Critical Care Medicine最新文献

Objective: The aim of this study was to develop and externally validate a machine-learning model that retrospectively identifies patients with acute respiratory distress syndrome (acute respiratory distress syndrome [ARDS]) using electronic health record (EHR) data.

Design: In this retrospective cohort study, ARDS was identified via physician-adjudication in three cohorts of patients with hypoxemic respiratory failure (training, internal validation, and external validation). Machine-learning models were trained to classify ARDS using vital signs, respiratory support, laboratory data, medications, chest radiology reports, and clinical notes. The best-performing models were assessed and internally and externally validated using the area under receiver-operating curve (AUROC), area under precision-recall curve, integrated calibration index (ICI), sensitivity, specificity, positive predictive value (PPV), and ARDS timing.

Patients: Patients with hypoxemic respiratory failure undergoing mechanical ventilation within two distinct health systems.

Interventions: None.

Measurements and main results: There were 1,845 patients in the training cohort, 556 in the internal validation cohort, and 199 in the external validation cohort. ARDS prevalence was 19%, 17%, and 31%, respectively. Regularized logistic regression models analyzing structured data (EHR model) and structured data and radiology reports (EHR-radiology model) had the best performance. During internal and external validation, the EHR-radiology model had AUROC of 0.91 (95% CI, 0.88-0.93) and 0.88 (95% CI, 0.87-0.93), respectively. Externally, the ICI was 0.13 (95% CI, 0.08-0.18). At a specified model threshold, sensitivity and specificity were 80% (95% CI, 75%-98%), PPV was 64% (95% CI, 58%-71%), and the model identified patients with a median of 2.2 hours (interquartile range 0.2-18.6) after meeting Berlin ARDS criteria.

Conclusions: Machine-learning models analyzing EHR data can retrospectively identify patients with ARDS across different institutions.

Objective: To determine the effectiveness and safety of andexanet and prothrombin complex concentrates (PCCs) when administered after intracranial hemorrhage (ICrH) associated with direct oral anticoagulants, specifically apixaban or rivaroxaban.

Design: A multicenter, retrospective, observational study of patients with apixaban or rivaroxaban-related ICrH who received andexanet or PCCs between January 1, 2015, and March 31, 2023. A predefined sensitivity analysis excluding patients with an admission Glasgow Coma Scale score of less than 7 was also performed.

Setting: Forty-two stroke centers in the United States.

Patients: A total of 1133 patients.

Interventions: None.

Measurements and main results: The primary efficacy outcome was the percentage of patients with excellent or good hemostasis as defined by the modified Sarode criteria. The primary safety outcome was the occurrence of a thrombotic event (TE) during their hospital stay. Of the 1133 patients evaluated, 1096 were included. In the full hemostatic efficacy analysis, patients receiving andexanet (87.8%) had higher odds of achieving excellent or good hemostasis (odds ratio [OR] 1.60; 95% CI, 1.00-2.56; p = 0.048) compared with PCCs (81.8%). Patients treated with andexanet (7.9%) had higher odds of a TE (OR 1.91; 95% CI, 1.13-3.20; p = 0.014) compared to those treated with PCCs (4.2%). No differences in hemostatic or thrombotic outcomes were observed when the sensitivity analysis was applied.

Conclusions: Despite statistically higher odds of achieving hemostatic efficacy with andexanet, we also observed higher odds of a TE with no difference in discharge outcomes observed between groups. When those with more severe neurologic injuries were excluded, efficacy and safety outcomes were similar between treatments.

Objectives: Mechanical power is a crucial concept in understanding ventilator-induced lung injury (VILI). We adopted the null hypothesis that under the same mechanical power, resulting from combinations of different static and dynamic variables-some with high stress per cycle and others without-would inflict similar degrees of damage on lung epithelial and endothelial cells as well as on the extracellular matrix in experimental acute respiratory distress syndrome (ARDS). To test this hypothesis, we varied tidal volume (Vt), which correlates with the stretching force per cycle, while adjusting respiratory rate (RR) to yield similar mechanical power values for identical durations across all experimental groups.

Design: Animal study.

Setting: Laboratory investigation.

Subjects: Thirty male Wistar rats (333 ± 26 g).

Interventions: Twenty-four hours after intratracheal administration of Escherichia coli lipopolysaccharide, animals were anesthetized and mechanically ventilated (positive end-expiratory pressure = 3 cm H2O) with combination of Vt and RR sufficient to induce similar mechanical power (n = 8/group): Vt = 6 mL/kg, RR = 140 breaths/minute (low Vt-high RR [LVT-HRR]); Vt = 12 mL/kg, RR = 70 breaths/minute (high Vt-low RR [HVT-LRR]); and Vt = 18 mL/kg, RR = 50 breaths/minute (very-high Vt-very-low RR [VHVT-VLRR]). All groups were ventilated for 80 minutes. A control group, not subjected to mechanical ventilation (MV), was used for molecular biology analyses.

Measurements and main results: After 80 minutes of MV, lung overdistension, alveolar/interstitial edema, fractional area of E-cadherin, and biomarkers of lung inflammation (interleukin-6), lung stretch (amphiregulin), damage to epithelial (surfactant protein B) and endothelial cells (vascular cell adhesion molecule 1 and angiopoietin-2), and extracellular matrix (versican and syndecan) were higher in group VHVT-VLRR than LVT-HRR. Plateau pressure and driving pressure increased progressively from LVT-HRR to HVT-LRR and VHVT-VLRR.

Conclusions: In the current experimental model of ARDS, mechanical power alone is insufficient to account for VILI. Instead, the manner in which its components are applied determines the extent of injury at a given mechanical power value.

Objectives: Weaning failure from mechanical ventilation (MV) is primarily caused by increased respiratory load and decreased respiratory neuromuscular competency, leading to a rapid shallow breathing pattern. We hypothesized that the product of diaphragmatic contraction velocity (a sonographic estimate of respiratory load) and tidal volume (an estimate of breathing pattern), termed the volume-velocity index (VVI), may predict weaning outcomes.

Design: The diagnostic accuracy of VVI (mL*cm/s) in predicting weaning outcomes was prospectively assessed, along with its relationship to indices of breathing effort, including esophageal pressure swings (ΔPes), the pressure-time product of esophageal pressure (PTPes), and maximal inspiratory pressure (MIP). A power analysis, informed by the results of an inception cohort, determined the required sample size for the validation cohort. Patients were enrolled through consecutive sampling. Weaning failure was defined as failure of the spontaneous breathing trial (SBT) or the need for MV within 48 hours.

Setting: The study was conducted in a tertiary academic ICU.

Patients: VVI was evaluated in critical care patients undergoing a SBT for the first time.

Interventions: None.

Measurements and main results: In the inception cohort (n = 30), VVI was significantly higher in successful weaning compared to failures (764.76 [±432.61] vs. 278 [±183.66], p < 0.001). It correlated with ΔPes (r = 0.74, R2 = 0.55), PTPes (r = 0.76, R2 = 0.58), and MIP (r = 0.75, R2 = 0.55) all p values less than 0.001. In the validation cohort (n = 40), VVI was higher in successful weaning (840 [550, 1220] vs. 250 [225, 302.5], p < 0.001) and predicted weaning success with an area under the receiver operating characteristic of 0.92 (95% CI, 0.83-1).

Conclusions: VVI effectively differentiates between weaning success and failure, shows a strong correlation with respiratory effort indices, and may enhance weaning protocols.

Objectives: Septic encephalopathy (SE) occurs in up to 50% of critically ill patients with sepsis and is associated with a high mortality and morbidity. The pathophysiology of SE is complex and involves increased levels of inflammatory mediators. Commonly used sedative drugs, such as propofol and midazolam, may worsen neuronal inflammation. Dexmedetomidine (DEX) has been shown to decrease the production of inflammatory mediators in experimental models of sepsis. The aim of this study was to investigate the effect of DEX on biomarkers associated with SE in critically ill patients with sepsis.

Design: Pilot, open-label, randomized controlled clinical trial.

Setting: Single-center University Hospital, Switzerland.

Patients: Adult patients with sepsis admitted to the ICU, who required intubation and ongoing sedative medication between September 1, 2019, and June 30, 2022.

Interventions: DEX-based sedation compared with propofol and/or midazolam-based sedation and serum S100-β level at 48 hr after randomization.

Measurements and main results: The study included 70 participants with 34 (48.6%) randomized to the DEX group and 36 (51.4%) to the propofol/midazolam group. Median S100-β levels in the DEX group at 48 hr were 0.103 (interquartile range 0.052-0.194) ng/ml, and in the propofol/midazolam group 0.189 (0.086-0.368) ng/mL (p = 0.064). Other biomarker showed no differences over time. In patients with a Glasgow Coma Scale less than or equal to 13, the median S100-β level in the DEX group was 0.13 ng/mL (0.06-0.18) compared to 0.91 ng/mL (0.43-0.96) in the propofol/midazolam group (p = 0.033).

Conclusions: DEX-based sedation compared to propofol/midazolam-based sedation did not show any significant difference in S100-β or any other markers of SE in critically ill patients with sepsis requiring mechanical ventilation. The finding of lower S100-β levels in DEX-sedated patients with GCS less than 13 warrants further investigation.

Objectives: Timely documentation of patient-concordant goals of care (GOC) in the ICU aims to promote patient autonomy and patient-centered care where the harms of interventions outweigh the potential benefits. This study examined the prevalence, timing, and predictors of ICU patients undergoing new and updated GOC documentation events while in the ICU.

Design: Multicenter retrospective study.

Setting and patients: All adults admitted to four ICUs from July 1, 2023, to December 31, 2023.

Interventions: None.

Main outcomes: The primary outcome was to determine the prevalence, timing, and predictors of new-GOC and updated-GOC documentation events following ICU admission.

Measurements and main results: We used multivariable logistic regression to identify predictors for new-GOC or updated-GOC documentation events using a backward stepwise elimination. Of the 2130 patients included, 13.3% (n = 284) had a new-GOC documentation event, and 16.3% (n = 346) had an updated-GOC documentation event. New-GOC events occurred sooner than updated-GOC events (median [interquartile range]: 18.3 [7.8-70.5] vs 73.7 [22.7-157.8] hr). Factors associated with GOC documentation events included age (odds ratio [OR] = 1.02, 95% CI, 1.01-1.03), frailty (OR = 1.30; 95% CI, 1.16-1.46), Sequential Organ Failure Assessment (SOFA) (OR = 1.23, 95% CI, 1.17-1.29), metastatic cancer (OR = 3.69, 95% CI, 2.17-6.26), ICU admission post medical emergency team review (OR = 1.94, 95% CI, 1.40-2.69), and cardiac arrest (OR = 2.23, 95% CI, 1.22-4.06) and if pre-ICU GOC had established treatment limitations, namely selective treatment goals (OR = 4.33, 95% CI, 3.18-5.90) or comfort-based treatment goals (OR = 7.66, 95% CI, 2.72-21.61). Apart from SOFA, all other factors remained significantly associated with GOC documentation events, even after accounting for ICU mortality as a competing outcome.

Conclusions: Almost 30% of patients had new- or updated-GOC documentation events while in ICU. Increasing age, higher SOFA scores, metastatic cancer, frailty, cardiac arrest, and ICU admission post-MET review predicted GOC changes while in ICU.

Objectives: After ICU admission, the quality of life (QoL) of ICU survivors is often significantly lower compared to their peers. However, recent studies showed that this impaired QoL cannot be fully explained by the physical, mental, and cognitive problems post-ICU, alluding to other determinants of QoL. Therefore, we aimed to explore ICU survivors' experienced QoL 1-2 years post-ICU, focusing on factors beyond functional outcomes.

Design: Qualitative interview study.

Setting: Seven hospitals in the Netherlands.

Patients: ICU survivors aged greater than or equal to 16 years admitted to the ICU between July 2022 and January 2023.

Interventions: None.

Measurements and main results: ICU patients were purposively sampled. Interviews were audiotaped, transcribed, and analyzed according to the principles of thematic content analysis. All interviews were coded independently by two researchers and participant recruitment was continued until no new themes were identified. Twenty-four semistructured interviews were performed between March and June 2024. The interviews resulted in 28 categories, from which seven main themes emerged regarding patients' experienced QoL: functional impairments (e.g., physical problems), participation (e.g., independence, work), support (e.g., informal care), environment (e.g., financial resources, personal circumstances), individual values (e.g., perspective on life, religion), comparison (e.g., expectations, reference), and coping (e.g., adaptation, acceptance). Patients described how these themes affected their QoL, both positively and negatively.

Conclusions: This study shows that perceived QoL after critical illness is impacted not only by patients' functional impairments but also by participation, support, environment, individual values, comparison, and coping. The themes identified in this study stress the importance of considering patients' individual and context factors to provide optimal post-ICU support.

Objectives: Blood transfusion is an essential therapy for patients receiving intensive care. The objective of the study was to better characterize the factors influencing the decision to transfuse RBCs during intensive care.

Design: Retrospective cohort analysis using a new analytic method called decision time-interval analysis, which segments care into objectively defined sequential time intervals.

Setting: Three ICUs during the period from January 2018 to June 2023.

Patients: Adult ICU patients 18 years old or older.

Interventions: None.

Measurements and main results: The pre-study planned outcome was the likelihood of RBC transfusion during each time interval of care. We analyzed 199,296 decision time-intervals occurring during 19,439 ICU encounters of 18,544 patients ranging in age from 18 to 103 years. The pre-transfusion hemoglobin concentration did not fully account for the likelihood of transfusion. Multivariable logistic regression analysis using either generalized linear mixed-effects or generalized estimating equations was used to model the impact of other clinical factors present at the time of the decision to transfuse for patients with hemoglobins in the 7-9 g/dL range. Factors that were significantly associated with an increased likelihood of transfusion included a decline in hemoglobin from the prior value, chest tube drainage, concurrent transfusion of plasma or platelets, prior RBC transfusion, use of extracorporeal membrane oxygenation, vasopressors, mechanical ventilation, renal replacement therapy, patient age, and male sex. The odds ratios for the principal clinical factors were significantly different in medical, surgical, and cardiac surgery ICUs.

Conclusions: The decision to transfuse RBC during ICU care is multifactorial and not adequately explained by the pre-transfusion hemoglobin concentration alone. Our findings have direct relevance for clinical guidelines regarding transfusion care, the interpretation of prior studies of transfusion based on hemoglobin concentration, and the opportunity for predictive analytics to improve patient outcomes.

Objectives: To evaluate the impact of prophylactic IV antibiotics on ventilator-associated pneumonia (VAP) and other outcomes in critically ill patients with acute brain injury.

Data sources: We searched PubMed, Embase, CINAHL, and other databases without language restrictions from inception to November 1, 2024.

Study selection: We included randomized trials that enrolled patients 16 years old or older with severe acute brain injury (e.g., trauma, stroke, and hypoxia) and compared prophylactic IV antibiotics to placebo or no antibiotics. Studies evaluating oral or gastric antibiotic prophylaxis were excluded.

Data extraction: Two reviewers independently evaluated study characteristics, pneumonia rates, mortality, duration of mechanical ventilation, ICU and hospital length of stay, post-admission bacteremia, antibiotic utilization, neurologic outcomes, and adverse events including antibiotic-resistance and Clostridioides difficile infections. Quality was assessed using the Cochrane risk-of-bias template. Studies were synthesized using inverse-variance random-effects models. Sensitivity analyses included stratifications by risk-of-bias, Glasgow Coma Scale, antibiotic duration, and mechanism of brain injury.

Data synthesis: We identified seven eligible trials (827 patients). Prophylactic antibiotics were associated with less VAP (risk ratio [RR], 0.61; 95% CI, 0.47-0.78; I2 = 23%) but not with changes in duration of mechanical ventilation (mean difference [MD], -0.45; 95% CI, -1.63 to +0.74), ICU length of stay (MD, -1.76 d; 95% CI, -3.94 to +0.42 d), hospital mortality (RR, 0.91; 95% CI, 0.74-1.12), or long-term neurologic outcomes. On stratification by mechanism of injury, antibiotic prophylaxis was associated with significantly less VAP, shorter ICU length of stay, and a trend toward lower mortality in patients with structural but not hypoxic brain injuries. Other sensitivity analyses were consistent with the primary findings.

Conclusions: Among critically ill patients with acute brain injury, prophylactic IV antibiotics were associated with less VAP but not with changes in length of stay, mortality, or neurologic outcomes. Patients with structural brain injuries may benefit more than those with hypoxic brain injury but further studies are needed to confirm or refute this possibility.