Mortality Prediction Model of Septic Shock Patients Based on Routinely Recorded Data

Prediction of mortality in patients with secondary pulmonary embolism based on primary admission indication: A short communication

We evaluated the prediction of mortality in patients admitted to the intensive care unit (ICU) who subsequently developed a pulmonary embolism (PE) (i.e., secondary PE) using three PE-specific scores, the Pulmonary Embolism Severity Index (PESI), simplified PESI (sPESI), and modified sPESI (ICU-sPESI) and compared them to the gold standard for the assessment of ICU all-cause mortality, the Acute Physiology and Chronic Health Evaluation-IV (APACHE-IV). All critical care admission indications were grouped into four major categories: post-operative, cardiovascular, infectious (sepsis), and other. The APACHE-IV displayed better discriminative ability to predict in-hospital mortality than the PESI and ICU-sPESI, but these two scores still performed fair for the ICU admissions related to postoperative, cardiovascular, and other admission types. Meanwhile, the sPESI displayed poor predictive performance across all four admission categories. Notably, discriminatory performance for patients with an infection-related admission was consistently low regardless of which score was used.

A nomogram for predicting mortality risk within 30 days in sepsis patients admitted in the emergency department: A retrospective analysis

OBJECTIVE

To establish and validate an individualized nomogram to predict mortality risk within 30 days in patients with sepsis from the emergency department.

METHODS

Data of 1205 sepsis patients who were admitted to the emergency department in a tertiary hospital between Jun 2013 and Sep 2021 were collected and divided into a training group and a validation group at a ratio of 7:3. The independent risk factors related to 30-day mortality were identified by univariate and multivariate analysis in the training group and used to construct the nomogram. The model was evaluated by receiver operating characteristic (ROC) curve, calibration chart and decision curve analysis. The model was validated in patients of the validation group and its performance was confirmed by comparing to other models based on SOFA score and machine learning methods.

RESULTS

The independent risk factors of 30-day mortality of sepsis patients included pro-brain natriuretic peptide, lactic acid, oxygenation index (PaO2/FiO2), mean arterial pressure, and hematocrit. The AUCs of the nomogram in the training and verification groups were 0.820 (95% CI: 0.780-0.860) and 0.849 (95% CI: 0.783-0.915), respectively, and the respective P-values of the calibration chart were 0.996 and 0.955. The DCA curves of both groups were above the two extreme curves, indicating high clinical efficacy. The AUC values were 0.847 for the model established by the random forest method and 0.835 for the model established by the stacking method. The AUCs of SOFA model in the model and validation groups were 0.761 and 0.753, respectively.

CONCLUSION

The sepsis nomogram can predict the risk of death within 30 days in sepsis patients with high accuracy, which will be helpful for clinical decision-making.

DYNAMIC PREDICTION OF RESIDUAL LIFE WITH LONGITUDINAL COVARIATES USING LONG SHORT-TERM MEMORY NETWORKS

Sepsis, a complex medical condition that involves severe infections with life-threatening organ dysfunction, is a leading cause of death worldwide. Treatment of sepsis is highly challenging. When making treatment decisions, clinicians and patients desire accurate predictions of mean residual life (MRL) that leverage all available patient information, including longitudinal biomarker data. Biomarkers are biological, clinical, and other variables reflecting disease progression that are often measured repeatedly on patients in the clinical setting. Dynamic prediction methods leverage accruing biomarker measurements to improve performance, providing updated predictions as new measurements become available. We introduce two methods for dynamic prediction of MRL using longitudinal biomarkers. in both methods, we begin by using long short-term memory networks (LSTMs) to construct encoded representations of the biomarker trajectories, referred to as "context vectors." In our first method, the LSTM-GLM, we dynamically predict MRL via a transformed MRL model that includes the context vectors as covariates. In our second method, the LSTM-NN, we dynamically predict MRL from the context vectors using a feed-forward neural network. We demonstrate the improved performance of both proposed methods relative to competing methods in simulation studies. We apply the proposed methods to dynamically predict the restricted mean residual life (RMRL) of septic patients in the intensive care unit using electronic medical record data. We demonstrate that the LSTM-GLM and the LSTM-NN are useful tools for producing individualized, real-time predictions of RMRL that can help inform the treatment decisions of septic patients.

Analysis of the correlation between the longitudinal trajectory of SOFA scores and prognosis in patients with sepsis at 72 hour after admission based on group trajectory modeling

BACKGROUND

To identify the distinct trajectories of the Sequential Organ Failure Assessment (SOFA) scores at 72 h for patients with sepsis in the Medical Information Mart for Intensive Care (MIMIC)-IV database and determine their effects on mortality and adverse clinical outcomes.

METHODS

A retrospective cohort study was carried out involving patients with sepsis from the MIMIC-IV database. Group-based trajectory modeling (GBTM) was used to identify the distinct trajectory groups for the SOFA scores in patients with sepsis in the intensive care unit (ICU). The Cox proportional hazards regression model was used to investigate the relationship between the longitudinal change trajectory of the SOFA score and mortality and adverse clinical outcomes.

RESULTS

A total of 16,743 patients with sepsis were included in the cohort. The median survival age was 66 years (interquartile range: 54-76 years). The 7-day and 28-day in-hospital mortality were 6.0% and 17.6%, respectively. Five different trajectories of SOFA scores according to the model fitting standard were determined: group 1 (32.8%), group 2 (30.0%), group 3 (17.6%), group 4 (14.0%) and group 5 (5.7%). Univariate and multivariate Cox regression analyses showed that, for different clinical outcomes, trajectory group 1 was used as the reference, while trajectory groups 2-5 were all risk factors associated with the outcome (P < 0.001). Subgroup analysis revealed an interaction between the two covariates of age and mechanical ventilation and the different trajectory groups of patients' SOFA scores (P < 0.05).

CONCLUSION

This approach may help identify various groups of patients with sepsis, who may be at different levels of risk for adverse health outcomes, and provide subgroups with clinical importance.

Understanding the complexity of sepsis mortality prediction via rule discovery and analysis: a pilot study

BACKGROUND

Sepsis, defined as life-threatening organ dysfunction caused by a dysregulated host response to infection, has become one of the major causes of death in Intensive Care Units (ICUs). The heterogeneity and complexity of this syndrome lead to the absence of golden standards for its diagnosis, treatment, and prognosis. The early prediction of in-hospital mortality for sepsis patients is not only meaningful to medical decision making, but more importantly, relates to the well-being of patients.

METHODS

In this paper, a rule discovery and analysis (rule-based) method is used to predict the in-hospital death events of 2021 ICU patients diagnosed with sepsis using the MIMIC-III database. The method mainly includes two phases: rule discovery phase and rule analysis phase. In the rule discovery phase, the RuleFit method is employed to mine multiple hidden rules which are capable to predict individual in-hospital death events. In the rule analysis phase, survival analysis and decomposition analysis are carried out to test and justify the risk prediction ability of these rules. Then by leveraging a subset of these rules, we establish a prediction model that is both more accurate at the in-hospital death prediction task and more interpretable than most comparable methods.

RESULTS

In our experiment, RuleFit generates 77 risk prediction rules, and the average area under the curve (AUC) of the prediction model based on 62 of these rules reaches 0.781 ([Formula: see text]) which is comparable to or even better than the AUC of existing methods (i.e., commonly used medical scoring system and benchmark machine learning models). External validation of the prediction power of these 62 rules on another 1468 sepsis patients not included in MIMIC-III in ICU provides further supporting evidence for the superiority of the rule-based method. In addition, we discuss and explain in detail the rules with better risk prediction ability. Glasgow Coma Scale (GCS), serum potassium, and serum bilirubin are found to be the most important risk factors for predicting patient death.

CONCLUSION

Our study demonstrates that, with the rule-based method, we could not only make accurate prediction on in-hospital death events of sepsis patients, but also reveal the complex relationship between sepsis-related risk factors through the rules themselves, so as to improve our understanding of the complexity of sepsis as well as its population.

Myocardial Strain and Cardiac Output are Preferable Measurements for Cardiac Dysfunction and Can Predict Mortality in Septic Mice

Background

Sepsis is the overwhelming host response to infection leading to shock and multiple organ dysfunction. Cardiovascular complications greatly increase sepsis‐associated mortality. Although murine models are routinely used for preclinical studies, the benefit of using genetically engineered mice in sepsis is countered by discrepancies between human and mouse sepsis pathophysiology. Therefore, recent guidelines have called for standardization of preclinical methods to document organ dysfunction. We investigated the course of cardiac dysfunction and myocardial load in different mouse models of sepsis to identify the optimal measurements for early systolic and diastolic dysfunction.

Methods and Results

We performed speckle‐tracking echocardiography and assessed blood pressure, plasma inflammatory cytokines, lactate, B‐type natriuretic peptide, and survival in mouse models of endotoxemia or polymicrobial infection (cecal ligation and puncture, [CLP]) of moderate and high severity. We observed that myocardial strain and cardiac output were consistently impaired early in both sepsis models. Suppression of cardiac output was associated with systolic dysfunction in endotoxemia or combined systolic dysfunction and reduced preload in the CLP model. We found that cardiac output at 2 hours post‐CLP is a negative prognostic indicator with high sensitivity and specificity that predicts mortality at 48 hours. Using a known antibiotic (ertapenem) treatment, we confirmed that this approach can document recovery.

Conclusions

We propose a non‐invasive approach for assessment of cardiac function in sepsis and myocardial strain and strain rate as preferable measures for monitoring cardiovascular function in sepsis mouse models. We further show that the magnitude of cardiac output suppression 2 hours post‐CLP can be used to predict mortality.

A Predictive Model Based on Machine Learning for the Early Detection of Late-Onset Neonatal Sepsis: Development and Observational Study

BACKGROUND

Neonatal sepsis is associated with most cases of mortalities and morbidities in the neonatal intensive care unit (NICU). Many studies have developed prediction models for the early diagnosis of bloodstream infections in newborns, but there are limitations to data collection and management because these models are based on high-resolution waveform data.

OBJECTIVE

The aim of this study was to examine the feasibility of a prediction model by using noninvasive vital sign data and machine learning technology.

METHODS

We used electronic medical record data in intensive care units published in the Medical Information Mart for Intensive Care III clinical database. The late-onset neonatal sepsis (LONS) prediction algorithm using our proposed forward feature selection technique was based on NICU inpatient data and was designed to detect clinical sepsis 48 hours before occurrence. The performance of this prediction model was evaluated using various feature selection algorithms and machine learning models.

RESULTS

The performance of the LONS prediction model was found to be comparable to that of the prediction models that use invasive data such as high-resolution vital sign data, blood gas estimations, blood cell counts, and pH levels. The area under the receiver operating characteristic curve of the 48-hour prediction model was 0.861 and that of the onset detection model was 0.868. The main features that could be vital candidate markers for clinical neonatal sepsis were blood pressure, oxygen saturation, and body temperature. Feature generation using kurtosis and skewness of the features showed the highest performance.

CONCLUSIONS

The findings of our study confirmed that the LONS prediction model based on machine learning can be developed using vital sign data that are regularly measured in clinical settings. Future studies should conduct external validation by using different types of data sets and actual clinical verification of the developed model.

Low vasopressin and progression of neonatal sepsis to septic shock: a prospective cohort study

The study objective was to analyze the association between low plasma vasopressin and progression of sepsis to septic shock in neonates < 34 weeks gestation. Septic neonates of < 34 weeks gestation were consecutively enrolled; moribund neonates and those with major malformations were excluded. Subjects were monitored for progression of sepsis to septic shock over the first 7 days from enrolment. Plasma vasopressin levels and inducible nitric oxide synthase levels were measured at the onset of sepsis (T0), severe sepsis (T1), and septic shock (T2). Primary outcome was plasma vasopressin levels at the point of sepsis in those who progressed to septic shock in comparison with matched nested controls in the non-progression group. Forty-nine (47%) enrolled subjects developed severe sepsis or septic shock. Plasma vasopressin levels (pg/ml) at the onset of sepsis were significantly low in those who progressed to septic shock (median (IQR), 31 (2.5-80) versus 100 (12-156); p = 0.02). After adjusting for confounders, vasopressin levels were independently associated with progression to septic shock (adjusted OR (95% CI), 0.97 (0.96, 0.99); p = 0.01).Conclusion: Preterm septic neonates who progressed to septic shock had suppressed vasopressin levels before the onset of shock. Low vasopressin levels were independently associated with progression to septic shock.What is known:• In animal sepsis models and adult septic patients, exuberant production of nitric oxide metabolites and low vasopressin levels have been reportedly associated with progression to septic shock.• Vasopressin levels have been variably reported as low as well as elevated in children with septic shock.What is New:• Preterm neonates who progressed from sepsis to septic shock had significantly lower levels of vasopressin before the onset of shock in comparison with those who did not progress.• Low vasopressin levels independently predicted the progression from sepsis to septic shock in this population.

Building Computational Models to Predict One-Year Mortality in ICU Patients with Acute Myocardial Infarction and Post Myocardial Infarction Syndrome

Heart disease remains the leading cause of death in the United States. Compared with risk assessment guidelines that require manual calculation of scores, machine learning-based prediction for disease outcomes such as mortality can be utilized to save time and improve prediction accuracy. This study built and evaluated various machine learning models to predict one-year mortality in patients diagnosed with acute myocardial infarction or post myocardial infarction syndrome in the MIMIC-III database. The results of the best performing shallow prediction models were compared to a deep feedforward neural network (Deep FNN) with back propagation. We included a cohort of 5436 admissions. Six datasets were developed and compared. The models applying Logistic Model Trees (LMT) and Simple Logistic algorithms to the combined dataset resulted in the highest prediction accuracy at 85.12% and the highest AUC at .901. In addition, other factors were observed to have an impact on outcomes as well.

A machine learning-based model for 1-year mortality prediction in patients admitted to an Intensive Care Unit with a diagnosis of sepsis

INTRODUCTION

Sepsis is associated to a high mortality rate, and its severity must be evaluated quickly. The severity of illness scores used are intended to be applicable to all patient populations, and generally evaluate in-hospital mortality. However, patients with sepsis continue to be at risk of death after hospital discharge.

OBJECTIVE

To develop a model for predicting 1-year mortality in critical patients diagnosed with sepsis.

PATIENTS

The data corresponding to 5650 admissions of patients with sepsis from the Medical Information Mart for Intensive Care (MIMIC-III) database were evaluated, randomly divided as follows: 70% for training and 30% for validation.

DESIGN

A retrospective register-based cohort study was carried out. The clinical information of the first 24h after admission was used to develop a 1-year mortality prediction model based on Stochastic Gradient Boosting (SGB) methodology. Variable selection was addressed using Least Absolute Shrinkage and Selection Operator (LASSO) and SGB variable importance methodologies. The predictive power was evaluated using the area under the ROC curve (AUROC).

RESULTS

An AUROC of 0.8039 (95% confidence interval (CI): [0.8033 0.8045]) was obtained in the validation subset. The model exceeded the predictive performances obtained with traditional severity of disease scores in the same subset.

CONCLUSION

The use of assembly algorithms, such as SGB, for the generation of a customized model for sepsis yields more accurate 1-year mortality prediction than the traditional scoring systems such as SAPS II, SOFA or OASIS.