Prognostic Value of Serum/Plasma Neurofilament Light Chain for COVID-19 Associated Mortality

Given the continued spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), early predictors of coronavirus disease 19 (COVID-19) mortality might improve patients outcomes. Increased levels of circulating neurofilament light chain (NfL), a biomarker of neuro-axonal injury, have been observed in patients with severe COVID-19. We investigated whether NfL provides non-redundant clinical value to previously identified predictors of COVID-19 mortality. We measured serum or plasma NfL concentrations in a blinded fashion in 3 cohorts totaling 338 COVID-19 patients. In cohort 1, we found significantly elevated NfL levels only in critically ill COVID-19 patients compared to healthy controls. Longitudinal cohort 2 data showed that NfL is elevated late in the course of the disease, following two other prognostic markers of COVID-19: decrease in absolute lymphocyte count (ALC) and increase in lactate dehydrogenase (LDH). Significant correlations between LDH and ALC abnormalities and subsequent rise of NfL implicate multi-organ failure as a likely cause of neuronal injury at the later stages of COVID-19. 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. In conclusion, although substantial increase in serum/plasma NfL reproducibly enhances COVID-19 mortality prediction, NfL has clinically meaningful prognostic value only close to death, which may be too late to alter medical management. When combined with other prognostic biomarkers, rising longitudinal NfL measurements triggered by LDH and ALC abnormalities would identify patients at risk of COVID-19 associated mortality who might still benefit from escalated care.

Deep Learning Based Long Term Mortality Prediction in the National Lung Screening Trial

In this study, the long-term mortality in the National Lung Screening Trial (NLST) was investigated using a deep learning-based method. Binary classification of the non-lung-cancer mortality (i.e. cardiovascular and respiratory mortality) was performed using neural network models centered around a 3D-ResNet. The models were trained on a participant age, gender, and smoking history matched cohort. Utilising both the 3D CT scan and clinical information, the models can achieve an AUC of 0.73 which outperforms humans at cardiovascular mortality prediction. By interpreting the trained models with 3D saliency maps, we examined the features on the CT scans that correspond to the mortality signal. The saliency maps can potentially assist the clinicians' and radiologists' to identify regions of concern on the image that may indicate the need to adopt preventative healthcare management strategies to prolong the patients' life expectancy.

COVID-19 Time of Intubation Mortality Evaluation (C-TIME): A System for Predicting Mortality of Patients with COVID-19 Pneumonia at the Time They Require Mechanical Ventilation.

Background: An accurate system to predict mortality in patients requiring intubation for COVID-19 could help to inform consent, frame family expectations and assist end-of-life decisions. Research objective: To develop and validate a mortality prediction system called C-TIME (COVID-19 Time of Intubation Mortality Evaluation) using variables available before intubation, determine its discriminant accuracy, and compare it to APACHE IVa and SOFA. Methods: A retrospective cohort was set in 18 medical-surgical ICUs, enrolling consecutive adults, positive by SARS-CoV 2 RNA by reverse transcriptase polymerase chain reaction or positive rapid antigen test, and undergoing endotracheal intubation. All were followed until hospital discharge or death. The combined outcome was hospital mortality or terminal extubation with hospice discharge. Twenty-five clinical and laboratory variables available 48 hours prior to intubation were entered into multiple logistic regression (MLR) and the resulting model was used to predict mortality of validation cohort patients. AUROC was calculated for C-TIME, APACHE IVa and SOFA. Results: The median age of the 2,440 study patients was 66 years; 61.6 percent were men, and 50.5 percent were Hispanic, Native American or African American. Age, gender, COPD, minimum mean arterial pressure, Glasgow Coma scale score, and PaO2/FiO2 ratio, maximum creatinine and bilirubin, receiving factor Xa inhibitors, days receiving non-invasive respiratory support and days receiving corticosteroids prior to intubation were significantly associated with the outcome variable. The validation cohort comprised 1,179 patients. C-TIME had the highest AUROC of 0.75 (95%CI 0.72-0.79), vs 0.67 (0.64-0.71) and 0.59 (0.55-0.62) for APACHE and SOFA, respectively (Chi2 P<0.0001). Conclusions: C-TIME is the only mortality prediction score specifically developed and validated for COVID-19 patients who require mechanical ventilation. It has acceptable discriminant accuracy and goodness-of-fit to assist decision-making just prior to intubation. The C-TIME mortality prediction calculator can be freely accessed on-line at https://phoenixmed.arizona.edu/ctime.

Unstructured clinical notes within the 24 hours since admission predict short, mid & long-term mortality in adult ICU patients

Mortality prediction for intensive care unit (ICU) patients is crucial for improving outcomes and efficient utilization of resources. Accessibility of electronic health records (EHR) has enabled data-driven predictive modeling using machine learning. However, very few studies rely solely on unstructured clinical notes from the EHR for mortality prediction. In this work, we propose a framework to predict short, mid, and long-term mortality in adult ICU patients using unstructured clinical notes from the MIMIC III database, natural language processing (NLP), and machine learning (ML) models. Depending on the statistical description of the patients’ length of stay, we define the short-term as 48-hour and 4-day period, the mid-term as 7-day and 10-day period, and the long-term as 15-day and 30-day period after admission. We found that by only using clinical notes within the 24 hours of admission, our framework can achieve a high area under the receiver operating characteristics (AU-ROC) score for short, mid and long-term mortality prediction tasks. The test AU-ROC scores are 0.87, 0.83, 0.83, 0.82, 0.82, and 0.82 for 48-hour, 4-day, 7-day, 10-day, 15-day, and 30-day period mortality prediction, respectively. We also provide a comparative study among three types of feature extraction techniques from NLP: frequency-based technique, fixed embedding-based technique, and dynamic embedding-based technique. Lastly, we provide an interpretation of the NLP-based predictive models using feature-importance scores.

Machine Learning Applied to Omics Datasets Predicts Mortality in Patients with Alcoholic Hepatitis

Alcoholic hepatitis is a major health care burden in the United States due to significant morbidity and mortality. Early identification of patients with alcoholic hepatitis at greatest risk of death is extremely important for proper treatments and interventions to be instituted. In this study, we used gradient boosting, random forest, support vector machine and logistic regression analysis of laboratory parameters, fecal bacterial microbiota, fecal mycobiota, fecal virome, serum metabolome and serum lipidome to predict mortality in patients with alcoholic hepatitis. Gradient boosting achieved the highest AUC of 0.87 for both 30-day mortality prediction using the bacteria and metabolic pathways dataset and 90-day mortality prediction using the fungi dataset, which showed better performance than the currently used model for end-stage liver disease (MELD) score.

Improving and Externally Validating Mortality Prediction Models for COVID-19 Using Publicly Available Data

We conducted a systematic survey of COVID-19 endpoint prediction literature to: (a) identify publications that include data that adhere to FAIR (findability, accessibility, interoperability, and reusability) principles and (b) develop and reuse mortality prediction models that best generalize to these datasets. The largest such cohort data we knew of was used for model development. The associated published prediction model was subjected to recursive feature elimination to find a minimal logistic regression model which had statistically and clinically indistinguishable predictive performance. This model could still not be applied to the four external validation sets that were identified, due to complete absence of needed model features in some external sets. Thus, a generalizable model (GM) was built which could be applied to all four external validation sets. An age-only model was used as a benchmark, as it is the simplest, effective, and robust predictor of mortality currently known in COVID-19 literature. While the GM surpassed the age-only model in three external cohorts, for the fourth external cohort, there was no statistically significant difference. This study underscores: (1) the paucity of FAIR data being shared by researchers despite the glut of COVID-19 prediction models and (2) the difficulty of creating any model that consistently outperforms an age-only model due to the cohort diversity of available datasets.