Comparative performance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data.

INTRODUCTION: Benchmarking intensive care units for audit and feedback is frequently based on comparing actual mortality versus predicted mortality. Traditionally, mortality prediction models rely on a limited number of input variables and significant manual data entry and curation. Using automatically extracted electronic health record data may be a promising alternative. However, adequate data on comparative performance between these approaches is currently lacking. METHODS: The AmsterdamUMCdb intensive care database was used to construct a baseline APACHE IV in-hospital mortality model based on data typically available through manual data curation. Subsequently, new in-hospital mortality models were systematically developed and evaluated. New models differed with respect to the extent of automatic variable extraction, classification method, recalibration usage and the size of collection window. RESULTS: A total of 13 models were developed based on data from 5,077 admissions divided into a train (80%) and test (20%) cohort. Adding variables or extending collection windows only marginally improved discrimination and calibration. An XGBoost model using only automatically extracted variables, and therefore no acute or chronic diagnoses, was the best performing automated model with an AUC of 0.89 and a Brier score of 0.10. DISCUSSION: Performance of intensive care mortality prediction models based on manually curated versus automatically extracted electronic health record data is similar. Importantly, our results suggest that variables typically requiring manual curation, such as diagnosis at admission and comorbidities, may not be necessary for accurate mortality prediction. These proof-of-concept results require replication using multi-centre data.

Between-trial heterogeneity in ARDS research.

PURPOSE: Most randomized controlled trials (RCTs) in patients with acute respiratory distress syndrome (ARDS) revealed indeterminate or conflicting study results. We aimed to systematically evaluate between-trial heterogeneity in reporting standards and trial outcome. METHODS: A systematic review of RCTs published between 2000 and 2019 was performed including adult ARDS patients receiving lung-protective ventilation. A random-effects meta-regression model was applied to quantify heterogeneity (non-random variability) and to evaluate trial and patient characteristics as sources of heterogeneity. RESULTS: In total, 67 RCTs were included. The 28-day control-group mortality rate ranged from 10 to 67% with large non-random heterogeneity (I2 = 88%, p < 0.0001). Reported baseline patient characteristics explained some of the outcome heterogeneity, but only six trials (9%) reported all four independently predictive variables (mean age, mean lung injury score, mean plateau pressure and mean arterial pH). The 28-day control group mortality adjusted for patient characteristics (i.e. the residual heterogeneity) ranged from 18 to 45%. Trials with significant benefit in the primary outcome reported a higher control group mortality than trials with an indeterminate outcome or harm (mean 28-day control group mortality: 44% vs. 28%; p = 0.001). CONCLUSION: Among ARDS RCTs in the lung-protective ventilation era, there was large variability in the description of baseline characteristics and significant unexplainable heterogeneity in 28-day control group mortality. These findings signify problems with the generalizability of ARDS research and underline the urgent need for standardized reporting of trial and baseline characteristics.

Achieving protein targets without energy overfeeding in critically ill patients: A prospective feasibility study.

BACKGROUND & AIMS: High protein delivery during early critical illness is associated with lower mortality, while energy overfeeding is associated with higher mortality. Protein-to-energy ratios of traditional enteral formulae are sometimes too low to reach protein targets without energy overfeeding. This prospective feasibility study aimed to evaluate the ability of a new enteral formula with a high protein-to-energy ratio to achieve the desired protein target while avoiding energy overfeeding. METHODS: Mechanically ventilated non-septic patients received the high protein-to-energy ratio nutrition during the first 4 days of ICU stay (n = 20). Nutritional prescription was 90% of measured energy expenditure. Primary endpoint was the percentage of patients reaching a protein target of ≥1.2 g/kg ideal body weight on day 4. Other endpoints included a comparison of nutritional intake to matched historic controls and the response of plasma amino acid concentrations. Safety endpoints were gastro-intestinal tolerance and plasma urea concentrations. RESULTS: Nineteen (95%) patients reached the protein intake target of ≥1.2 g/kg ideal body weight on day 4, compared to 65% in historic controls (p = 0.024). Mean plasma concentrations of all essential amino acids increased significantly from baseline to day 4. Predefined gastro-intestinal tolerance was good, but unexplained foul smelling diarrhoea occurred in two patients. In one patient plasma urea increased unrelated to acute kidney injury. CONCLUSIONS: In selected non-septic patients tolerating enteral nutrition, recommended protein targets can be achieved without energy overfeeding using a new high protein-to-energy ratio enteral nutrition.

Hyperoxia does not directly affect vascular tone in isolated arteries from mice.

Hospitalized patients often receive oxygen supplementation, which can lead to a supraphysiological oxygen tension (hyperoxia). Hyperoxia can have hemodynamic effects, including an increase in systemic vascular resistance. This increase suggests hyperoxia-induced vasoconstriction, yet reported direct effects of hyperoxia on vessel tone have been inconsistent. Furthermore, hyperoxia-induced changes in vessel diameter have not been studied in mice, currently the most used mammal model of disease. In this study we set out to develop a pressure-myograph model using isolated vessels from mice for investigation of pathways involved in hyperoxic vasoconstriction. Isolated conduit and resistance arteries (femoral artery and gracilis arteriole, respectively) from C57BL/6 mice were exposed to normoxia (PO2 of 80 mmHg) and three levels of hyperoxia (PO2 of 215, 375 and 665 mmHg) in a no-flow pressure myograph setup. Under the different PO2 levels, dose-response agonist induced endothelium-dependent vasodilation (acetylcholine, arachidonic acid), endothelium-independent vasodilation (s-nitroprusside), as well as vasoconstriction (norepinephrine, prostaglandin F2α) were examined. The investigated arteries did not respond to oxygen by a change in vascular tone. In the dose-response studies, maximal responses and EC50 values to any of the aforementioned agonists were not affected by hyperoxia either. We conclude that arteries and arterioles from healthy mice are not intrinsically sensitive to hyperoxic conditions. The present ex-vivo model is therefore not suitable for further research into mechanisms of hyperoxic vasoconstriction.

[Abbreviations in daily language: stop it]. / Afkortingen in het dagelijks taalgebruik: stop ermee.

Abbreviations are used more and more in physician common parlance and it seems they are on the way to becoming a new jargon. However, identical abbreviations may have different meanings, especially in different medical specialties. Moreover, many physicians do not know the meaning of specific abbreviations or are attributing the wrong meaning to the abbreviation. This will lead to misunderstanding and therefore danger to the patient. The authors are calling for a stop on the use of spoken abbreviations and for minimising the use of abbreviations in clinical notes and medical prescriptions.

[Healthcare professionals back in the driving seat]. / Zorgprofessionals moeten stuur weer overnemen.

This commentary discusses the increasingly observed managerilisation of healthcare. Managerilisation frequently results in a framework of rules, regulations and accompanying time-consuming forms and procedures to guide decision-making. Although likely developed with the best of intentions in mind, this framework may be of limited value and tends to leave healthcare professionals feeling frustrated and distrusted. In addition, overzealous bureaucracy and rigid adherence to protocols may be disadvantageous to patient care and outcomes. Instead, we advocate a renewed focus on common sense and in particular on a renewed trust in healthcare professionals. Their professional judgement is based on many years of education and bedside experience. Hospital management should once again seek to embrace their expertise, while healthcare professionals should actively seek to regain the reins when it comes to delivering healthcare.

Hyperoxia: At what level of SpO2 is a patient safe? A study in mechanically ventilated ICU patients.

BACKGROUND: Concerns have been expressed regarding a possible association between arterial hyperoxia and adverse outcomes in critically ill patients. Oxygen status is commonly monitored noninvasively by peripheral saturation monitoring (SpO2). However, the risk of hyperoxia above specific SpO2 levels in critically ill patients is unknown. The purpose of this study was to determine a threshold value of SpO2 above which the prevalence of arterial hyperoxia distinctly increases. METHODS: This is a cross-sectional study in adult mechanically ventilated intensive care patients in a tertiary referral center. In 100 patients, we collected 200 arterial blood gases (ABG) and simultaneously registered SpO2 levels, as well as hemodynamic and ventilation parameters and vasoactive medication. Patients under therapeutic hypothermia were excluded. RESULTS: The risk of arterial hyperoxia, defined as PaO2>100mmHg or >125mmHg, was negligible when SpO2 was ≤95% or ≤96%, respectively. The majority (89% and 54%, respectively for PaO2>100mmHg and 125mmHg) of ICU patients with SpO2 of 100% had arterial hyperoxia. The relation between SpO2 and PaO2 was not clearly affected by hemodynamic or other clinical variables (pH, pCO2, body temperature, recent blood transfusion). CONCLUSION: In critically ill patients, the prevalence of arterial hyperoxia increases when SpO2 is >95%. Above this saturation level, supplemental oxygen should be administered with caution in patients potentially susceptible to adverse effects of hyperoxia.

Pharmacokinetics of Anidulafungin in Critically Ill Intensive Care Unit Patients with Suspected or Proven Invasive Fungal Infections.

Echinocandins, such as anidulafungin, are the first-line treatment for candidemia or invasive candidiasis in critically ill patients. There are conflicting data on the pharmacokinetic properties of anidulafungin in intensive care unit (ICU) patients. Adult ICU patients (from 3 hospitals) receiving anidulafungin for suspected or proven fungal infections were included in the present study. Patients were considered evaluable if a pharmacokinetic curve for day 3 could be completed. Twenty-three of 36 patients (7 female and 16 male) were evaluable. The median (range) age and body weight were 66 (28 to 88) years and 76 (50 to 115) kg, respectively. Pharmacokinetic sampling on day 3 (n = 23) resulted in a median anidulafungin area under the concentration-time curve from 0 to 24 h (AUC0-24) of 72.1 (interquartile range [IQR], 61.3 to 94.0) mg · h · liter-1, a median daily trough concentration (C24) of 2.2 (IQR, 1.9 to 2.9) mg/liter, a median maximum concentration of drug in serum (Cmax) of 5.3 (IQR, 4.1 to 6.0) mg/liter, a median volume of distribution (V) of 46.0 (IQR, 32.2 to 60.2) liters, and a median clearance (CL) of 1.4 (IQR, 1.1 to 1.6) liters · h-1 Pharmacokinetic sampling on day 7 (n = 13) resulted in a median AUC0-24 of 82.7 (IQR, 73.0 to 129.5) mg · h · liter-1, a median minimum concentration of drug in serum (Cmin) of 2.8 (IQR, 2.2 to 4.2) mg/liter, a median Cmax of 5.9 (IQR, 4.6 to 8.0) mg/liter, a median V of 39.7 (IQR, 32.2 to 54.4) liters, and a median CL of 1.2 (IQR, 0.8 to 1.4) liters · h-1 The geometric mean ratio for the AUCday7/AUCday3 term was 1.13 (90% confidence interval [CI], 1.03 to 1.25). The exposure in the ICU patient population was in accordance with previous reports on anidulafungin pharmacokinetics in ICU patients but was lower than that for healthy volunteers or other patient populations. Larger cohorts of patients or pooled data analyses are necessary to retrieve relevant covariates. (This study has been registered at ClinicalTrials.gov under identifier NCT01438216.).

Neutrophil CD64 expression as a longitudinal biomarker for severe disease and acute infection in critically ill patients.

INTRODUCTION: Neutrophilic granulocytes express cluster of differentiation 64 (CD64) antigen upon activation. CD64 can be used as a marker of bacterial infection and sepsis. The goal of this study was to determine whether CD64 is a useful biomarker for critically ill patients and analyze longitudinal measurements with regard to outcome and sepsis severity. METHODS: In this prospective observational study, CD64 analysis was performed daily until discharge from ICU or death. Demographics, clinical, laboratory data, and outcome defined as 28-day survival were recorded. Patients were included when admitted to the ICU with sepsis, severe sepsis, or septic shock and within 24 h from start of antibiotic treatment. RESULTS: Hundred and fifty-five consecutive patients were enrolled. At baseline, a difference in CD64 of 2.26 (1.33-4.47) vs. 1.49 (0.89-2.24) (P = 0.004) was seen between patients with a positive culture and negative culture. CD64 at day 1 was higher with patients with septic shock when compared with sepsis (P = 0.012). No difference of CD64 between survivors and nonsurvivors was seen. CONCLUSION: This study demonstrated that CD64 discriminates between critically ill patients with culture positive and negative sepsis and correlates with severity of disease. However, CD64 index is not a good predictor for 28-day mortality in the critically ill patient.

Endothelial hyperpermeability after cardiac surgery with cardiopulmonary bypass as assessed using an in vitro bioassay for endothelial barrier function.

BACKGROUND: The mechanisms causing increased endothelial permeability after cardiopulmonary bypass (CPB) have not been elucidated. Using a bioassay for endothelial barrier function, we investigated whether endothelial hyperpermeability is associated with alterations in plasma endothelial activation and adhesion markers and can be attenuated by the use of pulsatile flow during CPB. METHODS: Patients undergoing cardiac surgery were randomized to non-pulsatile (n=20) or pulsatile flow CPB (n=20). Plasma samples were obtained before (pre-CPB) and after CPB (post-CPB), and upon intensive care unit (ICU) arrival. Changes in plasma endothelial activation and adhesion markers were determined by enzyme-linked immunosorbent assay. Using electric cell-substrate impedance sensing of human umbilical vein endothelial monolayers, the effects of plasma exposure on endothelial barrier function were assessed and expressed as resistance. RESULTS: Cardiopulmonary bypass was associated with increased P-selectin, vascular cell adhesion molecule-1, and von Willebrand factor plasma concentrations and an increase in the angiopoietin-2 to angiopoietin-1 ratio, irrespective of the flow profile. Plasma samples obtained after CPB induced loss of endothelial resistance of 21 and 23% in non-pulsatile and pulsatile flow groups, respectively. The negative effect on endothelial cell barrier function was still present with exposure to plasma obtained upon ICU admission. The reduction in endothelial resistance after exposure to post-CPB plasma could not be explained by CPB-induced haemodilution. CONCLUSION: The change in the plasma fingerprint during CPB is associated with impairment of in vitro endothelial barrier function, which occurs irrespective of the application of a protective pulsatile flow profile during CPB. CLINICAL TRIAL REGISTRATION: NTR2940.

Intravascular versus surface cooling speed and stability after cardiopulmonary resuscitation.

BACKGROUND AND OBJECTIVE: Mild therapeutic hypothermia (MTH) is used to limit neurological injury and improve survival after cardiac arrest (CA) and cardiopulmonary resuscitation, but the optimal mode of cooling is controversial. We therefore compared the effectiveness of MTH using invasive intravascular or non-invasive surface cooling with temperature feedback control. METHODS: This retrospective study in post-CA patients studied the effects of intravascular cooling (CoolGard, Zoll, n=97), applied on the intensive care unit (ICU) in one university hospital compared with those of surface cooling (Medi-Therm, Gaymar, n=76) applied in another university hospital. RESULTS: Time to reach target temperature and cooling speeds did not differ between groups. During the maintenance phase, mean core temperature was 33.1°C (range 32.7-33.7°C) versus 32.5°C (range 31.7-33.4°C) at targets of 33.0 and 32.5°C in intravascularly versus surface cooled patients, respectively. The variation coefficient for temperature during maintenance was higher in the surface than the intravascular cooling group (mean 0.85% vs 0.35%, p<0.0001). ICU survival was 60% and 50% in the intravascularly and surface cooled groups, respectively (NS). Lower age (OR 0.95; 95% CI 0.93 to 0.98; p<0.0001), ventricular fibrillation/ventricular tachycardia as presenting rhythm (OR 7.6; 95% CI 1.8 to 8.9; p<0.0001) and lower mean temperature during the maintenance phase (OR 0.52; 95% CI 0.25 to 1.08; p=0.081) might be independent determinants of ICU survival, while cooling technique and temperature variability did not contribute. CONCLUSIONS: In post-CA patients, intravascular cooling systems result in equal cooling speed, but less variation in temperature during the maintenance phase, as surface cooling. This may not affect the outcome.

Minimal effects on ex vivo coagulation during mild therapeutic hypothermia in post cardiac arrest patients.

OBJECTIVES: Mild therapeutic hypothermia (MTH) is being used to improve neurological outcome and survival in patients successfully resuscitated after cardiac arrest. The impact on coagulation may be difficult to assess since most coagulation parameters are measured at 37°C and not at actual body core temperature. Therefore we investigated the effects of MTH both at body core (target) temperature of 32°C and at 37°C. METHODS: Patients admitted at the ICU after cardiac arrest treated with MTH. Baseline blood samples, measured at 37°C were taken directly at arrival. The second and third samples were drawn within 1h and 24h after reaching target temperature and were measured at 32°C and 37°C. A final sample was drawn when the patient returned to normotemperature (measured at 37°C). Clotting time (CT) and maximum clotting formation (MCF) were measured with thromboelastometry. RESULTS: Upon reaching target temperature (32°C) Extem and Intem CT were increased compared to baseline with 57s (49-75) to 65s (59-72) and 165s (144-183) to 193s (167-212) respectively (median with IQR; P<0.05), with a further significant increase after 24h of hypothermia with 68s (57-80) and 221s (196-266). Samples analyzed at 32°C showed a significant longer CT of 12s in Extem and 33s in Intem compared to 37°C. MCF was not affected by MTH or adjustment of temperature. CONCLUSION: The mild effect of MTH on coagulation parameters remains unidentified when measured at 37°C. Although measurements at 32°C differ from those at 37°C, this does not appear to be of clinical relevance as all values were still within the reference range.

[End of life decisions, the Dutch form through Spanish eyes]. / Toma de decisiones al final de la vida, el modo neerlandés a traves de ojos españoles.

Abroad, but also in The Netherlands, there are many misunderstandings concerning end of life decisions and euthanasia. In general, euthanasia does not play any role in the intensive care units, simply because it does not fulfill the conditions to carry it out. However, there is still confusion, merely due to the assumption that the Dutch situation is different because of their legislation on euthanasia. The use of the unclear terminology such as "passive euthanasia", "voluntary euthanasia" or "involuntary euthanasia" contributes to the confusion of lay people and physicians, and should therefore be avoided. End of life decisions in intensive care patients are in fact a structural part of work of intensivists. Collecting all necessary information including the wishes and will of the patient, medical expertise and acknowledging limitations of medical treatment will help to determine futility of treatment goals. Once it is determined that surviving the intensive care unit with a quality of life acceptable for the patient is beyond reach, the goal of treatment should be improved and the dying process optimized. Stopping a treatment modality at the request of a will-competent patient or because of futility is not euthanasia.

Feasibility of online seizure detection with continuous EEG monitoring in the intensive care unit.

INTRODUCTION: Continuous EEG (cEEG) is of great interest in view of the reported high prevalence of non-convulsive seizures on intensive care units (ICUs). Here, we describe our experiences applying a seizure warning system using cEEG monitoring. METHODS: Fifty comatose ICU patients were included prospectively and monitored. Twenty-eight patients had post-anoxic encephalopathy (PAE) and 22 had focal brain lesions. A measure of neuronal interactions, synchronization likelihood, was calculated online over 10s EEG epochs and instances when the synchronization likelihood exceeded a threshold where marked as seizures. RESULTS: Five patients developed seizures. Our method detected seizures in three patients, in the other patients seizures were missed because of they were non-convulsive and had a focal character. The average false positive rate was 0.676/h. DISCUSSION: This is our first attempt to implement online seizure detection in the ICU. Despite problems with artifacts and that we missed focally oriented seizures, we succeeded in monitoring patients online. Given the relatively high occurrence of seizures, online seizure detection with cEEG merits further development for use in ICUs.

Computer-aided support improves early and adequate delivery of nutrients in the ICU.

BACKGROUND: In 2007 a national guideline on perioperative nutrition was issued in the Netherlands. As external indicator for adequacy of nutritional therapy, the percentage of malnourished patients who reach at least 1.2 grams of protein on day 4 after admission was chosen by the Netherlands Health Care Inspectorate. METHODS: We developed an algorithm that allows users to ask for advice on which artificial nutritional formula to prescribe and at which rate, assuring provision of adequate amounts of both protein and energy. Feedback on nutritional therapy is given to the users on a daily basis, and to the management per quarter. Both the advice and the feedback have been integrated in our data management system. The advice module is also available on-line. RESULTS: In the baseline situation over the first four quarters (2006) an average of 30.2% of patients who had a full day 4 in our unit reached the protein indicator. In the last six quarters post-implementation, the average percentage reached was 56.5% with values consistently over 50%. Changes were statistically significant at third quarter of 2007 (p<0.05) and thereafter (p<0.001). Results for day 7 of admission were unaffected, which indicates that targets were reached earlier during hospital stay. CONCLUSION: Our study shows that integration of nutritional advice and automatically generated feedback to users in a data management system consistently improves delivery of (early) nutrition.

The effect of rosiglitazone on asymmetric dimethylarginine (ADMA) in critically ill patients.

Asymmetric dimethylarginine (ADMA) plays a crucial role in the arginine-nitric oxide pathway. Critically ill patients have elevated levels of ADMA which proved to be a strong and independent risk factor for ICU mortality. The aim of this study was to investigate the effect of the peroxisome proliferator-activated receptor (PPAR)-gamma agonist rosiglitazone on ADMA plasma levels in critically ill patients. In a randomized controlled pilot study, ADMA, arginine and symmetric dimethylarginine (SDMA) were measured in 21 critically ill patients on the intensive care unit (ICU). Twelve patients received 4mg rosiglitazone once a day for a maximum of 6 weeks or until discharge or death. Nine patients served as control patients. In addition, total sequential organ failure assessment (SOFA score), kidney function and liver function were determined. Compared to the ADMA levels of healthy individuals as specified in earlier studies, ADMA plasma levels of critically ill patients were significantly higher (0.42+/-0.06 versus 0.73+/-0.2micromol/L, respectively; p<0.001). Both ADMA (B=3.5; 95% CI: 0.5-6.5; p=0.023) and SDMA (B=1.7; 95% CI: 0.7-2.7; p=0.001) were independently related to SOFA scores. Overall, rosiglitazone treatment had no effect on ADMA levels, which only significantly differed between the rosiglitazone and control groups at day 7 (p=0.028). The SOFA score in the rosiglitazone group was lower compared to the control group but the difference was only statistically significant at day 10 (p=0.01). In conclusion, in critically ill patients plasma ADMA levels were elevated and associated with the extent of multiple organ failure, but no significant ADMA-lowering effect of the PPAR-gamma agonist rosiglitazone was observed.