Critical care explorations最新文献

Context: To evaluate and compare sublingual microscopy (SLM) to urinary glycosaminoglycan (GAG) assays, dimethylmethylene blue (DMMB) assay, and liquid chromatography tandem mass spectrometry (LC-MS/MS) quantification of GAGs as biomarkers of microvascular dysfunction in patients with sepsis and septic shock.

Hypothesis: Indicators of microvascular dysfunction and markers of endothelial glycocalyx (eGC) degradation would be associated with sepsis.

Methods and models: Prospective, observational case-control study.

Setting: Denver Health Medical Center, a safety-net hospital in Denver, CO.

Subjects: Forty-four adult patients with sepsis or septic shock and 24 healthy control patients undergoing elective orthopedic procedures. Exclusion criteria included pregnancy and incarceration.

Results: Sublingual microvascular parameters (De Backer Density, proportion of perfused vessels) were measured using darkfield sidestream microscopy, and urinary GAGs were measured via DMMB colorimetric assay and LC-MS/MS targeting heparan sulfate (HS), dermatan sulfate, and keratan sulfate. Validation of HS quantification was performed on a subset using hydrophilic interaction liquid chromatography-mass spectrometry (HILIC-MS). LC-MS/MS HS was significantly higher in sepsis vs. controls (Area under the curve 0.85; 95% CI, 0.76-0.95), demonstrating higher diagnostic performance than SLM (De Backer Density AUC 0.71) and DMMB GAGs (AUC 0.62). LC-MS/MS and HILIC-MS HS levels were strongly correlated (R² = 0.97, p < 0.001). DMMB GAGs were associated with HS subtypes (p = 0.05) and SLM density (p = 0.03). No significant associations with in-hospital mortality or acute kidney injury were observed.

Interpretation and conclusions: Among the evaluated modalities, LC-MS/MS quantification of HS showed the greatest discriminative ability for identifying sepsis and correlated strongly with established mass spectrometric methods. SLM exhibited moderate diagnostic utility and significant associations with GAG levels, reinforcing its biologic relevance. However, its bedside application may be limited by challenges in image acquisition and analysis. The concordance across SLM, DMMB, and LC-MS/MS supports eGC degradation as a key feature of sepsis pathophysiology. These findings highlight the promise of LC-MS/MS as a scalable, rapid, and mechanistically informed platform for biomarker-driven enrichment in future sepsis trials and clinical care.

Background: Mortality prediction in ICU adults is only marginally improved when medication regimen complexity (MRC) data is incorporated into traditional regression models. Machine learning (ML) may improve this prediction.

Objective: To compare the performance of different ML approaches incorporating MRC data to both traditional and advanced regression approaches, with and without MRC data, to predict hospital mortality in ICU adults.

Derivation cohort: Nine hundred ninety-one ICU adults at the University of North Carolina (UNC) Health System.

Validation cohort: A temporally distinct cohort of 4,878 ICU adults at UNC and an external cohort of 12,290 ICU adults at the Oregon Health and Science University.

Prediction model: Supervised, classification-based ML models (e.g., Random Forest, Support Vector Machine [SVM], and XGBoost) were developed. Twenty-seven variables at ICU baseline (age, sex, service, diagnosis) and 24 hours (illness severity, supportive care use, fluid balance, laboratory values, MRC-ICU, vasopressor use) associated with mortality, and 14 missingness indicator variables, were included in each ML model. Traditional and advanced (equipped with linear predictors, predictors in nature cubic splines, predictors in smoothing cubic splines, and local linear predictors) regression models were optimized using stepwise selection by Bayesian Information Criterion. Area under the receiver operating characteristic (AUROC) was compared among models.

Results: Random Forest, SVM, and XGBoost achieved AUROCs of 0.83, 0.85, and 0.82, respectively, on the test set. Traditional regression models based on Sequential Organ Failure Assessment, Acute Physiology and Chronic Health Evaluation (APACHE) II, MRC-ICU + Sequential Organ Failure Assessment + APACHE II with and without an interaction term, and a full model including all 27 available variables demonstrated AUROCs of 0.81, 0.72, 0.82, 0.83, and 0.86, respectively. Advanced regression models yielded AUROCs of 0.85, 0.86, 0.85, and 0.84, respectively. The MRC-ICU exhibited a moderate level of feature importance in both XGBoost and Random Forest models. Models demonstrated lower performance in the validation cohorts.

Conclusions: Use of ML, compared with traditional and advanced regression methods, did not improve hospital mortality prediction despite medication data inclusion. The MRC-ICU demonstrates moderate feature importance in select ML models.

Objective: Severe infectious diseases are a leading cause of morbidity and mortality worldwide, particularly in sub-Saharan Africa (SSA), where young and middle-aged adults are disproportionately affected. Although age-related immune changes such as inflammaging and immunosenescence have been well characterized in high-income countries, their relevance to host responses during infection in SSA remains poorly understood. We aimed to characterize age-associated differences in immune, metabolic, and endothelial responses to severe infection in a prospective, multicenter cohort of adults in Uganda.

Design: Prospective cohort study.

Setting: Two public referral hospitals in Uganda.

Patients: Non-pregnant adults (18 yr old or older) hospitalized with severe, undifferentiated infection.

Interventions: None.

Measurements and main results: We analyzed clinical data and serum Olink proteomic profiles from 434 participants (median age: 45 yr old, interquartile range : 31-57). Clinically, illness severity and mortality were highest and comparable among adults 35-44, 45-59, and 60 years old or older, relative to younger adults. HIV prevalence peaked in the 35-44 and 45-59 age groups. Although most host responses were conserved across age groups after adjustment for sex and high-burden co-infections, patients 60 years old or older exhibited distinct immune dysregulation characterized by signs of Th1-predominant innate immune activation (increased CXCL9, CCL18, MCP1, and MCP4 expression, reduced interleukin-13 expression), dysregulated adaptive immunity (increased soluble CD27 and CD70 expression, reduced CD21 [CR2] expression), and increased cellular turnover and endothelial remodeling.

Conclusions: Older age (60 yr old or older) is associated with distinct host responses to severe infection in SSA. These findings may inform development of age-stratified, host-directed treatment strategies for severe infectious diseases.

Importance: Excessive cognitive load impairs task performance and contributes to burnout, but studies of cognitive load in pediatric critical care medicine (PCCM) settings are limited.

Objectives: To better understand cognitive load in an academic PCCM setting and how cognitive load differs based on experience, role, task type, and task frequency.

Design, settings, and participants: Prospective two-part survey at a quaternary children's hospital PCCM department. Part 1 (February to March 2022) assessed routine role-specific tasks; part 2 (June to August 2022) evaluated acute resuscitation. Participants were registered nurses (RNs), respiratory therapists (RTs), and physicians + advanced practice providers (APPs).

Main outcomes and measures: Raw cognitive load (1-9 Paas scale), net cognitive load (Paas × task frequency), and NASA-Task Load Index (NASA-TLX) subdomain scores (0-100) for acute resuscitation. Role was the primary exposure; between-group differences were analyzed using analysis of variance with pairwise comparisons.

Results: There were 109-part 1 and 79-part 2 survey respondents. Across all tasks, mean raw Paas scores were highest for physicians + APPs (5.2 ± 1.1), followed by RNs (4.8 ± 1.0) and RTs (4.0 ± 1.4; p = 0.004). In the three highest-load shared tasks-acute resuscitation, rescuing a decompensating patient, and managing advanced life-support devices-RNs reported significantly higher raw load than physicians + APPs and RTs. For bedside patient assessment, RNs had higher net cognitive load (25.0 ± 8.7) than physicians + APPs (20.3 ± 7.0; p = 0.01) and RTs (18.9 ± 8.9; p = 0.01). Nursing experience correlated with overall net cognitive load (r = 0.30; p = 0.02). During resuscitation, RNs reported higher NASA-TLX scores than other providers in all but two subdomains.

Conclusions and relevance: Cognitive load in PCCM varies significantly by role and task type. Nurses experience high raw cognitive load from critical events and net cognitive load from bedside patient assessment, suggesting opportunities for role-specific workflow redesign and cognitive load reduction strategies to benefit staff and patients.

Objective: To train and test supervised machine learning (ML) models to predict low cardiac output syndrome (LCOS) within the first 48 postoperative hours in neonates undergoing cardiothoracic surgery.

Design: Retrospective observational study. An efficient tree-based gradient-boosting algorithm (LightGBM) ML models were developed to predict LCOS occurrence at 2-, 4-, 6-, and 12-hour forecasting horizons, incorporating data from the prediction time and the two preceding hours. SHapley Additive exPlanations (SHAP) analyses were used for feature importance analyses.

Setting: Single center, January 2012 to April 2023.

Patients: Neonates 28 days old or younger who underwent cardiothoracic surgery.

Interventions: None.

Measurements and main results: A total of 181 neonates were included, with 14.9% experiencing LCOS. A multivariate time-series dataset was constructed using hourly clinical and laboratory variables recorded during the first 48 postoperative hours. The LightGBM ML models achieved area under the receiver operating characteristic curve values ranging from 0.91 to 0.98 and area under the precision-recall curve values ranging from 0.60 to 0.80 for LCOS prediction across 2-, 4-, 6-, and 12-hour forecasting horizons. SHAP analyses identified higher vasoactive inotrope score, lower urine output, and higher serum lactate as the most influential predictors.

Conclusions: This study demonstrates that the supervised machine learning models can accurately predict LCOS in neonates, offering high interpretability. The findings support further validation in multicenter settings and integration into clinical workflows to enhance postoperative critical cardiac care neonates.

Objective: To train and test supervised machine learning (ML) models to predict low cardiac output syndrome (LCOS) within the first 48 postoperative hours in neonates undergoing cardiothoracic surgery.

Design: Retrospective observational study. An efficient tree-based gradient-boosting algorithm (LightGBM) ML models were developed to predict LCOS occurrence at 2-, 4-, 6-, and 12-hour forecasting horizons, incorporating data from the prediction time and the two preceding hours. SHapley Additive exPlanations (SHAP) analyses were used for feature importance analyses.

Setting: Single center, January 2012 to April 2023.

Patients: Neonates 28 days old or younger who underwent cardiothoracic surgery.

Interventions: None.

Measurements and main results: A total of 181 neonates were included, with 14.9% experiencing LCOS. A multivariate time-series dataset was constructed using hourly clinical and laboratory variables recorded during the first 48 postoperative hours. The LightGBM ML models achieved area under the receiver operating characteristic curve values ranging from 0.91 to 0.98 and area under the precision-recall curve values ranging from 0.60 to 0.80 for LCOS prediction across 2-, 4-, 6-, and 12-hour forecasting horizons. SHAP analyses identified higher vasoactive inotrope score, lower urine output, and higher serum lactate as the most influential predictors.

Conclusions: This study demonstrates that the supervised machine learning models can accurately predict LCOS in neonates, offering high interpretability. The findings support further validation in multicenter settings and integration into clinical workflows to enhance postoperative critical cardiac care neonates.

The impact of common hemoglobinopathies, such as sickle cell trait (SCT), on outcomes in adults requiring extracorporeal membrane oxygenation (ECMO) remains understudied. Extracorporeal Life Support Organization registry data was analyzed to assess outcomes of adults with SCT or sickle cell disease (SCD) who underwent venoarterial or venovenous ECMO. Among 215 patients identified, 49 had SCT and 166 had SCD. The prevalence of SCT appeared grossly underestimated. Age-adjusted survival rates for SCT patients were favorable compared with those with SCD for venoarterial (43.5% vs. 19.5%; p = 0.04) and venovenous (73.3% vs. 44.8%; p = 0.11) ECMO. Bleeding and thrombotic event rates and renal complications in SCT patients were comparable to those with SCD and similar to reported rates in general adult ECMO populations. While our findings suggest that ECMO may be safely used in patients with SCT, further investigation is essential to determine the clinical impact of sickle cell and other hemoglobinopathies on ECMO therapy.

Background: Sepsis-induced cardiomyopathy (SICM) is prevalent yet remains difficult to diagnose using conventional echocardiography primarily due to its dependence on loading conditions, the dynamic nature of sepsis, and varied cardiovascular phenotypes. Recent advancements in noninvasive myocardial work (MW) analysis, particularly through left ventricle pressure-strain loop (LV PSL), offer a promising strategy for evaluating myocardial function by combining strain imaging with blood pressure data. This technique may address the limitations inherent in traditional measures such as ejection fraction, which can be influenced by fluctuating hemodynamics in sepsis and may not accurately reflect underlying myocardial function.

Case summary: This report presents three cases wherein patients exhibited either preserved or only mildly reduced left ventricular systolic function based on ejection fraction (LVEF), but were found to have diminished MW indices, including global work index, global constructive work, and global work efficiency along with low flow by left ventricle outflow-tract.

Conclusions: Relying solely on LVEF for diagnosing SICM is problematic due to numerous confounding variables. MW parameters constitute innovative, noninvasive echocardiographic indicators that have demonstrated value across a spectrum of cardiac disorders. Although these parameters appear promising as bedside assessment tools, their application within the context of sepsis warrants further investigation.

High-flow nasal cannula (HFNC) is a common noninvasive respiratory therapy for respiratory failure, offering positive airway pressure and dead space clearance. In critically ill patients, additional nasal tubes for feeding or monitoring are often required, but their effect on HFNC performance is not well understood. This bench study evaluated the impact of nasal tube placement on dead space clearance, nasopharyngeal pressure, and airway resistance using standard and asymmetrical cannulas. Inserting a 5-Fr tube with a standard cannula had very little effect, whereas bigger sizes (12-Fr and 16-Fr) slightly increased airway pressure and reduce Co2 clearance. An asymmetrical cannula exhibited variable effects depending on the side of the tube placement. Higher pressure and better clearance were achieved with a tube placed in the smaller cannula side. However, if occlusion of the two nares becomes excessive with the tube in place, downsizing the cannula might be recommended.

Objective: The objective of this systematic review was to characterize adverse event detection methods in the ICU setting, including neonatal, pediatric, and adult ICUs, to summarize the evidence of their performance characteristics.

Data sources: Ovid MEDLINE, Ovid Embase, CINAHL, the Cochrane Library, and Google Scholar.

Study selection: Title and abstract screening, as well as full text review, were performed by two reviewers independently using Covidence software. Articles were included if they consisted of original research in a peer-reviewed journal with implementation of an adverse event detection method and reported the total number or category of adverse events, level of harm, or implementation of quality improvement (QI).

Data extraction: Data were extracted by two reviewers with 20% in duplicate. Extracted data included the study type and period/date, the adverse event detection method, the setting (location and type of ICU), ICU bed base, and the data of interest outlined above in study selection.

Data synthesis: Fifty-nine studies in neonatal, pediatric, and adult ICUs were included. Every category of adverse event detection was represented, including incident reporting (IR) (38 studies), trigger tool use (14 studies), trained observation (TO; 11 studies), and structured review (10 studies). TO identified the most adverse events per 100 patient days (57.3), and IR the least (6.4). Only 12 studies (20%) described QI initiatives.

Conclusions: Detection methods likely need to be used in combination for comprehensive results. Definitions of adverse events and associated harms need to be standardized to facilitate future comparison and better understanding of the performance of individual methods of detection. In addition, more emphasis needs to be placed on the dissemination of practice change in response to detection. These will be important steps to better characterize high rates of adverse events in the ICU, thereby fueling patient safety initiatives.