- Retrospective study of 2,356 CABG patients from the MIMIC-IV 2.2 database, defining prolonged ventilation as more than 24 hours.
- Five machine learning algorithms were compared using 40 baseline features and 5 composite metrics, with SHAP used to interpret the top model.
- The random forest model performed best, with AUROC 0.9976 and AUPRC 0.9976, plus high accuracy, sensitivity and specificity.
- Delirium, oxygen saturation, glycemic variability and SOFA score contributed most, and the authors name delirium and glucose fluctuations as key predictors.
Critical care
Research from anesthesiology, pain and critical care journals, sourced from PubMed.
Sorted by publication date, not import date. Recent filters exclude papers with only a publication year; find those under All years.
AI-generated abstract summaries can contain errors. Verify findings against the original paper before applying them to patient care.
Last successful paper check: 16 Nov 2026 17:18 UTC.
0 newly discovered papers saved; these are not necessarily published today.
2 papers awaiting AI summaries. Summary processing had an error; paper-fetch status is separate.
Background checks run while the server is active. This does not guarantee a daily run if hosting sleeps.
2 papers awaiting AI summaries. Summary processing had an error; paper-fetch status is separate.
Background checks run while the server is active. This does not guarantee a daily run if hosting sleeps.
- Qualitative systematic review of 34 ethics articles (commentaries, opinion pieces, academic articles) on slow codes in U.S. hospitals, drawn from 12,863 screened records.
- The exposure is the "slow code," in which clinicians appear to give advanced cardiac life support without fully committing to resuscitation.
- Of the 34 articles, 61.8% (n = 21) argued physicians should be allowed to refuse CPR in futile cases, 23.5% (n = 8) supported slow codes when families request full resuscitation, and 26.5% (n = 9) took nuanced positions.
- The authors conclude most ethics literature deems slow codes impermissible, while physicians remain divided in practice, and call for clearer communication and ethical guidance.
- Prospective, single-center, within-patient crossover pilot of 52 adults undergoing radiofrequency AF ablation under general anesthesia.
- Compared high-frequency very low tidal volume ventilation with peak pressure limit (50 breaths/min, I:E 1:1, peak pressure ≤20 cmH2O) against standard ventilation.
- Catheter displacement fell from 6.22 ± 2.18 mm to 1.23 ± 0.69 mm (difference -4.99 mm, p < 0.001), with similar contact force.
- Partial pressure of carbon dioxide rose from 38.0 to 49.9 mmHg, exceeding 60 mmHg in 9.6% of patients; the authors call the findings a mechanistic signal.
- This Special Article builds a framework by synthesizing aggregated study-level data from pediatric cardiac surgery populations across multiple institutions, with no clinical intervention performed.
- Candidate variables linked to extubation timing and outcomes were grouped into domains based on recurrence in the evidence, biological plausibility, and fit with routine multidisciplinary assessment.
- The resulting PCERF has two layers: readiness factors (age, weight, procedural complexity, cardiopulmonary stability, pathway factors) and non-scored safety modifiers that may preclude early extubation.
- The authors describe the score structure, weights, and readiness bands as illustrative prototypes, not outcome-derived, and call the framework hypothesis-generating rather than a validated prediction model.
- This expert-driven, evidence-based consensus review covers simulation-based education and training across anaesthesiology, intensive care, and perioperative medicine, with no primary study population.
- It examines SBET across the professional lifespan, team performance, latent system threats, translational simulation, and digital and ecological competencies.
- It identifies barriers to adoption, including fragmented uptake, limited interprofessional integration, and few national frameworks, and does not report pooled numerical results.
- The authors call for coordinated, value-based adoption of SBET as strategic infrastructure, supported by sustained investment, protected time, faculty development, and interprofessional integration.
- Secondary analysis of a retrospective Dutch cohort of 18,798 adults directly admitted from the ED to the ICU across 12 hospitals, 2009-2020, in seven diagnostic groups.
- Logistic regression compared ED-to-ICU time quintiles with hospital mortality, adjusting for hospital and APACHE III, and was stratified by hospital type.
- In OHCA at academic hospitals, longer transfer times were linked to higher mortality (OR 2.94, 95% CI 1.80-4.78 above 3.4 hours), while NACT hospitals showed ORs below 1.0.
- A trauma association appeared only in the overall cohort (OR 1.90, 95% CI 1.12-3.21) and was not reproduced after hospital stratification; other diagnoses showed no association.
- Observational study of ChatGPT-4o answering 183 Society of Critical Care Medicine board-style multimodal questions, with responses reviewed by 14 clinical reviewers in a simulated exam setting.
- The model was run through a custom profile built on a validated framework, and its answers were compared with pooled clinician responses; there was no intervention arm.
- ChatGPT-4o answered 74.9% correctly versus 71.1% for pooled clinicians (p = 0.03), but image interpretation was 61.7% and critical care ultrasound 51.1%.
- The authors conclude the model is slightly above clinician benchmarks yet limited in image analysis, and 33.3% of its responses carried potential for clinical harm.
- Post hoc analysis of 70 of 110 severe TBI patients from the BOOST II randomized trial, across ten U.S. ICUs.
- Background electronic ICP data were compared with bedside-identified minimalist episodes (60 minutes or less, tier 1 treatments only), defined as 5 minutes at 20 mm Hg or higher.
- Electronic data confirmed only 47% of 509 bedside-identified episodes, and 76% of the 286 discordant episodes had average electronic ICP below 20 mm Hg.
- The authors conclude bedside identification may not reliably distinguish sustained ICP episodes and propose open-source, real-time, temporally synchronized electronic definitions.
- Retrospective single-center cohort of 30 critically ill adults with acute or acute-on-chronic liver failure who received MARS and piperacillin-tazobactam, with 121 measurements across 77 sessions.
- Piperacillin concentrations were measured immediately before and after each 8-hour MARS session, and the abstract describes no separate comparator group.
- The mean relative reduction per session was 34% (±7.9%; p < 0.05), with higher creatinine clearance linked to greater drug loss and higher bilirubin to less.
- The authors report that MARS was associated with about a one-third reduction per session and support systematic drug monitoring and individualized dosing, though 31.6% of post-session levels fell below 80 mg/L.
- This descriptive report presents an open, interoperable peri-operative data platform recording from 77 OR and ICU beds since March 2019.
- The platform streams data from monitors, syringe pumps, ventilators, EEG systems and lab values using Apache Kafka, with on-premise, anonymised storage meant to comply with GDPR.
- Recorded data cover 118,915 patients, including 56,212 with ventilation data, 78,091 with advanced ECG analysis, 9,862 with EEG and 3,458 with cardiac output monitoring.
- The authors conclude the work demonstrates feasibility of a scalable, vendor-neutral infrastructure and offers a reproducible blueprint, while noting that data quality standardisation and volume challenges remain.
- A single-day cross-sectional survey of 1058 physicians and nurses across 47 Dutch ICUs, with follow-up data for 215 patients.
- About a quarter (26%) of clinicians reported perceived inappropriateness of care for at least one patient, most often citing distributive injustice (70%).
- Being a nurse (aOR 1.78) and working in a culture that avoids end-of-life decisions (aOR 1.92) carried higher PIC risk, while mutual respect within teams was protective (aOR 0.84, 95% CI 0.70-1.00).
- PIC reported by multiple clinicians was independently associated with death (aOR 3.86; 95% CI 1.23-12.16; p=0.02) and with the combined adverse outcome (aOR 2.91; 95% CI 1.14-7.42; p=0.03).
- Retrospective population-based cohort of 1251 adult ICU patients who had pleural fluid cultured across eight ICUs in one Danish region over 10 years.
- Exposures were suspected pleural infection and parapneumonic effusion, with the association of each assessed against the risk of a positive first pleural culture.
- Only 51 patients (4.1%) had a positive first pleural culture; suspected infection was linked to higher risk (adjusted RR 4.88, 95% CI 2.70-8.83), while parapneumonic effusion was not (adjusted RR 1.41, 95% CI 0.83-2.38).
- A positive culture was associated with higher 90-day mortality (adjusted RR 1.35, 95% CI 1.04-1.73), and the authors suggest culturing may be most useful when infection is suspected.
- Systematic review and meta-analysis of 20 studies including nearly 800,000 adult cardiac surgery patients.
- Compared ultra-fast-track extubation (in the OR or within 1 hour) with fast-track extubation (within 6 hours in the ICU).
- UFTE was linked to lower 30-day mortality (OR 0.54), shorter ICU stay (MD -12.27 hours) and shorter hospital stay (MD -1.19 days).
- In propensity-matched cohorts, mortality and reintubation were comparable, though shorter stays persisted, and the authors call for randomized trials.
- Bicentric prospective registry study of 442 comatose adults after cardiac arrest from Lausanne and Paris (2020–2024), with pupillometry in the first 48–72 hours.
- Compares quantitative pupillary light reflex measures (NPi, minimum and maximum qPLR) with EEG, SSEPs, serum NSE, and clinical features; no intervention was given.
- Minimum qPLR below 2% and NPi ≤2 showed high specificity for poor outcome (96.4% and 99%) but low sensitivity (23.6% and 34%).
- Qualitative PLR was more sensitive than quantitative values (38.9% vs 23.6% and 19.0%) with similar specificity, and high qPLR predicted good outcome only modestly.
- Prospective study in two French university ICUs of 34 critically ill patients on pressure support ventilation; 14 (41%) were preload responders.
- Investigators tested whether PLR-induced changes in PPV or PP predict preload responsiveness, defined as a ≥12% rise in subaortic VTI, with patients split by median P0.1.
- PLR-induced change in PPV was associated with responsiveness (AUROC 0.76; p = 0.026; cutoff −3%), while baseline PPV and PLR-induced change in PP were not.
- In patients with P0.1 below 2.3 cm H2O, the PLR-induced PPV change had an AUROC of 0.90 (p = 0.050), versus 0.61 (p = 0.480) with higher effort.
- A retrospective tertiary-care ICU cohort of 2,478 adults with suspected infection in China tested a temperature-trajectory sepsis model first developed in U.S. centers.
- Investigators compared longitudinal immune and coagulation profiles across four subphenotypes and assessed immunoglobulin effects on 30-day mortality after propensity score matching.
- The hypothermic subphenotype (734 patients, 30%) had the highest 30-day mortality at 25% and persistent lymphopenia, low monocyte HLA-DR, and coagulopathy.
- In the hypothermic subphenotype, immunoglobulin was associated with a hazard ratio of 0.47 (p = 0.03) under primary subtyping, and the authors report heterogeneous responses across subphenotypes.
- Cross-sectional questionnaire survey of 338 ICU physicians and nurses at three Swiss centers, mostly nurses (78.1%) and women (74%).
- No intervention was given; the survey examined associations between brain death understanding and personal, educational, and professional factors.
- Self-perceived understanding and approval were high (median 9, IQR 8-10), but physicians showed greater approval than nurses (p = 0.003).
- Understanding rose with ICU experience (ρ = 0.194) and age (ρ = 0.213), and with prior brain death exposure, while sex and religion showed no significant association.
- This retrospective cohort included 22,156 ventilated adults with severe TBI who underwent tracheostomy at 417 ACS TQIP trauma centers from 2016 to 2021.
- Hospitals were grouped as low, medium, or high tendency for tracheostomy within 7 days of injury, based on risk-adjusted odds ratios, and this hospital-level tendency was the exposure.
- High-tendency hospitals had higher odds of withdrawal of life-sustaining treatment after tracheostomy (AOR 1.35; 95% CI 1.10-1.66; p = 0.004).
- The largest point estimate was in patients aged 40-59 (AOR 1.39; 95% CI 1.01-1.91), and the authors frame the findings as practice variability under prognostic uncertainty.
- Single-center retrospective target trial emulation of 4,635 adults on home beta-blockers admitted from the ED with suspected infection.
- Compared oral beta-blocker continuation within 48 hours of admission with no continuation, using inverse probability weights for baseline differences.
- Continuation was associated with lower 90-day all-cause mortality (HR 0.77, 95% CI 0.61-0.98, p=0.03) and shorter hospital stay (IRR 0.39, 95% CI 0.38-0.41).
- In-hospital mortality did not differ significantly (OR 0.60, 95% CI 0.30-1.20, p=0.15), and the authors call for controlled experimental studies.
- Animal study in C57BL6 mice and Yucatan mini swine, plus in vitro neutrophil assays using plasma from TBI patients and healthy volunteers.
- Animals received either sham anesthesia or mild TBI, then lung bacterial inoculation with S. aureus (mice) or A. pleuropneumoniae (pigs).
- TBI animals showed poorer lung bacterial clearance despite higher neutrophil counts in bronchoalveolar lavage and blood; the abstract gives no effect sizes or p values.
- In vitro, TBI patient plasma and extracellular vesicles reduced neutrophil reactive oxygen species production and phagocytosis, suggesting circulating factors contribute to the dysfunction.
- Substudy of three prospective cohort studies at one tertiary academic center, including 676 adults with respiratory failure and/or shock discharged alive from the ICU.
- Exposure was drug burden index (DBI) measured at admission, ICU transfer, and hospital discharge; no intervention was assigned.
- Median DBI rose from 1.96 at admission to 3.08 at discharge, but discharge DBI was not significantly associated with cognitive impairment (OR 1.25; 95% CI 0.89-1.76) or 90-day mortality (HR 0.87; 95% CI 0.61-1.23).
- No significant association was found with Katz ADL or FAQ disability at up to 6 months; the authors concluded that centrally acting drug burden was not linked to these long-term outcomes in this cohort.
- Retrospective registry cohort of 10,210 adult ICU patients admitted with pulmonary embolism across Dutch ICUs from 2013 to 2023.
- Compares high-risk with non-high-risk pulmonary embolism patients by risk category; no intervention was assigned.
- In-hospital mortality was 25.4% in high-risk versus 2.8% in non-high-risk patients, and 1-year HR was 3.98 (CI 3.59-4.40).
- Among hospital survivors, 1-year mortality remained higher in high-risk patients (HR 1.70; CI 1.49-1.96).
- This systematic review and meta-analysis pooled 32 studies of 3142 patients, including six diagnostic, 22 prognostic, and four studies reporting both.
- Diagnostic studies were not pooled, but individual studies suggested sensitivity of 78-95% and specificity of 80-90% for detecting raised central venous and right atrial pressure.
- In cardiac patients, VExUS of at least 2 was associated with acute kidney injury (OR 4.44, 95% CI 2.34-8.43) and mortality (OR 3.17, 95% CI 1.30-7.75), while in critically ill patients no association was found for either outcome.
- The authors conclude VExUS has moderate to good accuracy for raised central venous and right atrial pressure, and that its link to outcomes is clearer in cardiac than in critically ill patients; per-patient VExUS 3 was associated with mortality across all patients.
- This narrative review synthesizes RCTs, observational studies, physiologic investigations and prior reviews on perfusion monitoring in hemodynamically unstable patients.
- It examines tools including capillary refill time, peripheral perfusion indices, lactate, venous oxygenation, tissue oxygenation, spectral imaging and advanced hemodynamic assessment.
- It proposes a stepwise, physiology-driven framework for proportional escalation of monitoring, rather than normalizing isolated variables.
- The authors conclude that perfusion should be assessed as an integrated process, with capillary refill time as a pragmatic anchor and escalation guided by risk and uncertainty.
- This is a descriptive online survey of 570 respondents from 28 European countries about end-of-life decisions when extracorporeal life support (ECLS) goals are no longer achievable.
- No intervention was studied; the survey compared reported discontinuation practices and ethics and palliative care support across European regions and professional backgrounds.
- 41% reported discontinuing ECLS with comfort care, with 54% and 59% in Northern and Western Europe, while 24% of institutions lacked any ethics support.
- The authors conclude that regional and cultural variation is substantial and that structured interventions are needed to improve consistency and reduce moral distress.
- A prospective observational study over 12 months in an eight-bed brain-dead donor care ICU, covering 182 organ donors.
- All respiratory therapy, including full ventilator management, was delivered remotely through a telecritical platform, with in-person RT available as needed; there was no comparator group.
- Remote care completed 3872 procedures over 1782 hours, and in-person support was needed for 6.1% of those hours; no airway losses, cardiac arrests, or care delays were recorded.
- The authors estimate 2.2 FTEs and $306,952 in labor savings, with 520 organs procured at an observed-to-expected ratio of 1.19.
- Retrospective multicohort study of 15,732 adults with septic shock meeting CDC Adult Sepsis Event and Sepsis-3 criteria, drawn from a large U.S. health system during 2012 to 2024.
- Refractory septic shock was defined by coexisting norepinephrine (base) equivalent above 0.5 µg/kg/min and lactate above 2 mmol/L, compared against patients not meeting these criteria.
- Refractory septic shock occurred in 3,423 patients (21.8%), with hospital mortality of 64.4% and risk-adjusted mortality odds 4.87-fold higher (95% CI, 4.46-5.31).
- The authors conclude that about one in five patients with septic shock met these criteria and that they were associated with substantially higher mortality.
- Systematic review of 23 clinical studies, plus a retrospective cohort of 215,474 adult ICU patients in Japan's multicenter claims database (2010–2021).
- The exposure was the varying PICS definitions extracted from the review, applied to the same cohort; no intervention was studied.
- Patient counts per definition ranged from 211 to 47,888, prevalence from 0.1% to 22.2%, and in-hospital mortality from 11.4% to 64.0%.
- Despite this heterogeneity, PICS consistently identified patients with prolonged hospital stay and physical impairment, and the authors call for standardized criteria.
- Retrospective bicentric cohort of 5513 adult critically ill patients with moderate-to-severe TBI from two French level 1 trauma centers, 2010 to 2025.
- Investigators examined whether the link between admission hemoglobin and ICU mortality differed by sex; no interventions were assigned.
- Each 1 g/dL higher admission hemoglobin tracked with lower ICU mortality odds (OR 0.69; 97.5% CI 0.52–0.91; p=0.0027), with no significant sex interaction (OR 1.04; 97.5% CI 0.91–1.19; p=0.433).
- Causal analysis likewise showed no sex difference at equivalent hemoglobin, and the authors do not support sex-differentiated admission transfusion thresholds.
- Secondary analysis of the PROPPR trial, including 286 severely injured patients with complete data out of 680 enrolled across 12 level 1 centers.
- Early emergency department blood draws were analyzed for 48 coagulation and inflammatory biomarkers, grouped into phenotypes using principal component analysis; no intervention was given.
- The 30-day mortality rate was 24.4% (n = 70); after adjustment, the coagulopathy phenotype was associated with mortality (p = 0.0002), and mixed inflammatory phenotype 3 showed a trend (p = 0.0684).
- The authors conclude that coagulation and inflammation are coactivated early after severe injury and are linked to poor outcomes.
- Descriptive software development and validation study using 50 simulated scenarios derived from real cardiac surgical cases, reviewed by three senior cardiac anesthesiologists.
- The authors built a web-based app (CCA Pro v6.0) combining six modules: coagulation, blood gas, ventilation, hemodynamics, ECMO, and AI extraction of values from device photos.
- The app reportedly agreed with expert consensus across coagulation, acid-base, ventilator, and ECMO scenarios, but the abstract reports no quantitative agreement statistics.
- The authors suggest it could reduce cognitive load and dosing errors, and state that prospective clinical validation is warranted.
- This subgroup analysis of the MINT trial included 3504 adults with acute myocardial infarction and anemia across 144 sites, of whom 1679 (47.9%) were critically ill.
- Patients were randomized to a restrictive strategy (transfuse at hemoglobin below 7-8 g/dL) or a liberal strategy (below 10 g/dL), with 30-day outcomes compared by ICU status.
- The restrictive strategy showed nonsignificant increases in death in critically ill (RR 1.24; 95% CI 0.95-1.61) and noncritically ill patients (RR 1.09; 95% CI 0.77-1.54); interaction p = 0.55.
- Death and death/MI were more common in critically ill patients at 30 days, and the authors found no evidence that the overall liberal-favoring MINT results differ by ICU status.
- A UK survey of cardiothoracic anesthesiologists, sent via ACTACC email and departmental contacts, had 153 responses from 461 invited (33%).
- The survey asked about TEE accreditation status, pathways, and motivations and barriers for obtaining or maintaining it; no intervention was given.
- Fifty-three percent of respondents held valid TEE certification, and 80% considered maintaining accreditation valuable; the European pathway was the most popular (44%).
- The authors conclude current systems do not appear to meet UK cardiac anesthesia workforce needs and call for national guidelines and a tailored perioperative TEE curriculum.
- This is a critical narrative review of clinical controversies, not a primary study, and it reports no pooled sample size or quantitative results.
- It explains mechanical power as a single energy-per-unit-time measure that combines tidal volume, pressure, and flow, in contrast to static parameters such as tidal volume and PEEP.
- The authors conclude that current evidence is insufficient to show that mechanical power is independently associated with postoperative pulmonary complications.
- They suggest it may be a reasonable supplement to static parameters in high-risk patients, with excess power curbed by adjusting flow and respiratory rate, and call for standardized studies.
- Randomized controlled trial in anesthesia residents tested rested and after a 24-hour shift; 35 enrolled, 27 in the primary analysis.
- Sleep-deprived residents were randomized to a 30-minute nap opportunity or control, with actigraphy measuring sleep and nap duration.
- Simulated clinical performance was 14.8 points higher after napping (95% CI 2.8 to 26.9; P = 0.018), while technical skills did not differ significantly.
- The authors conclude napping appears to enhance performance, and exploratory analyses linked longer nap duration to several performance domains.
- This cross-sectional study of 100 invasively ventilated neonates included infants with RDS, neonatal ARDS, or evolving BPD, plus 28 controls without lung disease.
- Mechanical power and energy were calculated with four equations using simultaneously collected ventilation, oxygenation, and ultrasound lung aeration data.
- Power and energy were higher in infants with respiratory failure than in controls, with power differences of 0.21 to 0.9 J/min/kg (P<0.001).
- Power correlated with oxygenation (adjusted ρ 0.18 to 0.22) and with lung aeration impairment (ρ 0.25 to 0.27).
- This narrative expert review surveys machine learning applications across the postcardiotomy shock course, from early risk recognition through tMCS prognostication.
- The review covers ML use in escalation, device configuration, and weaning of temporary mechanical circulatory support, along with model development steps such as feature selection and external validation.
- Evidence is strongest for risk stratification, mortality prediction, and weaning support, mainly in venoarterial ECMO cohorts, but most models are retrospective and lack prospective PCS-specific validation.
- Automated device titration is promising but still preclinical, and the authors position ML as an adjunct to clinician judgment rather than a replacement.
- Prospective cohort of 214 postoperative ICU patients (77 nonseptic, 61 sepsis, 76 septic shock), with blood sampled within 24 hours of critical illness onset.
- Circulating endothelial subsets were profiled by high-dimensional flow cytometry with unsupervised clustering, and delirium was assessed daily with CAM-ICU.
- The CD32b+ subset was independently associated with 28-day delirium (HR 2.41, 95% CI 1.32 to 4.40; P = .004), with AUC 0.79 rising to 0.89 after age and sex adjustment.
- About 20% of the total effect was mediated through organ dysfunction events; the authors conclude the elevated subset supports an association with delirium and organ dysfunction.
- Retrospective three-phase analysis of 380 elective isolated CABG patients from 2024: pre-ERAS (n=238), ERAS without monitoring (n=119), and ERAS with FloTrac monitoring (n=23).
- Compared ERAS with FloTrac-guided hemodynamic monitoring against pre-ERAS care and ERAS without monitoring, with extubation time and ICU length of stay as primary outcomes.
- Mean extubation time fell from 10 hours 17 minutes to 5 hours 54 minutes in the FloTrac phase, a 43% reduction (95% CI -6.26 to -2.50 hours; p < 0.001). Mean hypotension duration fell 74% (p < 0.001).
- The authors attribute the gains to protocol adherence, workflow changes, and real-time feedback rather than FloTrac alone. ICU length of stay did not differ significantly from pre-ERAS (p = 0.472).
- Single-center retrospective cohort of 408 adults undergoing elective cardiovascular surgery (2022–2023), comparing a fixed postoperative dexmedetomidine infusion (n = 188) with no dexmedetomidine (n = 220).
- Exposure was a 24-hour infusion at 0.2 μg/kg/h versus no postoperative dexmedetomidine, with delirium assessed by CAM-ICU at 6 and 24 hours.
- Delirium was less frequent with dexmedetomidine at 6 hours (2.1% vs 7.0%, p = 0.032) and 24 hours (0.5% vs 3.6%, p = 0.042); adjusted OR at 6 hours was 0.25 (95% CI 0.07–0.88).
- Dexmedetomidine was also linked to shorter extubation time (6.85 ± 1.49 vs 9.47 ± 5.61 hours, p < 0.001) and similar adverse events and mortality, though the authors call for prospective confirmation.