 30 ml/kg of intravenous crystalloids in the initial phase, with adjustments based on clinical context and frequent reassessments (very low certainty of evidence). We suggest using an individualized approach in the optimization phase (very low certainty of evidence). No recommendation could be made for or against restrictive or liberal fluid strategies in the optimization phase (moderate certainty of no effect). For hemorrhagic shock, a restrictive fluid strategy is suggested after blunt trauma (moderate certainty) and penetrating trauma (low certainty), with fluid administration for non-traumatic hemorrhagic shock guided by hemodynamic and biochemical parameters (ungraded best practice). For circulatory failure due to left-sided cardiogenic shock, fluid resuscitation as the primary treatment is not recommended. Fluids should be administered cautiously for cardiac tamponade until definitive treatment and guided by surrogate markers of right heart congestion in acute pulmonary embolism (ungraded best practice). No recommendation could be made for circulatory failure associated with acute respiratory distress syndrome. CONCLUSIONS:The panel made four conditional recommendations and four ungraded best practice statements. No recommendations were made for two questions. Knowledge gaps were identified, and suggestions for future research were provided.