Accurate Predictor Of Mortality Research Articles

E-CatBoost: An efficient machine learning framework for predicting ICU mortality using the eICU Collaborative Research Database.

Improving the Intensive Care Unit (ICU) management network and building cost-effective and well-managed healthcare systems are high priorities for healthcare units. Creating accurate and explainable mortality prediction models helps identify the most critical risk factors in the patients’ survival/death status and early detect the most in-need patients. This study proposes a highly accurate and efficient machine learning model for predicting ICU mortality status upon discharge using the information available during the first 24 hours of admission. The most important features in mortality prediction are identified, and the effects of changing each feature on the prediction are studied. We used supervised machine learning models and illness severity scoring systems to benchmark the mortality prediction. We also implemented a combination of SHAP, LIME, partial dependence, and individual conditional expectation plots to explain the predictions made by the best-performing model (CatBoost). We proposed E-CatBoost, an optimized and efficient patient mortality prediction model, which can accurately predict the patients’ discharge status using only ten input features. We used eICU-CRD v2.0 to train and validate the models; the dataset contains information on over 200,000 ICU admissions. The patients were divided into twelve disease groups, and models were fitted and tuned for each group. The models’ predictive performance was evaluated using the area under a receiver operating curve (AUROC). The AUROC scores were 0.86 [std:0.02] to 0.92 [std:0.02] for CatBoost and 0.83 [std:0.02] to 0.91 [std:0.03] for E-CatBoost models across the defined disease groups; if measured over the entire patient population, their AUROC scores were 7 to 18 and 2 to 12 percent higher than the baseline models, respectively. Based on SHAP explanations, we found age, heart rate, respiratory rate, blood urine nitrogen, and creatinine level as the most critical cross-disease features in mortality predictions.

ISeeU2: Visually interpretable mortality prediction inside the ICU using deep learning and free-text medical notes

Accurate mortality prediction allows Intensive Care Units (ICUs) to adequately benchmark clinical practice and identify patients with unexpected outcomes. Traditionally, simple statistical models have been used to assess patient death risk, many times with sub-optimal performance. On the other hand Deep Learning holds promise to positively impact clinical practice by leveraging medical data to assist diagnosis and prediction, including mortality prediction. However, as the question of whether powerful Deep Learning models attend correlations backed by sound medical knowledge when generating predictions remains open, additional interpretability tools are needed to foster trust and encourage the use of AI by clinicians. In this work we show an interpretable Deep Learning model trained on MIMIC-III to predict mortality inside the ICU using raw nursing notes, together with visual explanations for word importance based on the Shapley Value. Our model reaches a ROC of 0.8629 (±0.0058), outperforming the traditional SAPS-II score and a LSTM recurrent neural network baseline while providing enhanced interpretability when compared with similar Deep Learning approaches. Supporting code can be found at https://github.com/williamcaicedo/ISeeU2.

Identification of Risk Factors and Development of Predictive Risk Score Model for Mortality after Open Ruptured Abdominal Aortic Aneurysm Repair.

Background and Objectives: Despite the relatively large number of publications concerning the validation of these models, there is currently no solid evidence that they can be used with absolute precision to predict survival. The goal of this study is to identify preoperative factors that influenced 30-day mortality and to create a predictive model after open ruptured abdominal aortic aneurysm (RAAA) repair. Materials and Methods: This was a retrospective single-center cohort study derived from a prospective collected database, between 1 January 2009 and 2016. Multivariate logistic regression analysis was used to identify all significant predictive factors. Variables that were identified in the multivariate analysis were dichotomized at standard levels, and logistic regression was used for the analysis. To ensure that dichotomized variables were not overly simplistic, the C statistic was evaluated for both dichotomized and continuous models. Results: There were 500 patients with complete medical data included in the analysis during the study period. Of them, 37.6% were older than 74 years, and 83.8% were males. Multivariable logistic regression showed five variables that were predictive of mortality: age > 74 years (OR = 4.01, 95%CI 2.43–6.26), loss of consciousness (OR = 2.21, 95%CI 1.11–4.40), previous myocardial infarction (OR = 2.35, 95%CI 1.19–4.63), development of ventricular arrhythmia (OR = 4.54, 95%CI 1.75–11.78), and DAP < 60 mmHg (OR = 2.32, 95%CI 1.17–4.62). Assigning 1 point for each variable, patients were stratified according to the preoperative RAAA mortality risk score (range 0–5). Patients with 1 point suffered 15.3% mortality and 3 points 68.2% mortality, while all patients with 5 points died. Conclusions: This preoperative RAAA score identified risk factors readily assessed at the bedside and provides an accurate prediction of 30-day mortality after open repair of RAAA.

The Association between Medical Utilization and Chronic Obstructive Pulmonary Disease Severity: A Comparison of the 2007 and 2011 Guideline Staging Systems.

(1) Background: This study aimed to investigate the associations between the Global Initiative for Chronic Obstructive Lung Disease (GOLD) staging systems, medical costs, and mortality among patients with chronic obstructive lung disease (COPD). Predictions of the effectiveness of the two versions of the staging systems were also compared. (2) Purpose: this study investigated the associations between the Global Initiative for Chronic Obstructive Lung Disease (GOLD) staging systems, medical costs, and mortality among patients with COPD. Predicting effectiveness between the two versions of the staging systems was also compared. (3) Procedure: This study used a secondary clinical database of a medical center in central Taiwan to examine records between 2011 and 2017. A total of 613 patients with COPD were identified. The independent variables comprised the COPD GOLD Guideline staging of the 2007 and 2011 versions, demographic characteristics, health status, and physician seniority. The dependent variables included total medical cost, average length of hospital stay, and mortality. The statistical methods included binomial logistic regression and the general linear model (GLM). (4) Discussion: The total medical cost during the observation period for patients with COPD averaged TWD 292,455.6. The average length of hospital stay was 9.7 days. The mortality rate was 9.6%, compared with that of patients in Grade 1 of the 2007 version; patients in Grade 4 of the 2007 version had significantly higher odds of death (OR = 4.07, p = 0.02). The accuracy of mortality prediction for both the 2007 and 2011 versions of the staging was equal, at 90.4%. The adjusted GLM analysis revealed that patients in Group D of the 2011 version had a significantly longer length of hospital stay than those in Group A of the 2011 version (p = 0.04). No difference between the 2007 and 2011 versions was found regarding the total medical cost. Complications were significantly associated with the total medical cost and average length of hospital stay. (5) Conclusions: The COPD staging 2011 version was associated with an average length of hospital stay, whereas the COPD staging 2007 version was related to mortality risk. Therefore, the 2011 version can estimate the length of hospital stay. However, in predicting prognosis and mortality, the 2007 version is recommended.

Using Machine Learning to Develop and Validate an In-Hospital Mortality Prediction Model for Patients with Suspected Sepsis.

Background: Early recognition of sepsis and the prediction of mortality in patients with infection are important. This multi-center, ED-based study aimed to develop and validate a 28-day mortality prediction model for patients with infection using various machine learning (ML) algorithms. Methods: Patients with acute infection requiring intravenous antibiotic treatment during the first 24 h of admission were prospectively recruited. Patient demographics, comorbidities, clinical signs and symptoms, laboratory test data, selected sepsis-related novel biomarkers, and 28-day mortality were collected and divided into training (70%) and testing (30%) datasets. Logistic regression and seven ML algorithms were used to develop the prediction models. The area under the receiver operating characteristic curve (AUROC) was used to compare different models. Results: A total of 555 patients were recruited with a full panel of biomarker tests. Among them, 18% fulfilled Sepsis-3 criteria, with a 28-day mortality rate of 8%. The wrapper algorithm selected 30 features, including disease severity scores, biochemical parameters, and conventional and few sepsis-related biomarkers. Random forest outperformed other ML models (AUROC: 0.96; 95% confidence interval: 0.93–0.98) and SOFA and early warning scores (AUROC: 0.64–0.84) in the prediction of 28-day mortality in patients with infection. Additionally, random forest remained the best-performing model, with an AUROC of 0.95 (95% CI: 0.91–0.98, p = 0.725) after removing five sepsis-related novel biomarkers. Conclusions: Our results demonstrated that ML models provide a more accurate prediction of 28-day mortality with an enhanced ability in dealing with multi-dimensional data than the logistic regression model.

Prognostic value of serum/plasma neurofilament light chain for COVID-19-associated mortality.

ObjectiveGiven the continued spread of coronavirus 2, the early predictors of coronavirus disease 19 (COVID‐19) associated mortality might improve patients' outcomes. Increased levels of circulating neurofilament light chain (NfL), a biomarker of neuronal injury, have been observed in severe COVID‐19 patients. We investigated whether NfL provides non‐redundant clinical value to previously identified predictors of COVID‐19 mortality.MethodsWe measured serum or plasma NfL concentrations in a blinded fashion in 3 cohorts totaling 338 COVID‐19 patients.ResultsIn cohort 1, we found significantly elevated NfL levels only in critically ill COVID‐19 patients. Longitudinal cohort 2 data showed that NfL is elevated late in the course of the disease, following the two other prognostic markers of COVID‐19: decrease in absolute lymphocyte count (ALC) and increase in lactate dehydrogenase (LDH). Significant correlations between ALC and LDH abnormalities and subsequent rise of NfL implicate that the multi‐organ failure is the most likely cause of neuronal injury in severe COVID‐19 patients. The addition of NfL to age and gender in cohort 1 significantly improved the accuracy of mortality prediction and these improvements were validated in cohorts 2 and 3.InterpretationA substantial increase in serum/plasma NfL reproducibly enhanced COVID‐19 mortality prediction. Combined with other prognostic markers, such as ALC and LDH that are routinely measured in ICU patients, NfL measurements might be useful to identify the patients at a high risk of COVID‐19‐associated mortality, who might still benefit from escalated care.

Discussion summary: MELD 3.0: The Model for End-Stage Liver Disease Updated for the Modern Era

BACKGROUND &amp; AIMS: The Model for End-Stage Liver Disease (MELD) has been established as a reliable indicator of shortterm survival in patients with end-stage liver disease. The current version (MELDNa), consisting of the international normalized ratio and serum bilirubin, creatinine, and sodium, has been used to determine organ allocation priorities for liver transplantation in the United States. The objective was to optimize MELD further by taking into account additional variables and updating coefficients with contemporary data. METHODS: All candidates registered on the liver transplant wait list in the US national registry from January 2016 through December 2018 were included. Uni- and multivariable Cox models were developed to predict survival up to 90 days after wait list registration. Model fit was tested using the concordance statistic (C-statistic) and reclassification, and the Liver Simulated Allocation Model was used to estimate the impact of replacing MELDNa with the new model. RESULTS: The final multivariable model was characterized by (1) additional variables of female sex and serum albumin, (2) interactions between bilirubin and sodium and between albumin and creatinine, and (3) an upper bound for creatinine at 3.0 mg/dL. The final model (MELD 3.0) had better discrimination than MELDNa (C-statistic, 0.869 vs 0.862; P &lt; .01). Importantly, MELD 3.0 correctly reclassified a net of 8.8% of decedents to a higher MELD tier, affording them a meaningfully higher chance of transplantation, particularly in women. In the Liver Simulated Allocation Model analysis, MELD 3.0 resulted in fewer wait list deaths compared to MELDNa (7788 vs 7850; P= .02). CONCLUSION: MELD 3.0 affords more accurate mortality prediction in general than MELDNa and addresses determinants of wait list outcomes, including the sex disparity.

Chest X-ray Severity Score as a Putative Predictor of Clinical Outcome in Hospitalized Patients: An Experience From a Vietnamese COVID-19 Field Hospital.

BackgroundThrough the coronavirus disease 2019 (COVID-19) pandemic, portable radiography was particularly useful for assessing and monitoring the COVID-19 disease in Vietnamese field hospitals. It provides a convenient and precise picture of the progression of the disease. The purpose of this study was to evaluate the predictive value of chest radiograph reporting systems (Brixia and total severity score (TSS)) and the National Early Warning Score (NEWS) clinical score in a group of hospitalized patients with COVID-19.MethodsThis retrospective cohort study used routinely collected clinical data from polymerase chain reaction (PCR)-positive severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) patients admitted to Field Hospital District 8, Ho Chi Minh City, Vietnam, from August 2021 to September 2021. The initial chest radiographs were scored based on the TSS and Brixia scoring systems to quantify the extent of lung involvement. After the chest radiograph score was reported, two residents calculated the rate of all-cause in-hospital mortality with the consultation of expert radiologists. In this study, NEWS2 scores on hospital admission were calculated. The gradient boosting machines (GBMs) and Shapley additive exPlanations (SHAP) were applied to access the important variable and improve the accuracy of mortality prediction. The adjusted odds ratio for predictor was presented by univariate analysis and multivariate analysis.ResultsThe chest X-rays (CXRs) at the admission of 273 patients (mean age 59 years +/-16, 42.1% were male) were scored. In the univariate analysis, age, vaccination status, previous disease, NEWS2, a saturation of peripheral oxygen (Sp02), the Brixia and TSS scores were significant predictors of mortality (p-value < 0.05). In multivariate analysis, there were statistically significant differences in mortality between age, Sp02, Brixia score, and patients with previous diseases were independent predictors of mortality and hospitalization.A gradient boosting machine was performed in the train data set, which showed that the best hyperparameters for predicting the mortality of patients are the Brixia score (exclude TSS score). In the top five predictors, an increase in Brixia, age, and BMI increased the logarithmic number of probability clarifying as death status. Although the TSS and Brixia scores evaluated chest imaging, the TSS score was not essential as the Brixia score (rank 6/11). It was clear that the BMI and NEWS2 score was positively correlated with the Brixia score, and age did not affect this correlation. Meanwhile, we did not find any trend between the TSS score versus BMI and NEWS2.ConclusionWhen integrated with the BMI and NEWS2 clinical classification systems, the severity score of COVID-19 chest radiographs, particularly the Brixia score, was an excellent predictor of all-cause in-hospital mortality.

Risk factors of the severe course and fatal outcome in COVID-19

Since its first detection, coronavirus disease 2019 (COVID-19) caused by coronavirus SARS-CoV-2 (severe acute respiratory syndrome coronavirus 2) infection has spread rapidly around the world. Although coronavirus SARS-CoV-2 primarily targets the respiratory system, complications in other organ systems (cardiovascular, neurological, and renal) can also contribute to death from the disease. Clinical experience thus far has shown substantial heterogeneity in the trajectory of SARS-CoV-2 infection, spanning from asymptomatic to mild, moderate, and severe disease forms with low survival rates. Accurate prediction of COVID-19 mortality and the identification of contributing factors would allow for targeted strategies in patients with the high risk of death. We aimed to identify clinical and laboratory features that contributed the most to this prediction. An improved understanding of predictive factors for COVID-19 is crucial for identify those with higher risk of mortality and for clinical decision making to reduce the risk of death. The main risk factors for the severe course of COVID-19, the development of complications and death include old age, concomitant diseases (cardiovascular diseases, chronic lung diseases, diabetes mellitus and hypertension), body temperature 37.8C, oxygen saturation 92%, quantitative and functional depletion of innate immunity, bilateral pulmonary infiltrates, increased levels of laboratory parameters of systemic inflammation, respiratory, cardiac, renal and/or hepatic failure. Proper assessment of prognostic factors and careful monitoring to ensure the necessary interventions at the appropriate time in high-risk patients can reduce the fatality rate from COVID-19.

The BUrn Mortality Prediction (BUMP) Score – An improved mortality prediction score based on data of the German burn registry

BackgroundBurn injuries constitute the fourth most common injuries globally. Patient outcomes must be currently assessed to provide appropriate patient care with high quality standards. However, existing mortality prediction scoring methods have been shown to lack accuracy in current burn patient populations. Therefore, this study aimed to validate existing scores using current patient data and assess whether new prediction parameters can provide better accuracy. MethodsA retrospective analysis of the patient data from the German Burn Registry between 2016 and 2019 was performed to evaluate all Abbreviated Burn Severity Index (ABSI) score parameters. All patients over 16 years of age who received intensive care were included. Descriptive statistics and logistic regression analysis were used to identify novel prediction parameters based on the parameters documented at admission and establish a new prediction score, the BUrn Mortality Prediction (BUMP) score. The quality of the new score was subsequently compared to that of the original ABSI, modified ABSI, Galeiras, Revised Baux score and TIMM. The new prediction score was then validated using patient data collected in the German Burn Registry in 2020. ResultsIn total, 7276 patients were included. Age; the presence of at least two comorbidities; burn injuries caused by work-related accidents, traffic accidents and suicide attempts; total burn surface area; inhalation trauma and full-thickness burns were identified as independent significant predictors of mortality (p < 0.001). Additionally, we evaluated new age groups to improve prediction accuracy. The number of comorbidities (p < 0.001) and the aetiology (burns occurring at work [p = 0.028], burns caused by traffic accidents [p < 0.001] or burns due to attempted suicide [p < 0.001]) had a significant influence on mortality. The BUMP score, which was developed based on these parameters, showed the best fitness and showed more accurate mortality prediction than all the above-mentioned scores (area under the receiver operating characteristic curve: 0.947 [0.939–0.954] compared to 0.926 [0.915–0.936], 0.928 [0.918–0.939], 0.937 [0.928–0.947], 0.939 [0.930–0.948], 0.940 [0.932–0.949] respectively). ConclusionsA novel score (BUMP score) was developed for the purpose of external quality assessment of burn centres participating in the German burn registry, where observed and expected outcomes are compared on a hospital level, and for scientifically applications. The clinical impact of this score and its generalisability to other patient populations needs to be evaluated.

Predicting mortality in elderly trauma patients: a review of the current literature.

Advances in medical care allow patients to live longer, translating into a larger geriatric patient population. Adverse outcomes increase with older age, regardless of injury severity. Age, comorbidities, and physiologic deterioration have been associated with the increased mortality seen in geriatric trauma patients. As such, outcome prediction models are critical to guide clinical decision making and goals of care discussions for this population. The purpose of this review was to evaluate the various outcome prediction models for geriatric trauma patients. There are several prediction models used for predicting mortality in elderly trauma patients. The Geriatric Trauma Outcome Score (GTOS) is a validated and accurate predictor of mortality in geriatric trauma patients and performs equally if not better to traditional scores such as the Trauma and Injury Severity Score. However, studies recommend medical comorbidities be included in outcome prediction models for geriatric patients to further improve performance. The ideal outcome prediction model for geriatric trauma patients has not been identified. The GTOS demonstrates accurate predictive ability in elderly trauma patients. The addition of medical comorbidities as a variable in outcome prediction tools may result in superior performance; however, additional research is warranted.

A proof-of-concept study on mortality prediction with machine learning algorithms using burn intensive care data.

IntroductionBurn injuries are a common traumatic injury. Large burns have high mortality requiring intensive care and accurate mortality predictions. To assess if machine learning (ML) could improve predictions, ML algorithms were tested and compared with the original and revised Baux score.MethodsAdmission data and mortality outcomes were collected from patients at Uppsala University Hospital Burn Centre from 2002 to 2019. Prognostic variables were selected, ML algorithms trained and predictions assessed by analysis of the area under the receiver operating characteristic curve (AUC). Comparison was made with Baux scores using DeLong testResultsA total of 17 prognostic variables were selected from 92 patients. AUCs in leave-one-out cross-validation for a decision tree model, an extreme boosting model, a random forest model, a support-vector machine (SVM) model and a generalised linear regression model (GLM) were 0.83 (95% confidence interval [CI] = 0.72–0.94), 0.92 (95% CI = 0.84–1), 0.92 (95% CI = 0.84–1), 0.92 (95% CI = 0.84–1) and 0.84 (95% CI = 0.74–0.94), respectively. AUCs for the Baux score and revised Baux score were 0.85 (95% CI = 0.75–0.95) and 0.84 (95% CI = 0.74–0.94). No significant differences were observed when comparing ML algorithms with Baux score and revised Baux score. Secondary variable selection was made to analyse model performance.ConclusionThis proof-of-concept study showed initial credibility in using ML algorithms to predict mortality in burn patients. The sample size was small and future studies are needed with larger sample sizes, further variable selections and prospective testing of the algorithms.Lay SummaryBurn injuries are one of the most common traumatic injuries especially in countries with limited prevention and healthcare resources. To treat a patient with large burns who has been admitted to an intensive care unit, it is often necessary to assess the risk of a fatal outcome. Physicians traditionally use simplified scores to calculate risks. One commonly used score, the Baux score, uses age of the patient and the size of the burn to predict the risk of death. Adding the factor of inhalation injury, the score is then called the revised Baux score. However, there are a number of additional causes that can influence the risk of fatal outcomes that Baux scores do not take into account. Machine learning is a method of data modelling where the system learns to predict outcomes based on previous cases and is a branch of artificial intelligence. In this study we evaluated several machine learning methods for outcome prediction in patients admitted for burn injury. We gathered data on 93 patients at admission to the intensive care unit and our experiments show that machine learning methods can reach an accuracy comparable with Baux scores in calculating the risk of fatal outcomes. This study represents a proof of principle and future studies on larger patient series are required to verify our results as well as to evaluate the methods on patients in real-life situations.

Nutritional assessment of patients with aneurysmal subarachnoid hemorrhage using the modified "Nutrition Risk in the Critically Ill" score, and its association with outcomes.

Introduction: subarachnoid hemorrhage (SAH) is a rare and life-threatening cerebrovascular disease. Mitigating the factors that compromise patient recovery during neurocritical care due to SAH is of clinical benefit. Objectives: to evaluate the nutritional risk of patients with aneurysmal subarachnoid hemorrhage using "The Modified Nutrition Risk in the Critically Ill" (mNUTRIC) score, and examine its association with outcomes such as mortality, time of mechanical ventilation, and functional status among survivors. Methods: we designed a cross-sectional study. Patients with SAH admitted to the neurointensive critical care unit (neuroICU) in a tertiary care public hospital were eligible. The inclusion criteria were a minimum stay in the intensive care unit (ICU) of 24 hrs for subarachnoid hemorrhage from a nontraumatic, spontaneously ruptured cerebral aneurysm, and hospital admission within 24 hrs after the onset of symptoms. Results: high nutritional risk as stratified by the mNUTRIC score was associated with discharge type (OR = 0.346; 95 % CI = 0.182-0.650; p = 0.001), acute hypertensive hydrocephalus (OR = 4.371; 95 % CI = 2.283-8.549; p < 0.001), and functional outcome (OR = 0.106; 95 % CI = 0.025-0.0388; p < 0.001). The mNUTRIC score was significantly different among median age (p < 0.001), length of stay in the neuroICU (p = 0.005), SOFA score (p < 0.001), and APACHE II score (p < 0.001) categories. Conclusions: this study demonstrated an association between nutritional risk assessment and outcomes such as length of stay in the neuroICU, type of discharge, functional status, and mortality prediction accuracy.

A Machine Learning Pipeline for Mortality Prediction in the ICU

ABSTRACT Mortality risk prediction for patients admitted into the intensive care unit (ICU) is a crucial and challenging task, so that clinicians are able to respond with timely and appropriate clinical intervention. This becomes more urgent under the background of COVID-19 as a global pandemic. In recent years, electronic health records (EHR) have been widely adopted, and have the potential to greatly improve clinical services and diagnostics. However, the large proportion of missing data in EHR poses challenges that may reduce the accuracy of prediction methods. We propose a cohort study that builds a pipeline that extracts ICD-9 codes and laboratory tests from public available electronic ICU databases, and improve the in-hospital mortality prediction accuracy using a combination of neural network missing data imputation approach and decision tree based outcome prediction algorithm. We show the proposed approach achieves a higher area under the ROC curve, ranging from 0.88-0.98, compared with other well-known machine learning methods applied to similar target population. It also offers clinical interpretations through variable selection. Our analysis also shows that mortality prediction for neonates was more challenging than for adults, and that prediction accuracy decreases as patients stayed longer in the ICU.

Using Artificial Neural Network Condensation to Facilitate Adaptation of Machine Learning in Medical Settings by Reducing Computational Burden: Model Design and Evaluation Study.

BackgroundMachine learning applications in the health care domain can have a great impact on people’s lives. At the same time, medical data is usually big, requiring a significant number of computational resources. Although this might not be a problem for the wide adoption of machine learning tools in high-income countries, the availability of computational resources can be limited in low-income countries and on mobile devices. This can limit many people from benefiting from the advancement in machine learning applications in the field of health care.ObjectiveIn this study, we explore three methods to increase the computational efficiency and reduce model sizes of either recurrent neural networks (RNNs) or feedforward deep neural networks (DNNs) without compromising their accuracy.MethodsWe used inpatient mortality prediction as our case analysis upon review of an intensive care unit dataset. We reduced the size of RNN and DNN by applying pruning of “unused” neurons. Additionally, we modified the RNN structure by adding a hidden layer to the RNN cell but reducing the total number of recurrent layers to accomplish a reduction of the total parameters used in the network. Finally, we implemented quantization on DNN by forcing the weights to be 8 bits instead of 32 bits.ResultsWe found that all methods increased implementation efficiency, including training speed, memory size, and inference speed, without reducing the accuracy of mortality prediction.ConclusionsOur findings suggest that neural network condensation allows for the implementation of sophisticated neural network algorithms on devices with lower computational resources.

Physiological Dysregulation, Frailty, and Impacts on Adverse Health and Functional Outcomes.

Background: Multi-system physiological dysregulation (PD) may represent a biological endo-phenotype of clinical frailty. We investigated the co-occurrence of PD with physical frailty and its contributions to the known impact of frailty on adverse health outcomes.Methods: Data of 2,725 participants from the Singapore Longitudinal Aging Studies (SLAS-2), included baseline measures of physical frailty and PD derived from Mahalanobis distance (Dm) value of 23 blood biomarkers. We analyzed their concurrent association and their impacts on 9-year mortality, MMSE cognition, GDS depression, number of medications, disability, and hospitalization at baseline and follow up (mean 4.5 years).Results: Global PD (Log10Dm, mean = 1.24, SD = 0.24) was significantly but weakly associated with pre-frailty-and-frailty. Controlling for age, sex and education, pre-frailty-and-frailty (HR = 2.12, 95% CI = 1.51–3.00) and PD (HR = 3.88, 95% CI = 2.15–6.98) predicted mortality. Together in the same model, mortality HR associated with pre-frailty-and-frailty (HR = 1.83, 95% CI = 1.22–2.73) and PD (HR = 3.06, 95% CI = 1.60–5.85) were reduced after additionally adding global PD to the prediction model. The predictive accuracy for mortality were both approximately the same (PD: AUC = 0.62, frailty: AUC = 0.64), but AUC was significantly increased to 0.68 when combined (p < 0.001). Taken into account in the same model, frailty remained significantly associated with all health and functional outcomes, and PD was significantly associated with only MMSE, disability and medications used. In secondary analyses, there were mixed associations of system-specific PDs with frailty and different adverse outcomes.Conclusions: Co-existing PD and physical frailty independently predict mortality and functional and health outcomes, with increased predictive accuracy when combined. PD appears to be a valid representation of a biological endo-phenotype of frailty, and the potential utility of such subclinical measures of frailty could be further studied.

Lactate Clearance Versus Base Deficit Correction in Monotoring of Septic Shock

Abstract Background Tissue perfusion and oxygenation monitoring is the fundamental of management to immediately identify and properly manage (impending) organ dysfunction in septic shock. This is a study of a prospective randomized trial that aims to comparing the prognostic value of lactate clearance versus base deficit correction as an indicator of tissue perfusion in septic shock, trying to prove the superiority of one over the other. Objective To assess the clinical utility of lactate clearance and arterial base deficit correction in predicting mortality in septic shock. Patients and Methods This is a Prospective randomized observational study that was conducted to assess the clinical utility of lactate clearance and arterial base deficit in predicting mortality in septic shock. This study included forty patients who were divided into two equal groups; each of twenty patients, group I (non-survivors), and group II (survivors). Results In this study, the base deficit values revealed statistically notable difference between the two groups at admission, 18 hours, 24 hours and 2 days of admission. The values revealed no notable difference at the 3rd. 4th, and 5th day of admission. At the 6th and 7th day of admission, the base deficit started to increase again in the non-survivor group with statistically notable increase in the values as compared to the survivor group. The cutoff point of base deficit to differentiate between nonsurvivors from survivors was &amp;gt; 11.43 with 62% sensitivity and 100% specificity. In this study, initial serum lactate levels at admission revealed higher accuracy in prediction of mortality as compared with initial base deficit. Conclusion Lactate clearance is a better predictor of mortality and morbidity than base deficit correction.

Risk prediction model of in-hospital mortality in heart failure with preserved ejection fraction and mid-range ejection fraction: a retrospective cohort study.

Aim: To develop and validate internally a multivariate risk model for predicting the in-hospital mortality of patients with heart failure with preserved ejection fraction (HFpEF) and heart failure with mid-range ejection fraction (HFmrEF). Methods &results: The clinical data of 8172 inpatients with HFpEF and HFmrEF was used to establish a retrospective database. These patients, among whom 307 in-hospital deaths (3.8%) occurred, were randomly assigned to derivation and verification cohort. Among the extracted data from the derivation cohort were nine variables significantly related to in-hospital mortality, which were scored 0-4, for a total score of 24, which allowed formation of a risk predictive model. The verification cohort was then used to validate the discrimination and calibration capacities of this predictive model: the area under curveequaled 0.8575 (0.8285, 0.8865) for the derivation cohort, and 0.8323 (0.7999, 0.8646) for the verification cohort. According to this risk score, we divided patients into four risk classes (low-, medium-, high- and extremely high-risk) and revealed that the risk of in-hospital mortality increased with increasing risk class with an obvious linear relationship between actual and predicted mortality (r=0.998, p<0.001). Conclusion: The model based on nine common clinical variables should provide an accurate prediction of in-hospital mortality and appears to be a reliable risk classification system for patients with HFpEF and HFmrEF.

Machine-learning-based COVID-19 mortality prediction model and identification of patients at low and high risk of dying

BackgroundThe coronavirus disease 2019 (COVID-19) pandemic caused by the SARS-Cov2 virus has become the greatest health and controversial issue for worldwide nations. It is associated with different clinical manifestations and a high mortality rate. Predicting mortality and identifying outcome predictors are crucial for COVID patients who are critically ill. Multivariate and machine learning methods may be used for developing prediction models and reduce the complexity of clinical phenotypes.MethodsMultivariate predictive analysis was applied to 108 out of 250 clinical features, comorbidities, and blood markers captured at the admission time from a hospitalized cohort of patients (N = 250) with COVID-19. Inspired modification of partial least square (SIMPLS)-based model was developed to predict hospital mortality. Prediction accuracy was randomly assigned to training and validation sets. Predictive partition analysis was performed to obtain cutting value for either continuous or categorical variables. Latent class analysis (LCA) was carried to cluster the patients with COVID-19 to identify low- and high-risk patients. Principal component analysis and LCA were used to find a subgroup of survivors that tends to die.ResultsSIMPLS-based model was able to predict hospital mortality in patients with COVID-19 with moderate predictive power (Q2 = 0.24) and high accuracy (AUC > 0.85) through separating non-survivors from survivors developed using training and validation sets. This model was obtained by the 18 clinical and comorbidities predictors and 3 blood biochemical markers. Coronary artery disease, diabetes, Altered Mental Status, age > 65, and dementia were the topmost differentiating mortality predictors. CRP, prothrombin, and lactate were the most differentiating biochemical markers in the mortality prediction model. Clustering analysis identified high- and low-risk patients among COVID-19 survivors.ConclusionsAn accurate COVID-19 mortality prediction model among hospitalized patients based on the clinical features and comorbidities may play a beneficial role in the clinical setting to better management of patients with COVID-19. The current study revealed the application of machine-learning-based approaches to predict hospital mortality in patients with COVID-19 and identification of most important predictors from clinical, comorbidities and blood biochemical variables as well as recognizing high- and low-risk COVID-19 survivors.

Deep Learning with Heterogeneous Graph Embeddings for Mortality Prediction from Electronic Health Records

Computational prediction of in-hospital mortality in the setting of an intensive care unit can help clinical practitioners to guide care and make early decisions for interventions. As clinical data are complex and varied in their structure and components, continued innovation of modelling strategies is required to identify architectures that can best model outcomes. In this work, we trained a Heterogeneous Graph Model (HGM) on electronic health record (EHR) data and used the resulting embedding vector as additional information added to a Convolutional Neural Network (CNN) model for predicting in-hospital mortality. We show that the additional information provided by including time as a vector in the embedding captured the relationships between medical concepts, lab tests, and diagnoses, which enhanced predictive performance. We found that adding HGM to a CNN model increased the mortality prediction accuracy up to 4%. This framework served as a foundation for future experiments involving different EHR data types on important healthcare prediction tasks.