Journal of critical care最新文献

Background: Machine learning models for predicting acute kidney injury (AKI) prognosis have primarily been developed in resource-rich settings, with limited validation in resource-limited environments. This study applied machine learning techniques to predict in-hospital mortality and major adverse kidney events within 28 days (MAKE-28) among critically ill patients with AKI in Southeast Asia.

Method: Data were derived from the Southeast Asia AKI cohort, a prospective multicenter study of critically ill patients. Demographic, clinical, and laboratory variables collected at ICU admission were analyzed. Logistic regression, random forest, and extreme gradient boosting (XGBoost) were used to develop prediction models, with recursive feature elimination applied for feature selection.

Results: Of 6993 ICU patients, 1650 individuals with AKI were included for analysis. Of these, 778 (47.1%) died during hospitalization and 1204 (73.9%) experienced MAKE-28. The three models demonstrated comparable performance in predicting MAKE-28 and hospital mortality (AUC 0.73-0.76 for MAKE-28 outcome and AUC 0.71-0.75 for hospital mortality). Discrimination ability was moderate, and all machine learning approaches outperformed conventional clinical scores. No difference in performance was observed between logistic regression and more complex machine learning models.

Conclusion: Machine learning models using routinely available clinical variables may offer useful prognostic information for AKI outcomes in resource-limited settings and outperform traditional scoring systems. External validation is required to confirm generalizability and support clinical implementation.

Cardiogenic shock (CS) remains a leading cause of death in intensive cardiac care. Outcomes are limited by delayed recognition of hypoperfusion, heterogeneous phenotypes, and late escalation of therapies. Diagnosis and risk stratification have progressed with the introduction of the SCAI staging system, which provides a common language for clinical severity and guides escalation of care. Echocardiography and invasive hemodynamics remain central for defining ventricular phenotype, detecting mechanical complications, and tailoring therapy. Early activation of multidisciplinary shock teams is increasingly adopted to coordinate rapid assessment and structured management. Treatment focuses on restoring perfusion, correcting the underlying cause, and preventing further organ injury. Norepinephrine is generally preferred as first-line vasopressor, while inotropes, including dobutamine and milrinone, are selected according to physiologic profile rather than theoretical advantages. Mechanical circulatory support (MCS) should be considered early in refractory hypoperfusion, using integrated clinical, metabolic, echocardiographic, and PAC-derived triggers when feasible. Multiorgan support (ventilation, renal replacement therapy, and ECMO-related strategies such as LV unloading/venting) should be aligned with shock trajectory and goals of care. CS management should shift from a "one-size-fits-all" model to an early, phenotype-driven strategy with explicit perfusion targets and timely MCS escalation, supported by shock teams and networks. Emerging biomarkers and machine-learning tools may further improve risk stratification and treatment timing.

Background: Acute kidney injury (AKI) is common in the intensive care unit (ICU) but is often detected only after creatinine rises or oliguria develops. Although novel biomarkers allow earlier detection, their cost limits use. The urine albumin-to-creatinine ratio (uACR) is inexpensive, yet its role in AKI risk stratification remains uncertain.

Methods: In a prospective single-centre cohort of mixed ICU patients, adults with existing AKI or at high risk (modified AKI Risk Based on Creatinine [ARBOC] score ≥ 3) were enrolled. uACR was measured at enrollment (uACR at Time 0) and 24 h (uACR at Time 1). Outcomes included the prevalence and prognostic value of elevated uACR (≥ 3.4 mg/mmol) for incident, progressive, or persistent AKI (> 48 h), and for ≥30% decline in estimated glomerular filtration rate (eGFR) at discharge. Predictive performance was assessed using the area under the receiver operating characteristic curve (AUC), change in AUC, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and decision-curve analysis.

Results: Of 1010 patients screened, 203 were analysed (89 with and 114 without AKI at enrollm). Elevated uACR was frequent (72% with AKI; 52% without). In AKI patients, high uACR correlated with longer and more persistent AKI. Discrimination for incident AKI was modest (AUC 0.61 and 0.66 for uACR at Time 0 and uACR at Time 1) but higher for persistent AKI (both AUC 0.68). Adding uACR at Time 0 to ARBOC improved reclassification (NRI 0.48; IDI 0.04) and clinical net benefit.

Conclusions: uACR modestly identified incident AKI and more strongly predicted persistent AKI. As a simple biomarker, uACR may serve as a low-cost adjunct to existing ICU risk stratification tools, but requires further validation before consideration for routine clinical use.

Background: Pulmonary congestion is a prognostic marker for heart failure (HF) morbidity and mortality; however, the current congestion evaluation depends on traditional physical examination, which lacks adequate sensitivity. Lung ultrasound (LUS) has been investigated as a more sensitive method to guide decongestion in decompensated HF.

Methods: A systematic review and meta-analysis synthesizing evidence from randomized controlled trials (RCTs) obtained from PubMed, CENTRAL, Scopus, and Web of Science until March 2025. Using Stata MP v. 17, we used the fixed-effects model to report dichotomous outcomes using the risk ratio (RR) and continuous outcomes using the standardized mean difference with a 95% confidence interval (CI).

Prospero id: CRD42024620337.

Results: Nine RCTs with 1095 patients were included. LUS-guided management significantly decreased the risk of HF hospitalization/all-cause mortality (RR: 0.72, [95% CI 0.56, 0.93], p = 0.01), HF hospitalization (RR: 0.65, [95% CI 0.48, 0.88], p = 0.01), and HF urgent visits (RR: 0.38, [95% CI 0.22, 0.66], p < 0.0001). There was no significant difference between LUS-guided management and standard of care regarding the incidence of hypotension (RR: 1.87, [95% CI 0.56, 6.20], p = 0.31), hypokalemia (RR: 0.93, [95% CI 0.48, 1.82], p = 0.83), hyperkalemia (RR: 0.98, [95% CI 0.62, 1.53], p = 0.91), and acute kidney injury/impaired renal function (RR: 1.08, [95% CI 0.66, 1.77], p = 0.75).

Conclusion: LUS-guided decongestion was associated with a significant decrease in the risk of HF re-hospitalization and HF urgent visits, with a tolerable safety profile, compared to standard care, with similar rates of hypotension, hypokalemia, hyperkalemia, and AKI.

Purpose: Despite advances in perioperative care, delayed oral fluid intake after extubation remains common and is often based on tradition rather than evidence. This study aimed to evaluate whether immediate oral fluid intake "sipping" after extubation reduces thirst and discomfort and is safe in an intensive care setting.

Methods: In this single-center, prospective, randomized controlled trial, 160 ICU patients who met extubation criteria were randomized 1:1 to either delayed fluid intake (2 h post-extubation) or immediate sipping (up to 3 ml/kg over 2 h). Thirst, discomfort, and adverse effects (nausea, vomiting, aspiration) were assessed at 0, 5, 30, 60, 90, and 120 min. Thirst relief was also evaluated in the experimental group. Statistical significance was set at p < 0.05.

Results: At 120 min, 64 of 80 patients in each group (80 %; 95 % CI: 70-88 %) reported thirst. The difference between groups was 0.0 % (95 % CI: -12 % to 12 %; p = 1.000). However, thirst relief between baseline and 120 min was observed in 11.3 % of patients in the sipping group (95 % CI: 5-20 %) vs. 1.3 % in the standard group (95 % CI: 0-7 %), with a difference of 10 % (95 % CI: 1-19 %; p = 0.0338). At 90 min, throat discomfort was present in 23.8 % of the sipping group (95 % CI: 15-35 %) vs. 42.5 % in the standard group (95 % CI: 32-54 %), with a difference of -18.7 % (95 % CI: -34 % to -3 %; p = 0.0118). Adverse effects (nausea, vomiting) were rare and comparable; no aspiration events were observed.

Conclusion: Immediate oral fluid intake "sipping"after extubation appears to be safe, improves thirst relief, and reduces discomfort in ICU patients without increasing adverse effects. These findings challenge traditional fasting practices and support early rehydration in post-extubation care.

Trial registration: The trial was registered at ClinicalTrials.gov on January 6, 2023 (Identifier: NCT05819645).