Serum potassium concentrations are commonly between 3.5 and 5.0 mmol/l. Standardised protocols for potassium range and supplementation in the ICU are lacking. The purpose of this retrospective analysis of ICU patients was to investigate potassium concentrations, variability and supplementation, and their association with in-hospital mortality. ICU patients ≥ 18 years, with ≥ 2 serum potassium values, treated at the Charité - Universitätsmedizin Berlin between 2006 and 2018 were eligible for inclusion. We categorised into groups of mean potassium concentrations: < 3.0, 3.0–3.5, > 3.5–4.0, > 4.0–4.5, > 4.5–5.0, > 5.0–5.5, > 5.5 mmol/l and potassium variability: 1st, 2nd and ≥ 3rd standard deviation (SD). We analysed the association between the particular groups and in-hospital mortality and performed binary logistic regression analysis. Survival curves were performed according to Kaplan–Meier and tested by Log-Rank. In a subanalysis, the association between potassium supplementation and in-hospital mortality was investigated. In 53,248 ICU patients with 1,337,742 potassium values, the lowest mortality (3.7%) was observed in patients with mean potassium concentrations between > 3.5 and 4.0 mmol/l and a low potassium variability within the 1st SD. Binary logistic regression confirmed these results. In a subanalysis of 22,406 ICU patients (ICU admission: 2013–2018), 12,892 (57.5%) received oral and/or intravenous potassium supplementation. Potassium supplementation was associated with an increase in in-hospital mortality in potassium categories from > 3.5 to 4.5 mmol/l and in the 1st, 2nd and ≥ 3rd SD (p < 0.001 each). ICU patients may benefit from a target range between 3.5 and 4.0 mmol/l and a minimal potassium variability. Clear potassium target ranges have to be determined. Criteria for widely applied potassium supplementation should be critically discussed. Trial registration German Clinical Trials Register, DRKS00016411. Retrospectively registered 11 January 2019, http://www.drks.de/DRKS00016411

Continuous glucose monitoring (CGM) has not yet been implemented in the intensive care unit (ICU) setting. The purpose of this study was to evaluate reliability, feasibility, nurse acceptance and accuracy of the Medtronic Sentrino® CGM system in critically ill patients. Sensors were inserted into the subcutaneous tissue of the patient’s thigh, quantifying interstitial glucose concentration for up to 72 h per sensor. Reliability and feasibility analysis included frequency of data display, data gaps and reasons for sensor removal. We surveyed nurse acceptance in a questionnaire. For the accuracy analysis, we compared sensor values to glucose values obtained via blood gas analysis. Potential benefits of CGM were investigated in intra-individual analyses of factors, such as glycemic variability or time in target range achieved with CGM compared to that achieved with intermittent glucose monitoring. The device generated 68,655 real-time values from 31 sensors in 20 critically ill patients. 532 comparative blood glucose values were collected. Data were displayed during 32.5 h [16.0/62.4] per sensor, which is 45.1 % of the expected time of 72 h and 84.8 % of 37.9 h actual monitoring time. 21 out of 31 sensors were removed prematurely. 79.1 % of the nursing staff rated the device as not beneficial; the response rate was one-third. Mean absolute relative difference was 15.3 % (CI 13.5–17.0 %). Clarke error grid: 76.9 % zone A, 21.6 % zone B, 0.2 % zone C, 0.9 % zone D, 0.4 % zone E. Bland–Altman plot: mean bias +0.53 mg/dl, limits of agreement +64.6 and −63.5 mg/dl. Accuracy deteriorated during elevated glycemic variability and in the hyperglycemic range. There was no reduction in dysglycemic events during CGM compared to 72 h before and after CGM. If CGM was measuring accurately, it identified more hyperglycemic events when compared to intermittent measurements. This study was not designed to evaluate potential benefits of CGM on glucose control. The subcutaneous CGM system did not perform with satisfactory accuracy, feasibility, or nursing acceptance when evaluated in 20 medical-surgical ICU patients. Low point accuracy and prolonged data gaps significantly limited the potential clinical usefulness of the CGM trend data. Accurate continuous data display, with a MARD < 14 %, showed potential benefits in a subgroup of our patients. Trial registration NCT02296372; Ethic vote Charité EA2/095/14

(1) Background: Female sex is considered a risk factor for Intensive Care Unit-Acquired Weakness (ICUAW). The aim is to investigate sex-specific aspects of skeletal muscle metabolism in the context of ICUAW. (2) Methods: This is a sex-specific sub-analysis from two prospectively conducted trials examining skeletal muscle metabolism and advanced muscle activating measures in critical illness. Muscle strength was assessed by Medical Research Council Score. The insulin sensitivity index was analyzed by hyperinsulinemic-euglycemic (HE) clamp. Muscular metabolites were studied by microdialysis. M. vastus lateralis biopsies were taken. The molecular analysis included protein degradation pathways. Morphology was assessed by myocyte cross-sectional area (MCSA). Multivariable linear regression models for the effect of sex on outcome parameters were performed. (3) Results: n = 83 (♂n = 57, 68.7%; ♀n = 26, 31.3%) ICU patients were included. ICUAW was present in 81.1%♂ and in 82.4%♀ at first awakening (p = 0.911) and in 59.5%♂ and in 70.6%♀ at ICU discharge (p = 0.432). Insulin sensitivity index was reduced more in women than in men (p = 0.026). Sex was significantly associated with insulin sensitivity index and MCSA of Type IIa fibers in the adjusted regression models. (4) Conclusion: This hypothesis-generating analysis suggests that more pronounced impairments in insulin sensitivity and lower MCSA of Type IIa fibers in critically ill women may be relevant for sex differences in ICUAW.

Background: Anticoagulation therapy with heparin is a frequent treatment in intensive care units and is monitored by activated partial thromboplastin clotting time (aPTT). It has been demonstrated that reaching an established anticoagulation target within 24 hours is associated with favorable outcomes. However, patients respond to heparin differently and reaching the anticoagulation target can be challenging. Machine learning algorithms may potentially support clinicians with improved dosing recommendations. Objective: This study evaluates a range of machine learning algorithms on their capability of predicting the patients’ response to heparin treatment. In this analysis, we apply, for the first time, a model that considers time series. Methods: We extracted patient demographics, laboratory values, dialysis and extracorporeal membrane oxygenation treatments, and scores from the hospital information system. We predicted the numerical values of aPTT laboratory values 24 hours after continuous heparin infusion and evaluated 7 different machine learning models. The best-performing model was compared to recently published models on a classification task. We considered all data before and within the first 12 hours of continuous heparin infusion as features and predicted the aPTT value after 24 hours. Results: The distribution of aPTT in our cohort of 5926 hospital admissions was highly skewed. Most patients showed aPTT values below 75 s, while some outliers showed much higher aPTT values. A recurrent neural network that consumes a time series of features showed the highest performance on the test set. Conclusions: A recurrent neural network that uses time series of features instead of only static and aggregated features showed the highest performance in predicting aPTT after heparin treatment.

(1) Background: Acute respiratory distress syndrome (ARDS) is a rare complication in multiply injured patients. Due to the rarity of ARDS development after trauma, little is known about outcomes of patients with trauma-associated ARDS compared to patients with non-trauma-associated ARDS. (2) Methods: This retrospective analysis included n = 1038 ARDS patients admitted to the ARDS center of Charité—Universitätsmedizin Berlin between 2007 and 2018. Patients with trauma-associated ARDS (n = 62) were compared to patients with non-trauma-associated ARDS (n = 976). In a secondary analysis, patients from the group with non-trauma-associated ARDS were 1:1 nearest neighbor matched to patients with trauma-associated ARDS. The primary outcomes were 28-day in-hospital mortality, 60-day in-hospital mortality, and overall in-hospital mortality. (3) Results: Overall in-hospital mortality in trauma-associated ARDS was 29.0% compared to 40.5% in all patients with non-trauma-associated ARDS (p = 0.074). The in-hospital mortality rate in matched patients with non-trauma-associated ARDS (33.9%) was comparable to the trauma-associated ARDS cohort (p = 0.701). Kaplan–Meier curves indicated time-sensitive variations in 28-day and 60-day in-hospital survival. (4) Conclusion: Mortality was not different in patients with trauma-associated ARDS compared to patients with non-trauma-associated ARDS. Survival rate in the Kaplan–Meier curves stabilized after the critical initial phase and throughout the further 60-day period in patients with trauma-associated ARDS compared to patients with non-trauma-associated ARDS. Since this divergence was less pronounced in the matched cohort, it may be related to the younger age, fewer comorbidities, and lower ARDS severity in patients with trauma-associated ARDS. Patients with trauma-associated ARDS remain a very different cohort compared to patients with non-trauma-associated ARDS. Therefore, the outcome comparison is limited, even after matching.

Due to an Editorial Office error during processing, a number of male and female symbols were incorrectly shown in the pdf version of the manuscript