Development and external validation of a machine learning model for the prediction of persistent acute kidney injury stage 3 in multi-centric, multi-national intensive care cohorts.

BACKGROUND: The aim of this retrospective cohort study was to develop and validate on multiple international datasets a real-time machine learning model able to accurately predict persistent acute kidney injury (AKI) in the intensive care unit (ICU). METHODS: We selected adult patients admitted to ICU classified as AKI stage 2 or 3 as defined by the "Kidney Disease: Improving Global Outcomes" criteria. The primary endpoint was the ability to predict AKI stage 3 lasting for at least 72 h while in the ICU. An explainable tree regressor was trained and calibrated on two tertiary, urban, academic, single-center databases and externally validated on two multi-centers databases. RESULTS: A total of 7759 ICU patients were enrolled for analysis. The incidence of persistent stage 3 AKI varied from 11 to 6% in the development and internal validation cohorts, respectively and 19% in external validation cohorts. The model achieved area under the receiver operating characteristic curve of 0.94 (95% CI 0.92-0.95) in the US external validation cohort and 0.85 (95% CI 0.83-0.88) in the Italian external validation cohort. CONCLUSIONS: A machine learning approach fed with the proper data pipeline can accurately predict onset of Persistent AKI Stage 3 during ICU patient stay in retrospective, multi-centric and international datasets. This model has the potential to improve management of AKI episodes in ICU if implemented in clinical practice.

Can carotid artery Doppler variations induced by the end-expiratory occlusion maneuver predict fluid responsiveness in septic shock patients?

BACKGROUND: An increase in cardiac index (CI) during an end-expiratory occlusion test (EEOt) predicts fluid responsiveness in ventilated patients. However, if CI monitoring is unavailable or the echocardiographic window is difficult, using the carotid Doppler (CD) could be a feasible alternative to track CI changes. This study investigates whether changes in CD peak velocity (CDPV) and corrected flow time (cFT) during an EEOt were correlated with CI changes and if CDPV and cFT changes predicted fluid responsiveness in patients with septic shock. METHODS: Prospective, single-center study in adults with hemodynamic instability. The CDPV and cFT on carotid artery Doppler and hemodynamic variables from the pulse contour analysis EV1000™ were recorded at baseline, during a 20-s EEOt, and after fluid challenge (500 mL). We defined responders as those who increased CI ≥ 15% after a fluid challenge. RESULTS: We performed 44 measurements in 18 mechanically ventilated patients with septic shock and without arrhythmias. The fluid responsiveness rate was 43.2%. The changes in CDPV were significantly correlated with changes in CI during EEOt (r = 0.51 [0.26-0.71]). A significant, albeit lower correlation, was found for cFT (r = 0.35 [0.1-0.58]). An increase in CI ≥ 5.35% during EEOt predicted fluid responsiveness with 78.9% sensitivity and 91.7% specificity, with an area under the ROC curve (AUROC) of 0.85. An increase in CDPV ≥ 10.5% during an EEOt predicted fluid responsiveness with 96.2% specificity and 53.0% sensitivity with an AUROC of 0.74. Sixty-one percent of CDPV measurements (from - 13.5 to 9.5 cm/s) fell within the gray zone. The cFT changes during EEOt did not accurately predict fluid responsiveness. CONCLUSIONS: In septic shock patients without arrhythmias, an increase in CDPV greater than 10.5% during a 20-s EEOt predicted fluid responsiveness with > 95% specificity. Carotid Doppler combined with EEOt may help optimize preload when invasive hemodynamic monitoring is unavailable. However, the 61% gray zone is a major limitation (retrospectively registered on Clinicaltrials.gov NCT04470856 on July 14, 2020).

Use of an innovative cuff pressure control and subglottic secretions drainage system in COVID-19 ARDS patients undergoing pronation.

We conducted a proof of concept study where Anapnoguard endotracheal tubes and its control unit were used in 15 patients with COVID-19 acute respiratory distress syndrome. Anapnoguard system provides suction, venting, rinsing of subglottic space and controls cuff pressure detecting air leakage through the cuff. Alpha-amylase and pepsin levels, as oropharyngeal and gastric microaspiration markers, were assessed from 85 tracheal aspirates in the first 72 h after connection to the system. Oropharyngeal microaspiration occurred in 47 cases (55%). Episodes of gastric microaspiration were not detected. Patient positioning, either prone or supine, did not affect alpha-amylase and pepsin concentration in tracheal secretions. Ventilator-associated pneumonia (VAP) rate was 40%. The use of the AG system provided effective cuff pressure control and subglottic secretions drainage. Despite this, no reduction in the incidence of VAP has been demonstrated, compared to data reported in the current COVID-19 literature. The value of this new technology is worth of being evaluated for the prevention of ventilator-associated respiratory tract infections.

Are Peripherally Inserted Central Catheters Suitable for Cardiac Output Assessment With Transpulmonary Thermodilution?

OBJECTIVES: Peripherally inserted central catheters are increasingly used in ICU as an alternative to centrally inserted central catheters for IV infusion. However, their reliability for hemodynamic measurements with transpulmonary thermodilution is currently unknown. We investigated the agreement between transpulmonary thermodilution measurements obtained with bolus injection through peripherally inserted central catheter and centrally inserted central catheter (reference standard) using a transpulmonary thermodilution-calibrated Pulse Contour hemodynamic monitoring system (VolumeView/EV1000). DESIGN: Prospective method-comparison study. SETTING: Twenty-bed medical-surgical ICU of a teaching hospital. PATIENTS: Twenty adult ICU patients who required hemodynamic monitoring because of hemodynamic instability and had both peripherally inserted central catheter and centrally inserted central catheter in place. INTERVENTION: The hemodynamic measurements obtained by transpulmonary thermodilution after injection of a cold saline bolus via both centrally inserted central catheter and either a single-lumen 4F or a double-lumen 5F peripherally inserted central catheter using were compared. In order to rule out bias related to manual injection, measurements were repeated using an automated rapid injection system. MEASUREMENTS AND MAIN RESULTS: A total of 320 measurements were made. Cardiac index was significantly higher when measured with double-lumen 5F peripherally inserted central catheter than with centrally inserted central catheter (mean, 4.5 vs 3.3 L/min/m; p < 0.0001; bias, 1.24 L/min/m [0.27, 2.22 L/min/m]; bias percentage, 31%). Global end-diastolic index, extravascular lung water index, and stroke volume index were also overestimated (853 ± 240 vs 688 ± 175 mL/m, 12.2 ± 4.2 vs 9.4 ± 2.9 mL/kg, and 49.6 ± 14.9 vs 39.5 ± 9.6 mL/m, respectively; p < 0.0001). Lower, albeit significant differences were found using single-lumen 4F peripherally inserted central catheter (mean cardiac index, 4.2 vs 3.7 L/min/m; p = 0.043; bias, 0.51 L/min/m [-0.53, 1.55 L/min/m]; bias percentage, 12.7%). All differences were confirmed, even after standardization of bolus speed with automated injection. CONCLUSIONS: Bolus injection through peripherally inserted central catheter for transpulmonary thermodilution using EV1000 led to a significant overestimation of cardiac index, global end-diastolic index, extravascular lung water index, and stroke volume index, especially when double-lumen 5F peripherally inserted central catheter was used (ClinicalTrial.gov NCT03834675).

Airway Closure during Surgical Pneumoperitoneum in Obese Patients.

BACKGROUND: Airway closure causes lack of communication between proximal airways and alveoli, making tidal inflation start only after a critical airway opening pressure is overcome. The authors conducted a matched cohort study to report the existence of this phenomenon among obese patients undergoing general anesthesia. METHODS: Within the procedures of a clinical trial during gynecological surgery, obese patients underwent respiratory/lung mechanics and lung volume assessment both before and after pneumoperitoneum, in the supine and Trendelenburg positions, respectively. Among patients included in this study, those exhibiting airway closure were compared to a control group of subjects enrolled in the same trial and matched in 1:1 ratio according to body mass index. RESULTS: Eleven of 50 patients (22%) showed airway closure after intubation, with a median (interquartile range) airway opening pressure of 9 cm H2O (6 to 12). With pneumoperitoneum, airway opening pressure increased up to 21 cm H2O (19 to 28) and end-expiratory lung volume remained unchanged (1,294 ml [1,154 to 1,363] vs. 1,160 ml [1,118 to 1,256], P = 0.155), because end-expiratory alveolar pressure increased consistently with airway opening pressure and counterbalanced pneumoperitoneum-induced increases in end-expiratory esophageal pressure (16 cm H2O [15 to 19] vs. 27 cm H2O [23 to 30], P = 0.005). Conversely, matched control subjects experienced a statistically significant greater reduction in end-expiratory lung volume due to pneumoperitoneum (1,113 ml [1,040 to 1,577] vs. 1,000 ml [821 to 1,061], P = 0.006). With airway closure, static/dynamic mechanics failed to measure actual lung/respiratory mechanics. When patients with airway closure underwent pressure-controlled ventilation, no tidal volume was inflated until inspiratory pressure overcame airway opening pressure. CONCLUSIONS: In obese patients, complete airway closure is frequent during anesthesia and is worsened by Trendelenburg pneumoperitoneum, which increases airway opening pressure and alveolar pressure: besides preventing alveolar derecruitment, this yields misinterpretation of respiratory mechanics and generates a pressure threshold to inflate the lung that can reach high values, spreading concerns on the safety of pressure-controlled modes in this setting.

Case Scenario: Perioperative Management of a Young Woman with Fontan Repair for Major Gynecologic Surgery.

Effective treatment for many congenital heart diseases diagnosed before birth has become available since the last three decades. Continuous improvements in surgical knowledge and techniques have allowed patients born with severe heart defects to survive through adulthood. However, palliative surgery often implies profound modifications of classical circulatory physiology, which must be taken into account particularly when general anesthesia is needed for major noncardiac surgery. Among the palliative surgeries, Fontan repair is an intervention aiming at excluding the right heart chambers with a total cavopulmonary conduit, which directs blood flow from both inferior and superior vena cavae directly to the right pulmonary artery. In such condition, patients are very sensitive to both preload reduction and pulmonary vascular resistances increase, so that a careful monitoring during anesthesia is required. Unfortunately, standard monitoring with a pulmonary artery catheter is not possible because of altered anatomy of right sections. In this case scenario, the authors report the perioperative management of a young woman who underwent major gynecologic surgery, who was managed using a transpulmonary thermodilution technique that was deemed more accurate than noncalibrated pulse-contour method and also able to provide more information regarding preload status. The authors adopted an integrated approach merging together hemodynamic and functional data (ScvO2 and venoarterial CO2 difference) to assess the appropriateness of hemodynamic management. The authors describe also pathophysiologic changes during such condition and also potential drawbacks of chosen technique.

How much oxygen in adult cardiac arrest?

Although experimental studies have suggested that a high arterial oxygen pressure (PaO2) might aggravate post-anoxic brain injury, clinical studies in patients resuscitated from cardiac arrest (CA) have given conflicting results. Some studies found that a PaO2 of more than 300 mm Hg (hyperoxemia) was an independent predictor of poor outcome, but others reported no association between blood oxygenation and neurological recovery in this setting. In this article, we review the potential mechanisms of oxygen toxicity after CA, animal data available in this field, and key human studies dealing with the impact of oxygen management in CA patients, highlighting some potential confounders and limitations and indicating future areas of research in this field. From the currently available literature, high oxygen concentrations during cardiopulmonary resuscitation seem preferable, whereas hyperoxemia should be avoided in the post-CA care. A specific threshold for oxygen toxicity has not yet been identified. The mechanisms of oxygen toxicity after CA, such as seizure development, reactive oxygen species production, and the development of organ dysfunction, need to be further evaluated in prospective studies.

Thirty-Day Complications, Unplanned Hospital Encounters, and Mortality after Endosonography and/or Guided Bronchoscopy: A Prospective Study.

BACKGROUND AND OBJECTIVE: Limited data exist regarding the adverse events of advanced diagnostic bronchoscopy, with most of the available information derived from retrospective datasets that primarily focus on early complications. METHODS: We conducted a 15-month prospective cohort study among consecutive patients undergoing endosonography and/or guided bronchoscopy under general anesthesia. We evaluated the 30-day incidence of severe complications, any complication, unplanned hospital encounters, and deaths. Additionally, we analyzed the time of onset (immediate, within 1 h of the procedure; early, 1 h-24 h; late, 24 h-30 days) and identified risk factors associated with these events. RESULTS: Thirty-day data were available for 697 out of 701 (99.4%) enrolled patients, with 85.6% having suspected malignancy and multiple comorbidities (median Charlson Comorbidity Index (IQR): 4 (2-5)). Severe complications occurred in only 17 (2.4%) patients, but among them, 10 (58.8%) had unplanned hospital encounters and 2 (11.7%) died within 30 days. A significant proportion of procedure-related severe complications (8/17, 47.1%); unplanned hospital encounters (8/11, 72.7%); and the two deaths occurred days or weeks after the procedure. Low-dose attenuation in the biopsy site on computed tomography was independently associated with any complication (OR: 1.87; 95% CI 1.13-3.09); unplanned hospital encounters (OR: 2.17; 95% CI 1.10-4.30); and mortality (OR: 4.19; 95% CI 1.74-10.11). CONCLUSIONS: Severe complications arising from endosonography and guided bronchoscopy, although uncommon, have significant clinical consequences. A substantial proportion of adverse events occur days after the procedure, potentially going unnoticed and exerting a negative clinical impact if a proactive surveillance program is not implemented.

Bloodstream infections in COVID-19 patients undergoing extracorporeal membrane oxygenation in ICU: An observational cohort study.

BACKGROUND: COVID-19 patients undergoing ECMO are at highly increased risk of nosocomial infections. OBJECTIVES: To study incidence, clinical outcomes and microbiological features of bloodstream infections (BSI) occurring during ECMO in COVID-19 patients. METHODS: Observational prospective cohort study enrolling consecutive COVID-19 patients undergoing veno-venous-ECMO in an Italian ICU from March 2020 to March 2022. RESULTS: In the study population of 68 patients (age 53 [49-60] years, 82% males), 30 (44%) developed bloodstream infections (BSI group) while 38 did not (N-BSI group) with an incidence of 32 events/1000 days of ECMO. In BSI group pre-ECMO respiratory support was shorter (6 [4-9] vs 9 [5-12] days, p = 0.02) and ECMO treatment was longer (18 [10-29] vs 11 [7-18] days, p = 0.03) than in N-BSI group. The overall ECMO and ICU mortality were 50% and 59%, respectively, without any inter-group difference (p = 1.00). A longer ECMO treatment was independently correlated with higher rate of BSI (p = 0.04, OR [95% CI] 1.06 [1.02-1.11]). Sixteen primary and 14 secondary infectious events were documented. Gram-positive pathogens were more common in primary than secondary BSI (88% vs 43%, p = 0.02) and Enterococcus faecalis (56%) was the most frequent one. Conversely, Gram-negative microorganisms were more often isolated in secondary rather than primary BSI (57% vs 13%, p = 0.02), with Acinetobacter baumannii (21%) and Pseudomonas aeruginosa (21%) as most represented species. The administration of Sars-CoV-2 antiviral drug showed independent correlation with a reduced rate of ICU mortality (p = 0.01, OR [95% CI] 0.22 [0.07-0.73]). CONCLUSIONS: Bloodstream infections represented a frequent complication without worsening clinical outcomes in our COVID-19 patients undergoing ECMO. Primary and secondary BSI events showed peculiar microbiological profiles.

Hemodynamic response to positive end-expiratory pressure and prone position in COVID-19 ARDS.

BACKGROUND: Use of high positive end-expiratory pressure (PEEP) and prone positioning is common in patients with COVID-19-induced acute respiratory failure. Few data clarify the hemodynamic effects of these interventions in this specific condition. We performed a physiologic study to assess the hemodynamic effects of PEEP and prone position during COVID-19 respiratory failure. METHODS: Nine adult patients mechanically ventilated due to COVID-19 infection and fulfilling moderate-to-severe ARDS criteria were studied. Respiratory mechanics, gas exchange, cardiac output, oxygen consumption, systemic and pulmonary pressures were recorded through pulmonary arterial catheterization at PEEP of 15 and 5 cmH2O, and after prone positioning. Recruitability was assessed through the recruitment-to-inflation ratio. RESULTS: High PEEP improved PaO2/FiO2 ratio in all patients (p = 0.004), and significantly decreased pulmonary shunt fraction (p = 0.012), regardless of lung recruitability. PEEP-induced increases in PaO2/FiO2 changes were strictly correlated with shunt fraction reduction (rho=-0.82, p = 0.01). From low to high PEEP, cardiac output decreased by 18 % (p = 0.05) and central venous pressure increased by 17 % (p = 0.015). As compared to supine position with low PEEP, prone positioning significantly decreased pulmonary shunt fraction (p = 0.03), increased PaO2/FiO2 (p = 0.03) and mixed venous oxygen saturation (p = 0.016), without affecting cardiac output. PaO2/FiO2 was improved by prone position also when compared to high PEEP (p = 0.03). CONCLUSIONS: In patients with moderate-to-severe ARDS due to COVID-19, PEEP and prone position improve arterial oxygenation. Changes in cardiac output contribute to the effects of PEEP but not of prone position, which appears the most effective intervention to improve oxygenation with no hemodynamic side effects.

Veno-arterial CO2 difference and respiratory quotient after cardiac arrest: An observational cohort study.

PURPOSE: To characterize venous-arterial CO2 difference (ΔpCO2) and the respiratory quotient (RQ) in post cardiac arrest patients and evaluate the association between these parameters and patient outcome. MATERIALS AND METHODS: Data were obtained retrospectively from post cardiac arrest patients admitted between 2007 and 2016 to a medical intensive care unit. Comatose, adult patients in whom arterial and venous blood gas analyses were concomitantly performed in the first 24 h were included. Patients were grouped according to the time-point of sampling; 0-6, 6-12 and 12-24 h after admission. RESULTS: 308 patients were included; 174 (56%) died before ICU discharge and 212 (69%) had an unfavorable neurologic outcome. RQ was associated with ICU mortality (OR:1.09 (95%CI: 1.04-1.14; p < 0.01)), although not with neurological outcome. ΔpCO2 was negatively associated with both ICU mortality (OR: 0.92 (95%CI: 0.86-0.99; p = 0.02)) and poor neurologic outcome (adjusted OR: 0.93 (95%CI: 0.87-0.99; p = 0.02)). ΔpCO2 predicted an elevated RQ; a ΔpCO2 above 8.5 mmHg identified a high RQ with reasonable sensitivity and specificity. CONCLUSIONS: RQ was associated with ICU mortality and ΔpCO2 identified elevated RQ in the early phase after cardiac arrest. However, ΔpCO2 were negatively associated with both ICU mortality and neurologic outcome.

Are single-lumen 5Fr and triple-lumen 6Fr PICCs suitable for hemodynamic assessment by trans-pulmonary thermodilution? A pilot study.

BACKGROUND: Single-lumen 4Fr or double-lumen 5Fr power injectable peripherally inserted central catheters (PICCs) are not accurate for trans-pulmonary thermodilution (TPTD), since they overestimate cardiac index and other TPTD-derived parameters when compared with centrally inserted central catheters (CICCs) because of the smaller size of their lumen. We hypothesize that PICCs with larger lumen size may be reliable for the cardiac index assessment using the TPTD. METHODS: This is a single-centre, prospective method-comparison study that included adult patients admitted in ICU who required a calibrated Pulse Contour hemodynamic monitoring system (VolumeView/EV1000™) for circulatory shock and had both PICC and CICC in place. We compared TPTD measurements via single-lumen 5Fr or triple-lumen 6Fr polyurethane power injectable PICCs with triple-lumen 7Fr CICC (reference standard). To rule out biases related to manual injection, measurements were repeated using an automated rapid injection system. We performed Bland-Altman analysis accounting for multiple observations per patient. RESULTS: A total of 320 measurements were performed in 15 patients. During the manual phase, the cardiac index measured with either single-lumen 5Fr or triple-lumen 6Fr PICCs were comparable with cardiac index measured with triple-lumen 7Fr CICC (3.2 ± 1.04 vs. 3.2 ± 1.06 L/min/m2, bias 2.2% and 3.3 ± 0.8 vs. 3.0 ± 0.7 L/min/m2, bias 8.5%, respectively). During the automated phase, triple-lumen 6Fr PICC slightly overestimated the cardiac index when compared to triple-lumen 7Fr CICC (CI 3.4 ± 0.7 vs. 3.0 ± 0.7 L/min/m2, bias 12.5%; p = 0.012). For both single-lumen 5Fr and triple-lumen 6Fr PICCs, percentage error vs. triple-lumen 7Fr CICC was below 20% (14.7% and 19% during the manual phase and 14.4% and 13.8% during the automated phase, respectively). Similar results were observed for TPTD-derived parameters. CONCLUSIONS: During hemodynamic monitoring with TPTD, both single-lumen 5Fr PICCs and triple-lumen 6Fr PICCs can be used for cold fluid bolus injection as an alternative to CICC (ClinicalTrials.gov NCT04241926).

Oxygen and carbon dioxide levels in patients after cardiac arrest.

BACKGROUND: The relationship of PaO2 and PaCO2 levels with outcome after cardiac arrest (CA) is controversial. Few studies have analysed both PaO2 and PaCO2 in this setting and the overall exposure to different PaO2 and PaCO2 levels has not been taken into account. METHODS: We reviewed blood gas data obtained within the first 24 h from all comatose adult patients who were admitted to the intensive care unit after successful resuscitation from CA. Exposure times to different PaO2 and PaCO2 thresholds were reported as areas under the curve (AUC) and the time above these thresholds was then calculated. The primary outcome measure was neurological outcome assessed using the Cerebral Performance Category (CPC) score at 3 months. An unfavourable outcome was defined as a CPC of 3-5 and a favourable outcome as a CPC of 1-2. RESULTS: A total of 356 patients were studied, with a median number of 9 [6-11] blood gas measurements within the first 24 h after admission. The highest and lowest PaO2 and PaCO2 were similar in patients with unfavourable and favourable neurological outcomes. There were no differences in the AUCs or times over different thresholds of PaO2 and PaCO2 in the two groups. In a multivariable analysis, high blood lactate concentrations on admission, presence of shock and a non-shockable initial rhythm were significantly associated with unfavourable outcome. CONCLUSIONS: There was no association between exposure to various levels of PaO2 and PaCO2 and neurological outcome after cardiac arrest.

Data on the application of early coagulation support protocol in the management of major trauma patients.

This article provides additional data on the application of early coagulation support protocol in the management of major trauma patients. Data come from a retrospective analysis reported in the article "Early coagulation support protocol: a valid approach in real-life management of major trauma patients. Results from two Italian centres" [1]. Data contain information about the relationship between differences in resource use and mortality outcomes, and patient demographic and clinical features at presentation. Furthermore, a comparison between resource consumption, the probability of multiple transfusions and the mortality outcomes among propensity-score matched patients is reported.

Early coagulation support protocol: A valid approach in real-life management of major trauma patients. Results from two Italian centres.

INTRODUCTION: Early coagulation support (ECS) includes prompt infusion of tranexamic acid, fibrinogen concentrate, and packed red blood cells for initial resuscitation of major trauma patients. The aim of this study was to determine the effects, in terms of blood product consumption, length of stay, and in-hospital mortality, of the ECS protocol, compared to the massive transfusion protocol (MTP) in the treatment of major trauma patients. PATIENTS AND METHODS: A retrospective analysis was conducted using the registry data of two Italian trauma centres. Adult major trauma patients with, or at risk of, active bleeding who were managed according to the MTP during the years 2011-2012, or the ECS protocol during the years 2013-2014 and were considered at risk of multiple transfusions, were enrolled. The primary endpoint was to determine whether the ECS protocol reduces the use of blood products in the acute management of trauma patients. Secondary endpoints were the outcome measures of length of stay in ICU, length of stay in hospital, and mortality at 24-hours and 28-days after hospital admission. RESULTS: Among the 518 major trauma patients admitted to the trauma centres during the study period, 235 patients (118 in the pre-ECS period and 117 in the ECS period) matched one of the inclusion criteria and were enrolled in the study. Compared with the pre-ECS period, the ECS period showed a reduction in the average consumption of packed red blood cells (-1.87 units, 95% confidence interval [CI], -2.40, -1.34), platelets (-1.28 units; 95% CI, -1.64, -0.91), and fresh frozen plasma (-1.69; 95% CI, -2.14, -1.25) in the first 24-hours. Furthermore, during the ECS period, we recorded a 10-day reduction in the hospital length of stay (-10 days, 95% CI, -11.6, -8.4) and a non-significant 28-day mortality increase. CONCLUSIONS: The ECS protocol was effective in reducing blood product consumption compared to the MTP and confirmed the importance of early fibrinogen administration as a strategy of rapid coagulation. This novel approach may be adopted in real-life management of major trauma patients.