Coronavirus disease 2019 (COVID-19) is associated with a high disease burden with 10% of confirmed cases progressing towards critical illness. Nevertheless, the disease course and predictors of mortality in critically ill patients are poorly understood. Following the critical developments in ICUs in regions experiencing early inception of the pandemic, the European-based, international RIsk Stratification in COVID-19 patients in the Intensive Care Unit (RISC-19-ICU) registry was created to provide near real-time assessment of patients developing critical illness due to COVID-19. As of April 22, 2020, 639 critically ill patients with confirmed SARS-CoV-2 infection were included in the RISC-19-ICU registry. Of these, 398 had deceased or been discharged from the ICU. ICU-mortality was 24%, median length of stay 12 (IQR, 5–21) days. ARDS was diagnosed in 74%, with a minimum P/F-ratio of 110 (IQR, 80–148). Prone positioning, ECCO2R, or ECMO were applied in 57%. Off-label therapies were prescribed in 265 (67%) patients, and 89% of all bloodstream infections were observed in this subgroup (n = 66; RR=3·2, 95% CI [1·7–6·0]). While PCT and IL-6 levels remained similar in ICU survivors and non-survivors throughout the ICU stay (p = 0·35, 0·34), CRP, creatinine, troponin, d-dimer, lactate, neutrophil count, P/F-ratio diverged within the first seven days (p<0·01). On a multivariable Cox proportional-hazard regression model at admission, creatinine, d-dimer, lactate, potassium, P/F-ratio, alveolar-arterial gradient, and ischemic heart disease were independently associated with ICU-mortality. The European RISC-19-ICU cohort demonstrates a moderate mortality of 24% in critically ill patients with COVID-19. Despite high ARDS severity, mechanical ventilation incidence was low and associated with more rescue therapies. In contrast to risk factors in hospitalized patients reported in other studies, the main mortality predictors in these critically ill patients were markers of oxygenation deficit, renal and microvascular dysfunction, and coagulatory activation. Elevated risk of bloodstream infections underscores the need to exercise caution with off-label therapies.

Accurate risk stratification of critically ill patients with coronavirus disease 2019 (COVID-19) is essential for optimizing resource allocation, delivering targeted interventions, and maximizing patient survival probability. Machine learning (ML) techniques are attracting increased interest for the development of prediction models as they excel in the analysis of complex signals in data-rich environments such as critical care. We retrieved data on patients with COVID-19 admitted to an intensive care unit (ICU) between March and October 2020 from the RIsk Stratification in COVID-19 patients in the Intensive Care Unit (RISC-19-ICU) registry. We applied the Extreme Gradient Boosting (XGBoost) algorithm to the data to predict as a binary outcome the increase or decrease in patients’ Sequential Organ Failure Assessment (SOFA) score on day 5 after ICU admission. The model was iteratively cross-validated in different subsets of the study cohort. The final study population consisted of 675 patients. The XGBoost model correctly predicted a decrease in SOFA score in 320/385 (83%) critically ill COVID-19 patients, and an increase in the score in 210/290 (72%) patients. The area under the mean receiver operating characteristic curve for XGBoost was significantly higher than that for the logistic regression model (0.86 vs. 0.69, P < 0.01 [paired t-test with 95% confidence interval]). The XGBoost model predicted the change in SOFA score in critically ill COVID-19 patients admitted to the ICU and can guide clinical decision support systems (CDSSs) aimed at optimizing available resources.

Direct assessment of capillary perfusion has been prioritized in hemodynamic management of critically ill patients in addition to optimizing blood flow on the global scale. Sublingual handheld vital microscopy has enabled online acquisition of moving image sequences of the microcirculation, including the flow of individual red blood cells in the capillary network. However, due to inherent content complexity, manual image sequence analysis remained gold standard, introducing inter-observer variability and precluding real-time image analysis for clinical therapy guidance. Here we introduce an advanced computer vision algorithm for instantaneous analysis and quantification of morphometric and kinetic information related to capillary blood flow in the sublingual microcirculation. We evaluated this technique in a porcine model of septic shock and resuscitation and cardiac surgery patients. This development is of high clinical relevance because it enables implementation of point-of-care goal-directed resuscitation procedures based on correction of microcirculatory perfusion in critically ill and perioperative patients.

BACKGROUND: Glycocalyx shedding after traumatic hemorrhagic or septic shock, as well as different resuscitation fluids, has been causally linked to increased vascular barrier permeability (VBP) resulting in tissue edema. In nontraumatic hemorrhagic shock (NTHS), it remains questionable whether glycocalyx degradation in itself results in an alteration of VBP. The composition of fluids can also have a modulatory effect on glycocalyx shedding and VBP. We hypothesized that the shedding of the glycocalyx during NTHS has little effect on VBP and that the composition of fluids can modulate these effects. METHODS: Fully instrumented Wistar-albino rats were subjected to a pressure-controlled NTHS (mean arterial pressure of 30 mm Hg) for 60 minutes. Animals were fluid resuscitated with Ringer’s acetate, balanced hydroxyethyl starch (HES) solution, or 0.9% normal saline to a mean arterial pressure of 80 mm Hg and compared with shams or nonresuscitated NTHS. Glycocalyx shed products were determined at baseline and 60 minutes after fluid resuscitation. Skeletal muscle microcirculation was visualized using handheld vital microscopy. VBP changes were assessed using plasma decay of 3 fluorescent dyes (40- and 500-kDa dextran and 70-kDa albumin), Evans blue dye exclusion, intravital fluorescence microscopy, and determination of tissue edema (wet/dry weight ratio). RESULTS: All glycocalyx shedding products were upgraded as a result of NTHS. Syndecan-1 significantly increased in NTHS (mean difference, −1668; 95% confidence interval [CI], −2336 to −1001; P < .0001), balanced crystalloid (mean difference, −964.2; 95% CI, −1492 to −436.4; P = .0001), and HES (mean difference, −1030; 95% CI, −1594 to −465.8; P = .0001) groups at the end of the experiment compared to baseline. Hyaluronan levels were higher at the end of the experiment in nonresuscitated NTHS (−923.1; 95% CI, −1216 to −630; P = .0001) and balanced crystalloid (−1039; 95% CI, −1332 to −745.5; P = .0001) or HES (−394.2; 95% CI, −670.1 to −118.3; P = .0027) groups compared to controls. Glycocalyx shedding resulted in microcirculation alterations as observed by handheld video microscopy. Total vessel density was altered in the normal saline (mean difference, 4.092; 95% CI, 0.6195–7.564; P = .016) and hemorrhagic shock (mean difference, 5.022; 95% CI, 1.55–8.495; P = .0024) groups compared to the control group, as well as the perfused vessel density and mean flow index. Despite degradation of endothelial glycocalyx, VBP as determined by 4 independent assays remained intact and continued to be so following fluid resuscitation. CONCLUSIONS: NTHS induced glycocalyx shedding and microcirculation alterations, without altering VBP. Fluid resuscitation partially restored the microcirculation without altering VBP. These results challenge the concept that the glycocalyx barrier is a significant contributor to VBP.

Oliguria is one of the leading triggers of fluid loading in patients in the intensive care unit (ICU). The purpose of this study was to assess the predictive value of urine Na+ (uNa+) and other routine urine biomarkers for cardiac fluid responsiveness in oliguric ICU patients. We conducted a prospective multicenter observational study in five university ICUs. Patients with urine output (UO) 65 mmHg received a fluid challenge. Cardiac fluid responsiveness was defined by an increase in stroke volume >15 % after fluid challenge. Urine and plasma biochemistry samples were examined before fluid challenge. We examined renal fluid responsiveness (defined as UO >0.5 ml/kg/h for 3 consecutive hours) after fluid challenge as a secondary endpoint. Fifty-four patients (age 51 ± 37 years, Simplified Acute Physiology Score II score 40 ± 20) were included. Most patients (72 %) were not cardiac responders (CRs), and 50 % were renal responders (RRs) to fluid challenge. Patient characteristics were similar between CRs and cardiac nonresponders. uNa+ (37 ± 38 mmol/L vs 25 ± 75 mmol/L, p = 0.44) and fractional excretion of sodium (FENa+) (2.27 ± 2.5 % vs 2.15 ± 5.0 %, p = 0.94) were not statistically different between those who did and those who did not respond to the fluid challenge. Areas under the receiver operating characteristic (AUROC) curves were 0.51 (95 % CI 0.35–0.68) and 0.56 (95 % CI 0.39–0.73) for uNa+ and FENa+, respectively. Fractional excretion of urea had an AUROC curve of 0.70 (95 % CI 0.54–0.86, p = 0.03) for CRs. Baseline UO was higher in RRs than in renal nonresponders (1.07 ± 0.78 ml/kg/3 h vs 0.65 ± 0.53 ml/kg/3 h, p = 0.01). The AUROC curve for RRs was 0.65 (95 % CI 0.53–0.78) for uNa+. In the present study, most oliguric patients were not CRs and half were not renal responders to fluid challenge. Routine urinary biomarkers were not predictive of fluid responsiveness in oliguric normotensive ICU patients.