A machine learning-based predictive model for the in-hospital mortality of critically ill patients with atrial fibrillation

BACKGROUND

Atrial fibrillation (AF) is common among intensive care unit (ICU) patients and significantly raises the in-hospital mortality rate. Existing scoring systems or models have limited predictive capabilities for AF patients in ICU. Our study developed and validated machine learning models to predict the risk of in-hospital mortality in ICU patients with AF.

METHODS AND RESULTS

Medical Information Mart for Intensive Care (MIMIC)-IV dataset and eICU Collaborative Research Database (eICU-CRD) were analyzed. Among ten classifiers compared, adaptive boosting (AdaBoost) showed better performance in predicting all-cause mortality in AF patients. A compact model with 15 features was developed and validated. Both the all variable and compact models exhibited excellent performance with area under the receiver operating characteristic curves (AUCs) of 1(95%confidence interval [CI]: 1.0-1.0) in the training set. In the MIMIC-IV testing set, the AUCs of the all variable and compact models were 0.978 (95% CI: 0.973-0.982) and 0.977 (95% CI: 0.972-0.982), respectively. In the external validation set, the AUCs of all variable and compact models were 0.825 (95% CI: 0.815-0.834) and 0.807 (95% CI: 0.796-0.817), respectively.

CONCLUSION

An AdaBoost-based predictive model was subjected to internal and external validation, highlighting its strong predictive capacity for assessing the risk of in-hospital mortality in ICU patients with AF.

Development and internal validation of an algorithm for estimating mortality in patients encountered by physician-staffed helicopter emergency medical services

BACKGROUND

Severity of illness scoring systems are used in intensive care units to enable the calculation of adjusted outcomes for audit and benchmarking purposes. Similar tools are lacking for pre-hospital emergency medicine. Therefore, using a national helicopter emergency medical services database, we developed and internally validated a mortality prediction algorithm.

METHODS

We conducted a multicentre retrospective observational register-based cohort study based on the patients treated by five physician-staffed Finnish helicopter emergency medical service units between 2012 and 2019. Only patients aged 16 and over treated by physician-staffed units were included. We analysed the relationship between 30-day mortality and physiological, patient-related and circumstantial variables. The data were imputed using multiple imputations employing chained equations. We used multivariate logistic regression to estimate the variable effects and performed derivation of multiple multivariable models with different combinations of variables. The models were combined into an algorithm to allow a risk estimation tool that accounts for missing variables. Internal validation was assessed by calculating the optimism of each performance estimate using the von Hippel method with four imputed sets.

RESULTS

After exclusions, 30 186 patients were included in the analysis. 8611 (29%) patients died within the first 30 days after the incident. Eleven predictor variables (systolic blood pressure, heart rate, oxygen saturation, Glasgow Coma Scale, sex, age, emergency medical services vehicle type [helicopter vs ground unit], whether the mission was located in a medical facility or nursing home, cardiac rhythm [asystole, pulseless electrical activity, ventricular fibrillation, ventricular tachycardia vs others], time from emergency call to physician arrival and patient category) were included. Adjusted for optimism after internal validation, the algorithm had an area under the receiver operating characteristic curve of 0.921 (95% CI 0.918 to 0.924), Brier score of 0.097, calibration intercept of 0.000 (95% CI -0.040 to 0.040) and slope of 1.000 (95% CI 0.977 to 1.023).

CONCLUSIONS

Based on 11 demographic, mission-specific, and physiologic variables, we developed and internally validated a novel severity of illness algorithm for use with patients encountered by physician-staffed helicopter emergency medical services, which may help in future quality improvement.

Comparative Predictive Accuracies of the Simplified Mortality Score for the Intensive Care Unit, Sepsis Severity Score, and Standard Severity Scores for 90-day Mortality in Sepsis Patients

Background

The standard severity scores were used for predicting hospital mortality of intensive care unit (ICU) patients. Recently, the new predictive score, Simplified Mortality Score for the ICU (SMS-ICU), was developed for predicting 90-day mortality.

Objective

To validate the ability of the SMS-ICU and compare with sepsis severity score (SSS) and original severity scores for predicting 90-day mortality in sepsis patients.

Method

An analysis of retrospective data was conducted in the ICU of a university teaching hospital. Also, 90-day mortality was used for the primary outcome.

Results

A total of 1,161 patients with sepsis were included. The 90-day mortality was 42.4%. The SMS-ICU presented the area under the receiver operating characteristic curve (AUROC) of 0.71, whereas the SSS had significantly higher AUROC than that of the SMS-ICU (AUROC 0.876, p < 0.001). The acute physiology and chronic health evaluation (APACHE) II and IV, and the simplified acute physiology scores (SAPS) II demonstrated good discrimination, with an AUROC above 0.90. The SMS-ICU provides poor calibration for 90-day mortality prediction, similar to the SSS and other standard severity scores. Furthermore, 90-day mortality was underestimated by the SMS-ICU, which had a standardized mortality ratio (SMR) of 1.36. The overall performance by Brier score demonstrated that the SMS-ICU was inferior to the SSS (0.222 and 0.169, respectively). Also, SAPS II presented the best overall performance with a Brier score of 0.092.

Conclusion

The SMS-ICU indicated lower performance compared to the SSS, standard severity scores. Consequently, modifications are required to enhance the performance of the SMS-ICU.

How to cite this article

Sathaporn N, Khwannimit B. Comparative Predictive Accuracies of the Simplified Mortality Score for the Intensive Care Unit, Sepsis Severity Score, and Standard Severity Scores for 90-day Mortality in Sepsis Patients. Indian J Crit Care Med 2024;28(4):343-348.

Machine-learning models are superior to severity scoring systems for the prediction of the mortality of critically ill patients in a tertiary medical center

BACKGROUND

Intensive care unit (ICU) mortality prediction helps to guide therapeutic decision making for critically ill patients. Several scoring systems based on statistical techniques have been developed for this purpose. In this study, we developed a machine-learning model to predict patient mortality in the very early stage of ICU admission.

METHODS

This study was performed with data from all patients admitted to the intensive care units of a tertiary medical center in Taiwan from 2009 to 2018. The patients' comorbidities, co-medications, vital signs, and laboratory data on the day of ICU admission were obtained from electronic medical records. We constructed random forest and extreme gradient boosting (XGBoost) models to predict ICU mortality, and compared their performance with that of traditional scoring systems.

RESULTS

Data from 12,377 patients was allocated to training (n = 9901) and testing (n = 2476) datasets. The median patient age was 70.0 years; 9210 (74.41%) patients were under mechanical ventilation in the ICU. The areas under receiver operating characteristic curves for the random forest and XGBoost models (0.876 and 0.880, respectively) were larger than those for the Acute Physiology and Chronic Health Evaluation II score (0.738), Sequential Organ Failure Assessment score (0.747), and Simplified Acute Physiology Score II (0.743). The fraction of inspired oxygen on ICU admission was the most important predictive feature across all models.

CONCLUSION

The XGBoost model most accurately predicted ICU mortality and was superior to traditional scoring systems. Our results highlight the utility of machine learning for ICU mortality prediction in the Asian population.

Risk Prediction Models for Hospital Mortality in General Medical Patients: A Systematic Review

Objective

To systematically review contemporary prediction models for hospital mortality developed or validated in general medical patients.

Methods

We screened articles in five databases, from January 1, 2010, through April 7, 2022, and the bibliography of articles selected for final inclusion. We assessed the quality for risk of bias and applicability using the Prediction Model Risk of Bias Assessment Tool (PROBAST) and extracted data using the Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modelling Studies (CHARMS) checklist. Two investigators independently screened each article, assessed quality, and extracted data.

Results

From 20,424 unique articles, we identified 15 models in 8 studies across 10 countries. The studies included 280,793 general medical patients and 19,923 hospital deaths. Models included 7 early warning scores, 2 comorbidities indices, and 6 combination models. Ten models were studied in all general medical patients (general models) and 7 in general medical patients with infection (infection models). Of the 15 models, 13 were developed using logistic or Poisson regression and 2 using machine learning methods. Also, 4 of 15 models reported on handling of missing values. None of the infection models had high discrimination, whereas 4 of 10 general models had high discrimination (area under curve >0.8). Only 1 model appropriately assessed calibration. All models had high risk of bias; 4 of 10 general models and 5 of 7 infection models had low concern for applicability for general medical patients.

Conclusion

Mortality prediction models for general medical patients were sparse and differed in quality, applicability, and discrimination. These models require hospital-level validation and/or recalibration in general medical patients to guide mortality reduction interventions.

Predictive Modeling Using Artificial Intelligence and Machine Learning Algorithms on Electronic Health Record Data: Advantages and Challenges

The rapid adoption of electronic health record (EHR) systems in US hospitals from 2008 to 2014 produced novel data elements for analysis. Concurrent innovations in computing architecture and machine learning (ML) algorithms have made rapid consumption of health data feasible and a powerful engine for clinical innovation. In critical care research, the net convergence of these trends has resulted in an exponential increase in outcome prediction research. In the following article, we explore the history of outcome prediction in the intensive care unit (ICU), the growing use of EHR data, and the rise of artificial intelligence and ML (AI) in critical care.

Machine Learning Algorithm Predicts Mortality Risk in Intensive Care Unit for Patients with Traumatic Brain Injury

BACKGROUND

Numerous mortality prediction tools are currently available to assist patients with moderate to severe traumatic brain injury (TBI). However, an algorithm that utilizes various machine learning methods and employs diverse combinations of features to identify the most suitable predicting outcomes of brain injury patients in the intensive care unit (ICU) has not yet been well-established.

METHOD

Between January 2016 and December 2021, we retrospectively collected data from the electronic medical records of Chi Mei Medical Center, comprising 2260 TBI patients admitted to the ICU. A total of 42 features were incorporated into the analysis using four different machine learning models, which were then segmented into various feature combinations. The predictive performance was assessed using the area under the curve (AUC) of the receiver operating characteristic (ROC) curve and validated using the Delong test.

RESULT

The AUC for each model under different feature combinations ranged from 0.877 (logistic regression with 14 features) to 0.921 (random forest with 22 features). The Delong test indicated that the predictive performance of the machine learning models is better than that of traditional tools such as APACHE II and SOFA scores.

CONCLUSION

Our machine learning training demonstrated that the predictive accuracy of the LightGBM is better than that of APACHE II and SOFA scores. These features are readily available on the first day of patient admission to the ICU. By integrating this model into the clinical platform, we can offer clinicians an immediate prognosis for the patient, thereby establishing a bridge for educating and communicating with family members.

Forecasting ICU Census by Combining Time Series and Survival Models

Capacity planning of ICUs is essential for effective management of health safety, quality of patient care, and the allocation of ICU resources. Whereas ICU length of stay (LOS) may be estimated using patient information such as severity of illness scoring systems, ICU census is impacted by both patient LOS and arrival patterns. We set out to develop and evaluate an ICU census forecasting algorithm using the Multiple Organ Dysfunction Score (MODS) and the Nine Equivalents of Nursing Manpower Use Score (NEMS) for capacity planning purposes.

DESIGN

Retrospective observational study.

SETTING

We developed the algorithm using data from the Medical-Surgical ICU (MSICU) at University Hospital, London, Canada and validated using data from the Critical Care Trauma Centre (CCTC) at Victoria Hospital, London, Canada.

PATIENTS

Adult patient admissions (7,434) to the MSICU and (9,075) to the CCTC from 2015 to 2021.

INTERVENTIONS

None.

MEASUREMENTS AND MAIN RESULTS

We developed an Autoregressive integrated moving average time series model that forecasts patients arriving in the ICU and a survival model using MODS, NEMS, and other factors to estimate patient LOS. The models were combined to create an algorithm that forecasts ICU census for planning horizons ranging from 1 to 7 days. We evaluated the algorithm quality using several fit metrics. The root mean squared error ranged from 2.055 to 2.890 beds/d and the mean absolute percentage error from 9.4% to 13.2%. We show that this forecasting algorithm provides a better fit when compared with a moving average or a time series model that directly forecasts ICU census. Additionally, we evaluated the performance of the algorithm using data during the global COVID-19 pandemic and found that the error of the forecasts increased proportionally with the number of COVID-19 patients in the ICU.

CONCLUSIONS

It is possible to develop accurate tools to forecast ICU census. This type of algorithm may be important to clinicians and managers when planning ICU capacity as well as staffing and surgical demand planning over a short time horizon.

Nurse staffing levels in critical care: The impact of patient characteristics

BACKGROUND

Intensive care is one of the most resource-intensive forms of care because seriously ill patients are cared for in units with high staffing levels. Studies show that the number of registered nurses (RNs) per patient and nurse education level affects patient outcome. However, there is a lack of studies that consider how nurses/patient ratio with an advanced educational level of specialized nurses in intensive care, affect the intensive care performed in different patient populations.

AIM

To investigate if differences in patient characteristics and nurse-patient ratio have an impact on the quality of care.

STUDY DESIGN

This is a retrospective observational study with a review of all patients >15 years receiving care at two general intensive care units with different nurse/patient ratio (unit A, 1:1 nurse/patient ratio and unit B, 0.5:1 nurse/patient ratio).

RESULTS

There was no significant difference in the initial severity of illness between the units. However, younger patients, male patients and patients requiring surgery entailed a higher workload and a longer intensive care unit (ICU) stay despite a 1:1 critical care nurse/patient ratio. A small difference, but not significant, with more unplanned re-intubations occurred at unit A compared with unit B.

CONCLUSION

The differences in the nurse/patient ratio did not reflect a difference in the severity of illness among admitted patients but might be explained by patient characteristics with different needs.

RELEVANCE TO CLINICAL PRACTICE

Health care managers should consider not only the number of nurses but also their educational level, specific competencies and skills mix and nursing-sensitive measures to provide high-quality ICU care in settings with different patient characteristics. Nursing-sensitive patient outcomes should be considered in relation to nurse/patient ratio, as important to measure to ensure a high quality of patient care in the ICU.

Using Second Measurement of De Ritis Ratio to Improve Mortality Prediction in Adult Trauma Patients in Intensive Care Unit

The De Ritis ratio (DRR), the ratio of serum levels of aspartate aminotransferase/alanine aminotransferase, has been reported to be a valuable biomarker in risk stratification for many liver and non-liver diseases. This study aimed to explore whether the inclusion of DRR at the date of intensive care unit (ICU) admission or days after ICU admission improves the predictive performance of various prognosis prediction models. This study reviewed 888 adult trauma patients (74 deaths and 814 survivors) in the trauma registered database between 1 January 2009, and 31 December 2020. Medical information with AST and ALT levels and derived DRR at the date of ICU admission (1st DRR) and 3-7 day after ICU admission (2nd DRR) was retrieved. Logistic regression was used to build new probability models for mortality prediction using additional DRR variables in various mortality prediction models. There was no significant difference in the 1st DRR between the death and survival patients; however, there was a significantly higher 2nd DRR in the death patients than the survival patients. This study showed that the inclusion of the additional DRR variable, measured 3-7 days after ICU admission, significantly increased the prediction performance in all studied prognosis prediction models.

Acute Kidney Injury in a Cohort of Critical Illness Patients Exposed to Non-Steroidal Anti-Inflammatory Drugs

To determine whether non-steroidal anti-inflammatory drug (NSAIDs) exposure prior to intensive care unit (ICU) admission affects the development of acute kidney injury (AKI) with renal replacement therapy (RRT). An administrative database is used to establish a cohort of patients who were admitted to the ICU. The exposure to NSAIDs that the patients had before admission to the ICU is determined. Demographic variables, comorbidities, AKI diagnoses requiring RRT, and pneumonia during the ICU stay are also measured. Multivariate logistic regression and inverse probability weighting (IPW) are used to calculate risks of exposure to NSAIDs for patients with AKI requiring RRT. In total, 96,235 patients were admitted to the ICU, of which 16,068 (16.7%) were exposed to NSAIDs. The incidence of AKI with RRT was 2.71% for being exposed to NSAIDs versus 2.24% for those not exposed (p < 0.001). For the outcome of AKI, the odds ratio weighted with IPW was 1.28 (95% CI: 1.15−1.43), and for the outcome of pneumonia as a negative control, the odds ratio was 1.07 (95% CI: 0.98−1.17). The impact of prior exposure to NSAIDs over critically ill patients in the development of AKI is calculated as 8 patients per 1000 exposures. The negative control with the same sources of bias did not show an association with NSAID exposure.

Ability of a modified Sequential Organ Failure Assessment score to predict mortality among sepsis patients in a resource-limited setting

Copyright © 2022 The Korean Society ofCritical Care MedicineThis is an Open Access article distributedunder the terms of Creative Attributions Non-Commercial License (https://creativecommons.org/li-censes/by-nc/4.0/) which permitsunrestricted noncommercial use, distribution,and reproduction in any medium, provided theoriginal work is properly cited.https://www.accjournal.org 363INTRODUCTIONSepsis is a life-threatening condition and constitutes major health care problems around the world [1,2]. Sepsis was associated with nearly 20% of all global deaths, and the majority of sepsis cases occurred in low- or middle-income countries [1]. In 2017, the World Health Organization recommended actions to reduce the global burden of sepsis [2]. Sepsis has been defined as acute life-threatening organ dysfunction due to dysregulation of host responses toBackground: Some variables of the Sequential Organ Failure Assessment (SOFA) score are not routinely measured in sepsis patients, especially in countries with limited resources. Therefore, this study was conducted to evaluate the accuracy of the modified SOFA (mSOFA) and compared its ability to predict mortality in sepsis patients to that of the original SOFA score.Methods: Sepsis patients admitted to the medical intensive care unit of Songklanagarind Hospital between 2011 and 2018 were retrospectively analyzed. The primary outcome was all-cause in-hospital mortality.Results: A total of 1,522 sepsis patients were enrolled. The mean SOFA and mSOFA scores were 9.7±4.3 and 8.8±3.9, respectively. The discrimination of the mSOFA score was significantly higher than that of the SOFA score for all-cause in-hospital mortality (area under the receiver operating characteristic curve, 0.891 [95% confidence interval, 0.875–0.907] vs. 0.879 [0.862–0.896]; P<0.001), all-cause intensive care unit (ICU) mortality (0.880 [0.863–0.898] vs. 0.871 [0.853–0.889], P=0.01) and all-cause 28-day mortality (0.887 [0.871–0.904] vs. 0.874 [0.856–0.892], P<0.001). The ability of mSOFA score to predict all-cause in-hospital and 28-day mortality was higher than that of the SOFA score within the subgroups of sepsis according to age, sepsis severity and serum lactate levels. The mSOFA score was demonstrated to have a performance similar to the original SOFA score regarding the prediction of mortality in sepsis patients with cirrhosis or hepatic dysfunction.Conclusions: The mSOFA score was a good alternative to the original SOFA core in predicting mortality among sepsis patients admitted to the ICU.

Probiotic Supplementation Prevents the Development of Ventilator-Associated Pneumonia for Mechanically Ventilated ICU Patients: A Systematic Review and Network Meta-analysis of Randomized Controlled Trials

Background

Ventilator-associated pneumonia (VAP) is one of the common critical complications of nosocomial infection (NI) in invasive mechanically ventilated intensive care unit (ICU) patients. The efficacy of total parenteral nutrition (TPN), enteral nutrition and/or adjuvant peripheral parenteral nutrition (EPN) supplemented with or without probiotic, prebiotic, and synbiotic therapies in preventing VAP among these patients has been questioned. We aimed to systematically and comprehensively summarize all available studies to generate the best evidence of VAP prevention for invasive mechanically ventilated ICU patients.

Methods

Randomized controlled trials (RCTs) for the administration of TPN, EPN, probiotics-supplemented EPN, prebiotics-supplemented EPN, and synbiotics-supplemented EPN for VAP prevention in invasive mechanically ventilated ICU patients were systematically retrieved from four electronic databases. The incidence of VAP was the primary outcome and was determined by the random-effects model of a Bayesian framework. The secondary outcomes were NI, ICU and hospital mortality, ICU and hospital length of stay, and mechanical ventilation duration. The registration number of Prospero is CRD42020195773.

Results

A total of 8339 patients from 31 RCTs were finally included in network meta-analysis. The primary outcome showed that probiotic-supplemented EPN had a higher correlation with the alleviation of VAP than EPN in critically invasive mechanically ventilated patients (odds ratio [OR] 0.75; 95% credible intervals [CrI] 0.58–0.95). Subgroup analyses showed that probiotic-supplemented EPN prevented VAP in trauma patients (OR 0.30; 95% CrI 0.13–0.83), mixed probiotic strain therapy was more effective in preventing VAP than EPN therapy (OR 0.55; 95% CrI 0.31–0.97), and low-dose probiotic therapy (less than 10¹⁰ CFU per day) was more associated with lowered incidence of VAP than EPN therapy (OR 0.16; 95% CrI 0.04–0.64). Secondary outcomes indicated that synbiotic-supplemented EPN therapy was more significantly related to decreased incidence of NI than EPN therapy (OR 0.34; 95% CrI 0.11–0.85). Prebiotic-supplemented EPN administration was the most effective in preventing diarrhea (OR 0.05; 95% CrI 0.00–0.71).

Conclusion

Probiotic supplementation shows promise in reducing the incidence of VAP in critically invasive mechanically ventilated patients. Currently, low quality of evidence reduces strong clinical recommendations. Further high-quality RCTs are needed to conclusively prove these findings.

Systamatic Review Registration

[https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020195773], identifier [CRD42020195773].

Predict models for prolonged ICU stay using APACHE II, APACHE III and SAPS II scores: A Japanese multicenter retrospective cohort study

Prolonged ICU stays are associated with high costs and increased mortality. Thus, early prediction of such stays would help clinicians to plan initial interventions, which could lead to efficient utilization of ICU resources. The aim of this study was to develop models for predicting prolonged stays in Japanese ICUs using APACHE II, APACHE III and SAPS II scores. In this multicenter retrospective cohort study, we analyzed the cases of 85,558 patients registered in the Japanese Intensive care Patient Database between 2015 and 2019. Prolonged ICU stay was defined as an ICU stay of >14 days. Multivariable logistic regression analyses were performed to develop three predictive models for prolonged ICU stay using APACHE II, APACHE III and SAPS II scores, respectively. After exclusions, 79,620 patients were analyzed, 2,364 of whom (2.97%) experienced prolonged ICU stays. Multivariable logistic regression analyses showed that severity scores, BMI, MET/RRT, postresuscitation, readmission, length of stay before ICU admission, and diagnosis at ICU admission were significantly associated with higher risk of prolonged ICU stay in all models. The present study developed predictive models for prolonged ICU stay using severity scores. These models may be helpful for efficient utilization of ICU resources.

General Severity of Illness Scoring Systems and COVID-19 Mortality Predictions: Is "Old Still Gold?"

How to cite this article: Siddiqui SS, Patnaik R, Kulkarni AP. General Severity of Illness Scoring Systems and COVID-19 Mortality Predictions: Is "Old Still Gold?" Indian J Crit Care Med 2022;26(4):416-418.

Prognostic Value of an Estimate-of-Risk Model in Critically Ill Obstetric Patients in Brazil

The CIPHER (Collaborative Integrated Pregnancy High-Dependency Estimate of Risk) prognostic model was not predictive of risk of death, prolonged organ support, or lifesaving intervention among critically ill patients in Brazil.

To externally validate the CIPHER (Collaborative Integrated Pregnancy High-Dependency Estimate of Risk) prognostic model for pregnant and postpartum women admitted to the intensive care unit.

Withholding or withdrawing life support in long-term neurointensive care patients: a single-centre, prospective, observational pilot study

PURPOSE

Scarce evidence exists regarding end-of-life decision (EOLD) in neurocritically ill patients. We investigated the factors associated with EOLD making, including the group and individual characteristics of involved healthcare professionals, in a multiprofessional neurointensive care unit (NICU) setting.

MATERIALS AND METHODS

A prospective, observational pilot study was conducted between 2013 and 2014 in a 10-bed NICU. Factors associated with EOLD in long-term neurocritically ill patients were evaluated using an anonymised survey based on a standardised questionnaire.

RESULTS

8 (25%) physicians and 24 (75%) nurses participated in the study by providing their 'treatment decisions' for 14 patients at several time points. EOLD was 'made' 44 (31%) times, while maintenance of life support 98 (69%) times. EOLD patterns were not significantly different between professional groups. The individual characteristics of the professionals (age, gender, religion, personal experience with death of family member and NICU experience) had no significant impact on decisions to forgo or maintain life-sustaining therapy. EOLD was patient-specific (intraclass correlation coefficient: 0.861), with the presence of acute life-threatening disease (OR (95% CI): 18.199 (1.721 to 192.405), p=0.038) and low expected patient quality of life (OR (95% CI): 9.276 (1.131 to 76.099), p=0.016) being significant and independent determinants for withholding or withdrawing life-sustaining treatment.

CONCLUSIONS

Our findings suggest that EOLD in NICU relies mainly on patient prognosis and not on the characteristics of the healthcare professionals.

Cohort study of the APACHE II score and mortality for different types of intensive care unit patients

OBJECTIVES

Find the discriminant and calibration of APACHE II (Acute Physiology And Chronic Health Evaluation) score to predict mortality for different type of intensive care unit (ICU) patients.

METHODS

This is a cohort retrospective study using secondary data of ICU patients admitted to Siloam Hospital of Lippo Village from 2014 to 2018 with minimum age ≥17 years. The analysis uses the receiver operating characteristic curve, student t-test and logistic regression to find significant variables needed to predict mortality.

RESULTS

A total of 2181 ICU patients: men (55.52%) and women (44.48%) with an average age of 53.8 years old and length of stay 3.92 days were included in this study. Patients were admitted from medical emergency (30.5%), neurosurgical (52.1%) and surgical (17.4%) departments, with 10% of mortality proportion. Patients admitted from the medical emergency had the highest average APACHE score, 23.14±8.5, compared with patients admitted from neurosurgery 15.3±6.6 and surgical 15.8±6.8. The mortality rate of patients from medical emergency (24.5%) was higher than patients from neurosurgery (3.5%) or surgical (5.3%) departments. Area under curve of APACHE II score showed 0.8536 (95% CI 0.827 to 0.879). The goodness of fit Hosmer-Lemeshow show p=0.000 with all ICU patients' mortality; p=0.641 with medical emergency, p=0.0001 with neurosurgical and p=0.000 with surgical patients.

CONCLUSION

APACHE II has a good discriminant for predicting mortality among ICU patients in Siloam Hospital but poor calibration score. However, it demonstrates poor calibration in neurosurgical and surgical patients while demonstrating adequate calibration in medical emergency patients.

Comparison of prognosis predictive value of 4 disease severity scoring systems in patients with acute respiratory failure in intensive care unit: A STROBE report

Various disease severity scoring systems were currently used in critically ill patients with acute respiratory failure, while their performances were not well investigated.The study aimed to investigate the difference in prognosis predictive value of 4 different disease severity scoring systems in patients with acute respiratory failure.With a retrospective cohort study design, adult patients admitted to intensive care unit (ICU) with acute respiratory failure were screened and relevant data were extracted from an open-access American intensive care database to calculate the following disease severity scores on ICU admission: acute physiology score (APS) III, Sequential Organ Failure Assessment score (SOFA), quick SOFA (qSOFA), and Oxford Acute Severity of Illness Score (OASIS). Hospital mortality was chosen as the primary outcome. Multivariable logistic regression analyses were performed to analyze the association of each scoring system with the outcome. Receiver operating characteristic curve analyses were conducted to evaluate the prognosis predictive performance of each scoring system.A total of 4828 patients with acute respiratory failure were enrolled with a hospital mortality rate of 16.78%. APS III (odds ratio [OR] 1.03, 95% confidence interval [CI] 1.02-1.03), SOFA (OR 1.15, 95% CI 1.12-1.18), qSOFA (OR 1.26, 95% CI 1.11-1.42), and OASIS (OR 1.06, 95% CI 1.05-1.08) were all significantly associated with hospital mortality after adjustment for age and comorbidities. Receiver operating characteristic analyses showed that APS III had the highest area under the curve (AUC) (0.703, 95% CI 0.683-0.722), and SOFA and OASIS shared similar predictive performance (area under the curve 0.653 [95% CI 0.631-0.675] and 0.664 [95% CI 0.644-0.685], respectively), while qSOFA had the worst predictive performance for predicting hospital mortality (0.553, 95% CI 0.535-0.572).These results suggested the prognosis predictive value varied among the 4 different disease severity scores for patients admitted to ICU with acute respiratory failure.

Intensive Care Unit Scoring Systems

BACKGROUND

Illness severity scoring systems are commonly used in critical care. When applied to the populations for whom they were developed and validated, these tools can facilitate mortality prediction and risk stratification, optimize resource use, and improve patient outcomes.

OBJECTIVE

To describe the characteristics and applications of the scoring systems most frequently applied to critically ill patients.

METHODS

A literature search was performed using MEDLINE to identify original articles on intensive care unit scoring systems published in the English language from 1980 to 2020. Search terms associated with critical care scoring systems were used alone or in combination to find relevant publications.

RESULTS

Two types of scoring systems are most frequently applied to critically ill patients: those that predict risk of in-hospital mortality at the time of intensive care unit admission (Acute Physiology and Chronic Health Evaluation, Simplified Acute Physiology Score, and Mortality Probability Models) and those that assess and characterize current degree of organ dysfunction (Multiple Organ Dysfunction Score, Sequential Organ Failure Assessment, and Logistic Organ Dysfunction System). This article details these systems' differing features and timing of use, score calculation, patient populations, and comparative performance data.

CONCLUSION

Critical care nurses must be aware of the strengths, limitations, and specific characteristics of severity scoring systems commonly used in intensive care unit patients to effectively employ these tools in clinical practice and critically appraise research findings based on their use.

Perioperative clinical parameters associated with short-term mortality after colorectal perforation

Purpose

Although early prediction of mortality is useful for the management of patients with colorectal perforations, no significant perioperative predictive factors have been identified. The purpose of this study was to identify useful prognostic factors for patients with colorectal perforation.

Methods

This single-center retrospective study included consecutive patients undergoing emergency surgery for colorectal perforation from January 2012 to December 2019. The primary outcome was combined 30 day and in-hospital mortality. Patient- and disease-related factors obtained perioperatively were evaluated for mortality prediction. A scoring system was developed to enhance clinical utility.

Results

Overall, 146 patients were included and 20 (14%) died after surgery. Multivariate logistic regression identified five predictive factors: age, hemodialysis, uncommon perforation etiology, plasma albumin level, and decreased platelet count. The area under the receiver operating curve for the scoring system using these parameters was 0.894 (95% CI 0.835–0.952). Patients at high-risk of mortality were classified by the proposed score with a sensitivity of 90.0% and negative predictive value of 98.0%.

Conclusion

This study identified five perioperative factors significantly associated with mortality of patients with colorectal perforation. Although these parameters predict mortality of patients with colorectal perforation using a score with high discrimination, further study is required to confirm these findings.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00068-021-01719-8.

Ventilator-associated bacterial pneumonia in coronavirus 2019 disease, a retrospective monocentric cohort study

Introduction

Severe coronavirus 2019 disease (CoViD-19) may lead to respiratory failure and mechanical ventilation. Therefore, ventilator associated pneumonia (VAP) may complicate the course of the disease. The aim of the current article was to investigate possible predictive factors for bacterial VAP on a retrospective manner, in a cohort of mechanically ventilated CoViD-19 patients. Additionally, determinant factors of lethality were analyzed.

Methods

Medical records of patients hospitalized in the intensive care units (ICU) at the university hospital UZ Brussel during the epidemic were reviewed. VAP was defined following the National Healthcare Safety Network 2017 criteria. Univariate and multivariate logistic regressions analyses were performed.

Results

Among the 39 patients included in the study, 54% were diagnosed with bacterial VAP. Case fatality rate was 44%, but 59% of the deceased patients had a do-not-resuscitate status. Multivariate logistic regression for prediction of VAP showed significant differences in duration of ICU hospitalization and in minimal lung compliance.

Additional analyses were performed on CoViD-19 patients who were affected by bacterial respiratory superinfection. The responsible pathogens correspond to the commonly found bacteria in VAP. However, 71% of the isolated germs were multi-drug resistant and bacteraemia was reported in 38%. Multivariate analyses for prediction of lethality found significant difference in SOFA score.

Conclusions

Mechanically ventilated CoViD-19 patients might frequently develop VAP. Longer ICU hospitalization was associated with pulmonary superinfection in the current cohort. Moreover, decreased minimal lung compliance was correlated to VAP and higher SOFA score at VAP diagnosis was associated with lethality.

Characteristics and outcomes of patients admitted to adult intensive care units in Hong Kong: a population retrospective cohort study from 2008 to 2018

Background

Globally, mortality rates of patients admitted to the intensive care unit (ICU) have decreased over the last two decades. However, evaluations of the temporal trends in the characteristics and outcomes of ICU patients in Asia are limited. The objective of this study was to describe the characteristics and risk adjusted outcomes of all patients admitted to publicly funded ICUs in Hong Kong over a 11-year period. The secondary objective was to validate the predictive performance of Acute Physiology And Chronic Health Evaluation (APACHE) IV for ICU patients in Hong Kong.

Methods

This was an 11-year population-based retrospective study of all patients admitted to adult general (mixed medical-surgical) intensive care units in Hong Kong public hospitals. ICU patients were identified from a population electronic health record database. Prospectively collected APACHE IV data and clinical outcomes were analysed.

Results

From 1 April 2008 to 31 March 2019, there were a total of 133,858 adult ICU admissions in Hong Kong public hospitals. During this time, annual ICU admissions increased from 11,267 to 14,068, whilst hospital mortality decreased from 19.7 to 14.3%. The APACHE IV standard mortality ratio (SMR) decreased from 0.81 to 0.65 during the same period. Linear regression demonstrated that APACHE IV SMR changed by − 0.15 (95% CI − 0.18 to − 0.11) per year (Pearson’s R = − 0.951, p < 0.001). Observed median ICU length of stay was shorter than that predicted by APACHE IV (1.98 vs. 4.77, p < 0.001). C-statistic for APACHE IV to predict hospital mortality was 0.889 (95% CI 0.887 to 0.891) whilst calibration was limited (Hosmer–Lemeshow test p < 0.001).

Conclusions

Despite relatively modest per capita health expenditure, and a small number of ICU beds per population, Hong Kong consistently provides a high-quality and efficient ICU service. Number of adult ICU admissions has increased, whilst adjusted mortality has decreased over the last decade. Although APACHE IV had good discrimination for hospital mortality, it overestimated hospital mortality of critically ill patients in Hong Kong.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40560-020-00513-9.

Clinical Characteristics and Risk Factors for Critically Ill Patients with Carbapenem-Resistant Klebsiella pneumonia e (CrKP): A Cohort Study from Developing Country

BACKGROUND

Increasing evidence indicates carbapenem-resistant Klebsiella pneumoniae (CrKP) is increasingly prevalent in intensive care unit (ICU), but its clinical characteristics and risk factors remain unknown.

AIM

The aim of the present study was to evaluate clinical characteristics, risk factors in critically ill patients with CrKP infection.

METHODS

A retrospective study was included in patients from January 2013 to October 2019. Clinical data were collected from CrKP patients on the day of specimen collection admitted to ICU. Multivariable logistic regression was used for risk factors. Receiver operating curve (ROC) and the area under the curve (AUC) with DeLong method of MedCalc software were used. Two-way repeated-measures ANOVA analysis was used to analyze the characteristics of independent risk factors over time.

FINDINGS

A total of 147 adult patients with CrKP were screened, among them, 89 (median age 64.0 years, 66 (74.15%) males) patients with CrKP were finally included, of which 38 patients (42.7%) were non-survival group. Multivariate logistic regression analysis indicated that lactic acid (OR3.04 95% CI 1.38-6.68, P = 0.006), APACHE II score (OR 1.20, 95% CI 1.09-1.33, P < 0.001), tigecycline combined with fosfomycin treatment (OR0.15, 95% CI 0.04-0.65, P = 0.011) are independent risk factors for 28-day mortality in patients with CRKP infection (P<0.05). Combined lactic acid with APACHE II score could predict 28-day mortality, of which AUC value was 0.916 (95% CI, 0.847-0.985), with sensitivity 0.76 and specificity 0.98. ANOVA analysis showed that APACHE II score and lactic acid between the two groups at three-time points were statistically significant, which interactive with time and showed an upward and downward trend with time (P < 0.05).

CONCLUSION

Therapeutic strategy based on improving lactic acid and APACHE II would contribute to the outcome in patients with CrKP infection. Tigecycline combined with fosfomycin could reduce the 28-day mortality in patients with CrKP infection in developing country.

Performance of PRISM III, PELOD-2, and P-MODS Scores in Two Pediatric Intensive Care Units in China

Objective: The performances of the pediatric risk of mortality score III (PRISM III), pediatric logistic organ dysfunction score-2 (PELOD-2), and pediatric multiple organ dysfunction score (P-MODS) in Chinese patients are unclear. This study aimed to assess the performances of these scores in predicting mortality in critically ill pediatric patients. Methods: This retrospective observational study was conducted at two tertiary-care PICUs of teaching hospitals in China. A total of 1,253 critically ill pediatric patients admitted to the two Pediatric Intensive Care Units (PICUs) of the First Affiliated Hospital, Sun Yat-Sen University from August 2014 to December 2019 and Shen-Zhen Children's Hospital from January 2019 to December 2019 were analyzed. The indexes of discrimination and calibration were applied to evaluate score performance for the three models (PRISM III, PELOD-2, and P-MODS scores). The receiver operating characteristic (ROC) curve was plotted, and the efficiency of PRISM III, PELOD-2, and P-MODS in predicting death were evaluated by the area under ROC curve (AUC). Hosmer-Lemeshow goodness-of-fit test was used to evaluate the degree of fitting between the mortality predictions of each scoring system and the actual mortality. Results: A total of 1,253 pediatric patients were eventually enrolled in this study (median age, 38 months; overall mortality rate, 8.9%; median length of PICU stay, 8 days). Compared to the survival group, the non-survival group showed significantly higher PRISM III, PELOD-2, and P-MODS scores [PRISM III: 18 (12, 23) vs. 11 (0, 16); PELOD-2, 8 (4, 10) vs. 4 (0, 6); and P-MODS: 5 (4, 9) vs. 3 (0, 4), all P < 0.001]. ROC curve analysis showed that the AUCs of PRISM III, PELOD-2, and P-MODS for predicting the death of critically ill children were 0.858, 0.721, and 0.596, respectively. Furthermore, in the Hosmer-Lemeshow goodness-of-fit test, PRISM III and PELOD-2 showed the better calibration between predicted mortality and observed mortality (PRISM III: χ² = 5.667, P = 0.368; PELOD-2: χ² = 9.582, P = 0.276; P-MODS: χ² = 12.449, P = 0.015). Conclusions: PRISM III and PELOD-2 can discriminate well between survivors and non-survivors. PRISM III and PELOD-2 showed the better calibration between predicted and observed mortality, while P-MODS showed poor calibration.

Accuracy of Postoperative Risk Scores for Survival Prediction in Interagency Registry for Mechanically Assisted Circulatory Support Profile 1 Continuous-Flow Left Ventricular Assist Device Recipients

In this study, we sought to determine the accuracy of several critical care risk scores for predicting survival of Interagency Registry for Mechanically Assisted Circulatory Support (INTERMACS) Profile 1 patients after continuous-flow left ventricular assist device (CF-LVAD) placement. We retrospectively analyzed the records of 605 patients who underwent CF-LVAD implantation between 2003 and 2016. We calculated the preoperative HeartMate II Risk Score (HMRS) and preoperative Right Ventricular Failure Risk Score (RVFRS) and the following risk scores for postoperative days 1-5: HMRS, RVFRS, Model for End-stage Liver Disease (MELD), MELD-eXcluding International Normalized Ratio, Post Cardiac Surgery (POCAS) risk score, Sequential Organ Failure Assessment (SOFA) risk score, and Acute Physiology and Chronic Health Evaluation III. The preoperative scores and the postoperative day 1, 5-day mean, and 5-day maximum scores were entered into a receiver operating characteristic curve analysis to examine accuracy for predicting 30-day, 90-day, and 1-year survival. The mean POCAS score was the best predictor of 30-day and 90-day survival (area under the curve [AUC] = 0.869 and 0.816). The postoperative mean RVFRS was the best predictor of 1-year survival (AUC = 0.7908). The postoperative maximum and mean RVFRS and HMRS were more accurate than the preoperative scores. Both of these risk score measurements of acuity in the postoperative intensive care unit setting help predict early mortality after LVAD implantation.

Prognostic association of routinely measured biomarkers in patients admitted to critical care: a systematic review

PURPOSE

To examine reported prognostic associations of routine blood measurements in the intensive care unit.

MATERIALS AND METHODS

We searched PubMed, EMBASE through 28^th May 2020 to identify all studies in adult critical care investigating associations between parameters measured routinely in whole blood, plasma or serum, and length of stay or mortality. Registration: PROSPERO; CRD42019122058.

RESULTS

A total of 128 studies, reporting 28 different putative prognostic biomarkers, met eligibility criteria. Those most frequently examined were red cell distribution width, neutrophil-to-lymphocyte ratio, C-reactive protein, and platelet count. A higher red cell distribution width, a lower platelet count, and a higher neutrophil-to-lymphocyte ratio were consistently associated with both increased mortality and length of stay. A lower level of albumin was consistently associated with greater mortality. C-reactive protein was inconsistent. Most studies (n = 110) used regression modelling with wide variation in variable selection and covariate-adjustment; none externally validated the proposed predictive models.

CONCLUSIONS

Simple regression models have so far proved inadequate for the complexity of data available from routine blood sampling in critical care. Adoption of a direct causal framework may help better assess mechanistic processes, aid design of future studies, and guide clinical decision making using routine data.

Overall and subgroup specific performance of the pediatric index of mortality 2 score in Switzerland: a national multicenter study

Pediatric Index of Mortality (PIM) 2 score is used in pediatric intensive care unit (PICU) to predict the patients' risk of death. The performance of this model has never been assessed in Switzerland. The aim of this study was to evaluate the performance of the PIM2 score in the whole cohort and in pre-specified diagnostic subgroups of patients admitted to PICUs in Switzerland. All children younger than 16 years admitted to any PICU in Switzerland between January 1, 2012 and December 31, 2017 were included in the study. A total of 22,382 patients were analyzed. Observed mortality was 2%, whereas mortality predicted by PIM2 was 4.2% (SMR = 0.47, 95% CI, 0.42-0.52). Calibration was also poor across the deciles of mortality risks (p < 0.001). The AUC-ROC for the entire cohort was 0.88 (95% CI, 0.87-0.90). Calibration varied significantly according to primary diagnosis.Conclusion: The performance of the PIM 2 score in a cohort of Swiss patients is poor with adequate discrimination and poor calibration. The PIM 2 score tends to under predict the number of deaths among septic patients and in patients admitted after a cardiorespiratory arrest. What is Known: •PIM2 score is a widely used mortality prediction model in PICU. •PIM2 performance among uncommon but clinically relevant diagnostic subgroups of patients is unknown. •The performance of PIM2 score has never been assessed in Switzerland. What is New: •The performance of the PIM 2 score in a cohort of Swiss patients is poor with adequate discrimination and poor calibration. •Calibration varies significantly according to primary diagnosis. The PIM 2 score under predict the number of deaths among septic patients and in patients admitted after a cardiorespiratory arrest.

PRISM 4-C: An Adapted PRISM IV Algorithm for Children With Cancer

We evaluated the performance of PRISM IV for pediatric cancer patients, and adapted and calibrated the algorithm to calculate mortality probabilities for these patients. An ambidirectional cohort was used, and data were collected from March 2017 prospectively to April 2019, and retrospectively to November 2014. The derivation set for model building contained 500 patients, and a validation set of 503 patients. Risk variables for hospital death were tested in logistic regression models encompassing PRISM IV components. There were 128 deaths (12.7%), being 65 deaths in the validation set. In the validation set, the PRISM IV algorithm had an area under the receiver operating characteristic curve of 0.89, with P=0.13 by Hosmer-Lemeshow test, and predicted 33 of the 65 deaths for a standardized mortality rate of 1.8 (95% confidence interval, 1.4-2.9; P<0.001 by Mid-P test). Our modified algorithm had an area under the receiver operating characteristic curve of 0.93, with P=0.3 by Hosmer-Lemeshow test and an standardized mortality rate of 1.02 (95% confidence interval, 0.79-1.19). The modified algorithm predicted 63.7 of 65 deaths. The PRISM IV algorithm was a poor predictor of mortality in children with cancer. The modified algorithm improved the predictive performance.

End-of-life care in patients with a highly transmissible respiratory virus: implications for COVID-19

Symptom management and end-of-life care are core skills for all physicians, although in ordinary times many anesthesiologists have fewer occasions to use these skills. The current coronavirus disease (COVID-19) pandemic has caused significant mortality over a short time and has necessitated an increase in provision of both critical care and palliative care. For anesthesiologists deployed to units caring for patients with COVID-19, this narrative review provides guidance on conducting goals of care discussions, withdrawing life-sustaining measures, and managing distressing symptoms.

Biomarkers in critical care nutrition

The goal of nutrition support is to provide the substrates required to match the bioenergetic needs of the patient and promote the net synthesis of macromolecules required for the preservation of lean mass, organ function, and immunity. Contemporary observational studies have exposed the pervasive undernutrition of critically ill patients and its association with adverse clinical outcomes. The intuitive hypothesis is that optimization of nutrition delivery should improve ICU clinical outcomes. It is therefore surprising that multiple large randomized controlled trials have failed to demonstrate the clinical benefit of restoring or maximizing nutrient intake. This may be in part due to the absence of biological markers that identify patients who are most likely to benefit from nutrition interventions and that monitor the effects of nutrition support. Here, we discuss the need for practical risk stratification tools in critical care nutrition, a proposed rationale for targeted biomarker development, and potential approaches that can be adopted for biomarker identification and validation in the field.

What factors predict length of stay in the intensive care unit? Systematic review and meta-analysis

PURPOSE

Studies have shown that a small percentage of ICU patients have prolonged length of stay (LoS) and account for a large proportion of resource use. Therefore, the identification of prolonged stay patients can improve unit efficiency. In this study, we performed a systematic review and meta-analysis to understand the risk factors of ICU LoS.

MATERIALS AND METHODS

We searched MEDLINE, Embase and Scopus databases from inception to November 2018. The searching process focused on papers presenting risk factors of ICU LoS. A meta-analysis was performed for studies reporting appropriate statistics.

RESULTS

From 6906 citations, 113 met the eligibility criteria and were reviewed. A meta-analysis was performed for six factors from 28 papers and concluded that patients with mechanical ventilation, hypomagnesemia, delirium, and malnutrition tend to have longer stay, and that age and gender were not significant factors.

CONCLUSIONS

This work suggested a list of risk factors that should be considered in prediction models for ICU LoS, as follows: severity scores, mechanical ventilation, hypomagnesemia, delirium, malnutrition, infection, trauma, red blood cells, and PaO2:FiO2. Our findings can be used by prediction models to improve their predictive capacity of prolonged stay patients, assisting in resource allocation, quality improvement actions, and benchmarking analysis.

External validation of a prognostic model for intensive care unit mortality: a retrospective study using the Ontario Critical Care Information System

PURPOSE

To externally validate an intensive care unit (ICU) mortality prediction model that was created using the Ontario Critical Care Information System (CCIS), which includes the Multiple Organ Dysfunction Score (MODS).

METHODS

We applied the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) recommendations to a prospective longitudinal cohort of patients discharged between 1 July 2015 and 31 December 31 2016 from 90 adult level-3 critical care units in Ontario. We used multivariable logistic regression with measures of discrimination, calibration-in-the-large, calibration slope, and flexible calibration plots to compare prediction model performance of the entire data set and for each ICU subtype.

RESULTS

Among 121,201 CCIS records with ICU mortality of 11.3%, the C-statistic for the validation data set was 0.805. The C-statistic ranged from 0.775 to 0.846 among the ICU subtypes. After intercept recalibration to adjust the baseline risk, the mean predicted risk of death matched actual ICU mortality. The calibration slope was close to 1 with all CCIS data and ICU subtypes of cardiovascular and community hospitals with low ventilation rates. Calibration slopes significantly less than 1 were found for ICUs in teaching hospitals and community hospitals with high ventilation rates whereas coronary care units had a calibration slope significantly higher than 1. Calibration plots revealed over-prediction in high risk groups to a varying degree across all cohorts.

CONCLUSIONS

A risk prediction model primarily based on the MODS shows reproducibility and transportability after intercept recalibration. Risk adjusting models that use existing and feasible data collection can support performance measurement at the individual ICU level.

Gene Expression Patterns Distinguish Mortality Risk in Patients with Postsurgical Shock

Nowadays, mortality rates in intensive care units are the highest of all hospital units. However, there is not a reliable prognostic system to predict the likelihood of death in patients with postsurgical shock. Thus, the aim of the present work is to obtain a gene expression signature to distinguish the low and high risk of death in postsurgical shock patients. In this sense, mRNA levels were evaluated by microarray on a discovery cohort to select the most differentially expressed genes between surviving and non-surviving groups 30 days after the operation. Selected genes were evaluated by quantitative real-time polymerase chain reaction (qPCR) in a validation cohort to validate the reliability of data. A receiver-operating characteristic analysis with the area under the curve was performed to quantify the sensitivity and specificity for gene expression levels, which were compared with predictions by established risk scales, such as acute physiology and chronic health evaluation (APACHE) and sequential organ failure assessment (SOFA). IL1R2, CD177, RETN, and OLFM4 genes were upregulated in the non-surviving group of the discovery cohort, and their predictive power was confirmed in the validation cohort. This work offers new biomarkers based on transcriptional patterns to classify the postsurgical shock patients according to low and high risk of death. The results present more accuracy than other mortality risk scores.

Inter-Rater Reliability and Impact of Disagreements on Acute Physiology and Chronic Health Evaluation IV Mortality Predictions

Acute Physiology and Chronic Health Evaluation is a well-validated method to risk-adjust ICU patient outcomes. However, predictions may be affected by inter-rater reliability for manually entered elements. We evaluated inter-rater reliability for Acute Physiology and Chronic Health Evaluation IV manually entered elements among clinician abstractors and assessed the impacts of disagreements on mortality predictions.

A Framework to Understand the Progression of Cardiovascular Disease for Type 2 Diabetes Mellitus Patients Using a Network Approach

The prevalence of chronic disease comorbidity has increased worldwide. Comorbidity-i.e., the presence of multiple chronic diseases-is associated with adverse health outcomes in terms of mobility and quality of life as well as financial burden. Understanding the progression of comorbidities can provide valuable insights towards the prevention and better management of chronic diseases. Administrative data can be used in this regard as they contain semantic information on patients' health conditions. Most studies in this field are focused on understanding the progression of one chronic disease rather than multiple diseases. This study aims to understand the progression of two chronic diseases in the Australian health context. It specifically focuses on the comorbidity progression of cardiovascular disease (CVD) in patients with type 2 diabetes mellitus (T2DM), as the prevalence of these chronic diseases in Australians is high. A research framework is proposed to understand and represent the progression of CVD in patients with T2DM using graph theory and social network analysis techniques. Two study cohorts (i.e., patients with both T2DM and CVD and patients with only T2DM) were selected from an administrative dataset obtained from an Australian health insurance company. Two baseline disease networks were constructed from these two selected cohorts. A final disease network from two baseline disease networks was then generated by weight adjustments in a normalized way. The prevalence of renal failure, fluid and electrolyte disorders, hypertension and obesity was significantly higher in patients with both CVD and T2DM than patients with only T2DM. This showed that these chronic diseases occurred frequently during the progression of CVD in patients with T2DM. The proposed network-based model may potentially help the healthcare provider to understand high-risk diseases and the progression patterns between the recurrence of T2DM and CVD. Also, the framework could be useful for stakeholders including governments and private health insurers to adopt appropriate preventive health management programs for patients at a high risk of developing multiple chronic diseases.

Arterial blood pressure correlates with 90-day mortality in sepsis patients: a retrospective multicenter derivation and validation study using high-frequency continuous data

OBJECTIVE

To identify the outcome of patients with sepsis using high-frequency blood pressure data.

MATERIALS AND METHODS

This retrospective observational study was conducted at a university hospital ICU (derivation study) and at two urban hospitals (validation study) with data from adult sepsis patients who visited the centers during the same period. The area under the curve (AUC) of blood pressure falling below threshold was calculated. The predictive 90-day mortality (primary endpoint) area under threshold (AUT) and critical blood pressure were calculated as the maximum area under the curve of the receiver operating characteristic curve (AUCROC) and the threshold minus average AUT (derivation study), respectively. For the validation study, the derived 90-day mortality AUCROC (using critical blood pressure) was compared with Sequential Organ Failure Assessment (SOFA), Simplified Acute Physiology Score (SAPS) II, Acute Physiology and Chronic Health Evaluation (APACHE) II, and APACHE III.

RESULTS

Derivation cohort (N = 137): the drop area from the mean blood pressure of 70 mmHg at 24-48 hours most accurately predicted 90-day mortality [critical blood pressure, 67.8 mmHg; AUCROC, 0.763; 95% confidence interval (CI), 0.653-0.890]. Validation cohort (N = 141): the 90-day mortality AUCROC (0.776) compared with the AUCROC for SOFA (0.711), SAPSII (0.771), APACHE II (0.745), and APACHE III (0.710) was not significantly different from the critical blood pressure 67.8 mmHg (P = 0.420).

CONCLUSION

High-frequency arterial blood pressure data of the period and extent of blood pressure depression can be useful in predicting the clinical outcomes of patients with sepsis.

SAPS3-CNIV score to predict hospital mortality following noninvasive ventilation: methodology insights

We have read with great interest and congratulate Huseiniet al. [1] on the external validation of the SAPS3-CNIV score (Simplified Acute Physiology Score 3 customised for noninvasive ventilation) to predict hospital mortality. Huseiniet al. [1] concluded that SAPS3-CNIV did not improve prediction of mortality in patients over SAPS3. This scoring system included the SAPS3, haemoglobin, carbon dioxide tension, lactate, do not resuscitate (DNR) orders and aetiology of respiratory failure. However, we consider that there are some key aspects that need to be taken into account for a proper clinical extrapolation.

Further work towards the development of scores to predict hospital mortality is warranted, to overcome the methodological limitations of the SAPS3-CNIVhttp://bit.ly/2QE1LoB

Mid-regional pro-adrenomedullin predicts poor outcome in non-selected patients admitted to an intensive care unit

Background Mortality risk and outcome in critically ill patients can be predicted by scoring systems, such as APACHE and SAPS. The identification of prognostic biomarkers, simple to measure upon admission to an intensive care unit (ICU) is an open issue. The aim of this observational study was to assess the prognostic value of plasma mid-regional pro-adrenomedullin (MR-proADM) at ICU admission in non-selected patients in comparison to Acute Physiology and Chronic Health Evaluation II (APACHEII) and Simplified Acute Physiology Score II (SAPSII) scores. Methods APACHEII and SAPSII scores were calculated after 24 h from ICU admission. Plasma MR-proADM levels were measured by TRACE-Kryptor on admission (T0) and after 24 h (T24). The primary endpoint was intra-hospital mortality; secondary endpoint was length of stay (LOS). Results One hundred and twenty-six consecutive non-selected patients admitted to an ICU were enrolled. Plasma MR-proADM levels were correlated with LOS (r=0.28; p=0.0014 at T0; r=0.26; p=0.005 at T24). Multivariate analysis showed that T0 MR-proADM was a significant predictor of mortality (odds ratio [OR]: 1.27; 95% confidence interval [95%CI]: 1.03-1.55; p=0.022). Receiver operating characteristic curves analysis revealed that MR-proADM on ICU admission identified non-survivors with high accuracy, not inferior to the one of APACHEII and SAPSII scores (area under the curve [AUC]: 0.71; 95%CI: 0.62-0.78; p=0.0002 for MR-proADM; AUC: 0.71; 95%CI: 0.62-0.79; p<0.0001 for APACHEII; AUC: 0.8; 95%CI: 0.71-0.87; p<0.0001 for SAPSII). Conclusions Our findings point out a role of MR-proADM as a prognostic tool in non-selected patients in ICUs being a reliable predictor of mortality and LOS and support its use on admission to an ICU to help the management of critically ill patients.

THAI-ICU score as a simplified severity score for critically ill patients in a resource limited setting: Result from SEA-AKI study group

PURPOSE

To create a simplified ICU scoring system to predict mortality in critically ill patients that can be feasibly applied in resource limited setting with good performance of predicting hospital mortality.

MATERIALS AND METHODS

A retrospective study from prospective cohort was created consisting of adult patients who were admitted to an ICU of 17 centers across Thailand from 2013 to 2015. A development cohort (n = 3503) and a validation cohort (n = 1909) were randomly selected from the available enrollment data.

RESULTS

In the development cohort, the predictors of the simplified score 6 variable model were low Glasgow coma score (GCS), low mean arterial pressure or need vasopressor, positive net-fluid balance, tachypnea, thrombocytopenia, and high blood urea nitrogen. In the validation study of THAI-ICU, AUC (95%CI) was 0.81(0.78-0.83). At the optimum cutoff value of 9; the sensitivity, specificity, positive likelihood ratio were 72%, 73%, and 2.72 respectively. The Hosmer-Lemeshow - C statistic was 13.5 (p = .2) and the Brier score 95% CI was 0.16 (0.15, 0.17).

CONCLUSIONS

The THAI-ICU score is a new simplified severity score for predicting hospital mortality. The simplicity of the score will increase the possibility to apply in resource limited settings.

Performance of Critical Care Outcome Prediction Models in an Intermediate Care Unit

BACKGROUND

Intermediate care units (IMCUs) are heterogeneous in design and operation, which makes comparative effectiveness studies challenging. A generalizable outcome prediction model could improve such comparisons. However, little is known about the performance of critical care outcome prediction models in the intermediate care setting. The purpose of this study is to evaluate the performance of the Acute Physiology and Chronic Health Evaluation version II (APACHE II), Simplified Acute Physiology Score version II (SAPS II) and version 3 (SAPS 3), and Mortality Probability Model version III (MPM₀III) in patients admitted to a well-characterized IMCU.

MATERIALS AND METHODS

In the IMCU of an academic medical center (July to December 2012), the discrimination and calibration of each outcome prediction model were evaluated using the area under the receiver-operating characteristic and Hosmer-Lemeshow goodness-of-fit test, respectively. Standardized mortality ratios (SMRs) were also calculated.

RESULTS

The cohort included data from 628 unique IMCU admissions with an inpatient mortality rate of 8.3%. All models exhibited good discrimination, but only the SAPS II and MPM₀III were well calibrated. While the APACHE II and SAPS 3 both markedly overestimated mortality, the SMR for the SAPS II and MPM₀III were 0.91 and 0.91, respectively.

CONCLUSIONS

The SAPS II and MPM₀III exhibited good discrimination and calibration, with slight overestimation of mortality. Each model should be further evaluated in multicenter studies of patients in the intermediate care setting.

Semi-quantitative visual assessment of chest radiography is associated with clinical outcomes in critically ill patients

Background

Respiratory pathology is a major driver of mortality in the intensive care unit (ICU), even in the absence of a primary respiratory diagnosis. Prior work has demonstrated that a visual scoring system applied to chest radiographs (CXR) is associated with adverse outcomes in ICU patients with Acute Respiratory Distress Syndrome (ARDS). We hypothesized that a simple, semi-quantitative CXR score would be associated with clinical outcomes for the general ICU population, regardless of underlying diagnosis.

Methods

All individuals enrolled in the Registry of Critical Illness at Brigham and Women’s Hospital between June 2008 and August 2018 who had a CXR within 24 h of admission were included. Each patient’s CXR was assigned an opacification score of 0–4 in each of four quadrants with the total score being the sum of all four quadrants. Multivariable negative binomial, logistic, and Cox regression, adjusted for age, sex, race, immunosuppression, a history of chronic obstructive pulmonary disease, a history of congestive heart failure, and APACHE II scores, were used to assess the total score’s association with ICU length of stay (LOS), duration of mechanical ventilation, in-hospital mortality, 60-day mortality, and overall mortality, respectively.

Results

A total of 560 patients were included. Higher CXR scores were associated with increased mortality; for every one-point increase in score, in-hospital mortality increased 10% (OR 1.10, CI 1.05–1.16, p < 0.001) and 60-day mortality increased by 12% (OR 1.12, CI 1.07–1.17, p < 0.001). CXR scores were also independently associated with both ICU length of stay (rate ratio 1.06, CI 1.04–1.07, p < 0.001) and duration of mechanical ventilation (rate ratio 1.05, CI 1.02–1.07, p < 0.001).

Conclusions

Higher values on a simple visual score of a patient’s CXR on admission to the medical ICU are associated with increased in-hospital mortality, 60-day mortality, overall mortality, length of ICU stay, and duration of mechanical ventilation.

Development and Evaluation of an Automated Machine Learning Algorithm for In-Hospital Mortality Risk Adjustment Among Critical Care Patients

OBJECTIVES

Risk adjustment algorithms for ICU mortality are necessary for measuring and improving ICU performance. Existing risk adjustment algorithms are not widely adopted. Key barriers to adoption include licensing and implementation costs as well as labor costs associated with human-intensive data collection. Widespread adoption of electronic health records makes automated risk adjustment feasible. Using modern machine learning methods and open source tools, we developed and evaluated a retrospective risk adjustment algorithm for in-hospital mortality among ICU patients. The Risk of Inpatient Death score can be fully automated and is reliant upon data elements that are generated in the course of usual hospital processes.

SETTING

One hundred thirty-one ICUs in 53 hospitals operated by Tenet Healthcare.

PATIENTS

A cohort of 237,173 ICU patients discharged between January 2014 and December 2016.

DESIGN

The data were randomly split into training (36 hospitals), and validation (17 hospitals) data sets. Feature selection and model training were carried out using the training set while the discrimination, calibration, and accuracy of the model were assessed in the validation data set.

MEASUREMENTS AND MAIN RESULTS

Model discrimination was evaluated based on the area under receiver operating characteristic curve; accuracy and calibration were assessed via adjusted Brier scores and visual analysis of calibration curves. Seventeen features, including a mix of clinical and administrative data elements, were retained in the final model. The Risk of Inpatient Death score demonstrated excellent discrimination (area under receiver operating characteristic curve = 0.94) and calibration (adjusted Brier score = 52.8%) in the validation dataset; these results compare favorably to the published performance statistics for the most commonly used mortality risk adjustment algorithms.

CONCLUSIONS

Low adoption of ICU mortality risk adjustment algorithms impedes progress toward increasing the value of the healthcare delivered in ICUs. The Risk of Inpatient Death score has many attractive attributes that address the key barriers to adoption of ICU risk adjustment algorithms and performs comparably to existing human-intensive algorithms. Automated risk adjustment algorithms have the potential to obviate known barriers to adoption such as cost-prohibitive licensing fees and significant direct labor costs. Further evaluation is needed to ensure that the level of performance observed in this study could be achieved at independent sites.

Developing well-calibrated illness severity scores for decision support in the critically ill

Illness severity scores are regularly employed for quality improvement and benchmarking in the intensive care unit, but poor generalization performance, particularly with respect to probability calibration, has limited their use for decision support. These models tend to perform worse in patients at a high risk for mortality. We hypothesized that a sequential modeling approach wherein an initial regression model assigns risk and all patients deemed high risk then have their risk quantified by a second, high-risk-specific, regression model would result in a model with superior calibration across the risk spectrum. We compared this approach to a logistic regression model and a sophisticated machine learning approach, the gradient boosting machine. The sequential approach did not have an effect on the receiver operating characteristic curve or the precision-recall curve but resulted in improved reliability curves. The gradient boosting machine achieved a small improvement in discrimination performance and was similarly calibrated to the sequential models.

Improving mortality models in the ICU with high-frequency data

BACKGROUND

Assessment of the performance of Intensive Care Units (ICU) is of vital importance for an effective healthcare system. Such assessment ensures that the limited resources of the healthcare system are allocated where they are most needed. Severity scoring systems are employed for this purpose and improving these systems is a continuing area of research which has focused on the use of more complex techniques and new variables.

OBJECTIVES

This paper investigates whether scoring systems could be improved through use of metrics which better summarise the high frequency data collected by automated systems for patients in the ICU.

METHODS AND DATA

3128 admissions to the Gold Coast University Hospital ICU are used to construct three logistic regressions based on the most widely used scoring system (APACHE III) to compare performance with and without predictors leveraging available high frequency information. Performance is assessed based on model accuracy, calibration, and discrimination. High frequency information was considered for existing pulse and mean arterial pressure physiology fields and resulting models compared against a baseline logistic regression using only APACHE III physiology variables.

RESULTS

Model discrimination and accuracy were better for models which included high frequency predictors, with calibration remaining good in all cases. The most influential high frequency summaries were the number of turning points in a patient's mean arterial pressure or pulse in the first 24 h of ICU admission.

CONCLUSIONS

The findings indicate that scoring systems can be improved by better accounting for high frequency data.

Factors associated with physicians' predictions of six-month mortality in critically ill patients

Background

Physician's estimates of a patient's prognosis are an important component in shared decision-making. However, the variables influencing physician's judgments are not well understood. We aimed to determine which physician and patient factors are associated with physicians' predictions of critically ill patients' six-month mortality and the accuracy and confidence of these predictions.

Methods

Prospective cohort study evaluating physicians' predictions of six-month mortality. Using univariate and multivariable generalized estimating equations, we assessed the association between baseline physician and patient characteristics with predictions of six-month death, as well as accuracy and confidence of these predictions.

Results

Our cohort was comprised 300 patients and 47 physicians. Physicians were asked to predict if patients would be alive or dead at six months and to report their confidence in these predictions. Physicians predicted that 99 (33%) patients would die. The key factors associated with both the direction and accuracy of prediction were older age of the patient, the presence of malignancy, being in a medical ICU, and higher APACHE III scores. The factors associated with lower confidence included older physician age, being in a medical ICU and higher APACHE III score.

Conclusions

Patient level factors are associated with predictions of mortality at six months. The accuracy and confidence of the predictions are associated with both physician and patients' factors. The influence of these factors should be considered when physicians reflect on how they make predictions for critically ill patients.