Machine learning for early prediction of sepsis-associated acute brain injury

Machine learning prediction models for multidrug-resistant organism infections in ICU ventilator-associated pneumonia patients: Analysis using the MIMIC-IV database

OBJECTIVE

This study aims to construct and compare four machine learning models using the MIMIC-IV database to identify high-risk factors for multidrug-resistant organism (MDRO) infection in Ventilator-associated pneumonia (VAP) patients.

METHODS

The study included 972 VAP patients from the MIMIC-IV database. Data encompassing demographic information, vital signs, laboratory results, and other relevant variables were collected. The class imbalance issue was addressed using the Synthetic Minority Over-sampling Technique (SMOTE). The dataset was randomly split into training and testing sets (8:2). LASSO regression and feature importance scores were used for feature selection. Clinical prediction models were built using logistic regression, XGBoost, random forest and gradient boosting machine. The performance of the models was evaluated through receiver operating characteristic(ROC) curve analysis.Model calibration was assessed using calibration curves and Brier scores. The effectiveness was evaluated through Decision Curve Analysis (DCA). SHAP was utilized for model interpretation.

RESULTS

Among 972 patients, 824 were non-MDROs-VAP and 128 were MDROs-VAP. Comparative analysis revealed statistically significant differences in various clinical parameters. XGBoost exhibited the best predictive performance, incorporating 20 features with an AUC of 0.831 (95 % CI: 0.785-0.877) on the test set. Calibration curves demonstrated robust consistency, corroborated by Decision Curve Analysis (DCA) affirming the clinical utility. SHAP analysis identified the most important features: red cell distribution width, duration of mechanical ventilation, anion gap, basophil percentage, and neutrophil percentage.

CONCLUSION

This study established and compared four machine learning models for MDROs infections in VAP patients. XGBoost was identified as the optimal predictor, and SHAP values provided insights into 20 independent risk factors, confirming its excellent predictive value.

IMPLICATIONS FOR CLINICAL PRACTICE

VAP is a common infection in ICU patients with a heightened risk of MDRO and increased mortality. The recognition of high bias in existing models calls for future research to employ rigorous methodologies and robust data sources, aiming to develop and validate more accurate and clinically applicable predictive models for MDROs infections in VAP patients.

Early prediction of sepsis associated encephalopathy in elderly ICU patients using machine learning models: a retrospective study based on the MIMIC-IV database

Background

Sepsis associated encephalopathy (SAE) is prevalent among elderly patients in the ICU and significantly affects patient prognosis. Due to the symptom similarity with other neurological disorders and the absence of specific biomarkers, early clinical diagnosis remains challenging. This study aimed to develop a predictive model for SAE in elderly ICU patients.

Methods

The data of elderly sepsis patients were extracted from the MIMIC IV database (version 3.1) and divided into training and test sets in a 7:3 ratio. Feature variables were selected using the LASSO-Boruta combined algorithm, and five machine learning (ML) models, including Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost),Light Gradient Boosting Machine(LGBM), Multilayer Perceptron (MLP), and Support Vector Machines (SVM), were subsequently developed using these variables. A comprehensive set of performance metrics was used to assess the predictive accuracy, calibration, and clinical applicability of these models. For the machine learning model with the best performance, we employed the SHapley Additive Explanations(SHAP) method to visualize the model.

Results

Based on strict inclusion and exclusion criteria, a total of 3,156 elderly sepsis patients were enrolled in the study, with an SAE incidence rate of 48.7%. The mortality rate of elderly sepsis patients who developed SAE was significantly higher than that of patients in the non-SAE group (28.78% vs. 12.59%, P < 0.001). A total of 18 feature variables were selected for the construction of the ML model using the LASSO-Boruta combined algorithm. Compared to the other four models and traditional scoring systems, the XGBoost model demonstrated the best overall predictive performance, with Area Under the Curve(AUC)=0.898, accuracy=0.830, recall=0.819, F1-Score=0.820, specificity=0.840, and Precision=0.821. Furthermore, the results from the Decision Curve Analysis (DCA) and calibration curves demonstrated that the XGBoost model has significant clinical value and stable predictive performance. The ten-fold cross-validation method further confirmed the robustness and generalizability of the model. In addition, we simplified the model based on the SHAP feature importance ranking, and the results indicated that the simplified XGBoost model retains excellent predictive ability (AUC=0.858).

Conclusions

The XGBoost model effectively predicts SAE in elderly ICU patients and may serve as a reliable tool for clinicians to identify high-risk patients.

Development and external validation of a machine learning model for brain injury in pediatric patients on extracorporeal membrane oxygenation

BACKGROUND

Patients supported by extracorporeal membrane oxygenation (ECMO) are at a high risk of brain injury, contributing to significant morbidity and mortality. This study aimed to employ machine learning (ML) techniques to predict brain injury in pediatric patients ECMO and identify key variables for future research.

METHODS

Data from pediatric patients undergoing ECMO were collected from the Chinese Society of Extracorporeal Life Support (CSECLS) registry database and local hospitals. Ten ML methods, including random forest, support vector machine, decision tree classifier, gradient boosting machine, extreme gradient boosting, light gradient boosting machine, Naive Bayes, neural networks, a generalized linear model, and AdaBoost, were employed to develop and validate the optimal predictive model based on accuracy and area under the curve (AUC). Patients were divided into retrospective cohort for model development and internal validation, and one cohort for external validation.

RESULTS

A total of 1,633 patients supported by ECMO were included in the model development, of whom 181 experienced brain injury. In the external validation cohort, 30 of the 154 patients experienced brain injury. Fifteen features were selected for the model construction. Among the ML models tested, the random forest model achieved the best performance, with an AUC of 0.912 for internal validation and 0.807 for external validation.

CONCLUSION

The Random Forest model based on machine learning demonstrates high accuracy and robustness in predicting brain injury in pediatric patients supported by ECMO, with strong generalization capabilities and promising clinical applicability.

The application of metagenomics, radiomics and machine learning for diagnosis of sepsis

Introduction

Sepsis poses a serious threat to individual life and health. Early and accessible diagnosis and targeted treatment are crucial. This study aims to explore the relationship between microbes, metabolic pathways, and blood test indicators in sepsis patients and develop a machine learning model for clinical diagnosis.

Methods

Blood samples from sepsis patients were sequenced. α-diversity and β-diversity analyses were performed to compare the microbial diversity between the sepsis group and the normal group. Correlation analysis was conducted on microbes, metabolic pathways, and blood test indicators. In addition, a model was developed based on medical records and radiomic features using machine learning algorithms.

Results

The results of α-diversity and β-diversity analyses showed that the microbial diversity of sepsis group was significantly higher than that of normal group (p < 0.05). The top 10 microbial abundances in the sepsis and normal groups were Vitis vinifera, Mycobacterium canettii, Solanum pennellii, Ralstonia insidiosa, Ananas comosus, Moraxella osloensis, Escherichia coli, Staphylococcus hominis, Camelina sativa, and Cutibacterium acnes. The enriched metabolic pathways mainly included Protein families: genetic information processing, Translation, Protein families: signaling and cellular processes, and Unclassified: genetic information processing. The correlation analysis revealed a significant positive correlation (p < 0.05) between IL-6 and Membrane transport. Metabolism of other amino acids showed a significant positive correlation (p < 0.05) with Cutibacterium acnes, Ralstonia insidiosa, Moraxella osloensis, and Staphylococcus hominis. Ananas comosus showed a significant positive correlation (p < 0.05) with Poorly characterized and Unclassified: metabolism. Blood test-related indicators showed a significant negative correlation (p < 0.05) with microorganisms. Logistic regression (LR) was used as the optimal model in six machine learning models based on medical records and radiomic features. The nomogram, calibration curves, and AUC values demonstrated that LR performed best for prediction.

Discussion

This study provides insights into the relationship between microbes, metabolic pathways, and blood test indicators in sepsis. The developed machine learning model shows potential for aiding in clinical diagnosis. However, further research is needed to validate and improve the model.

Optimal target mean arterial pressure for patients with sepsis-associated encephalopathy: a retrospective cohort study

BACKGROUND

Sepsis-associated encephalopathy (SAE) patients often experience changes in intracranial pressure and impaired cerebral autoregulation. Mean arterial pressure (MAP) plays a crucial role in cerebral perfusion pressure, but its relationship with mortality in SAE patients remains unclear. This study aims to investigate the relationship between MAP and the risk of 28-day and in-hospital mortality in SAE patients, providing clinicians with the optimal MAP target.

METHODS

We retrospectively collected clinical data of patients diagnosed with SAE on the first day of ICU admission from the MIMIC-IV (v2.2) database. Patients were divided into four groups based on MAP quartiles. Kruskal-Wallis H test and Chi-square test were used to compare clinical characteristics among the groups. Restricted cubic spline and segmented Cox regression models, both unadjusted and adjusted for multiple variables, were employed to elucidate the relationship between MAP and the risk of 28-day and in-hospital mortality in SAE patients and to identify the optimal MAP. Subgroup analyses were conducted to assess the stability of the results.

RESULTS

A total of 3,816 SAE patients were included. The Q1 group had higher rates of acute kidney injury and vasoactive drug use on the first day of ICU admission compared to other groups (P < 0.01). The Q1 and Q4 groups had longer ICU and hospital stays (P < 0.01). The 28-day and in-hospital mortality rates were highest in the Q1 group and lowest in the Q3 group. Multivariable adjustment restricted cubic spline curves indicated a nonlinear relationship between MAP and mortality risk (P for nonlinearity < 0.05). The MAP ranges associated with HRs below 1 for 28-day and in-hospital mortality were 74.6-90.2 mmHg and 74.6-89.3 mmHg, respectively.The inflection point for mortality risk, determined by the minimum hazard ratio (HR), was identified at a MAP of 81.5 mmHg. The multivariable adjusted segmented Cox regression models showed that for MAP < 81.5 mmHg, an increase in MAP was associated with a decreased risk of 28-day and in-hospital mortality (P < 0.05). In Model 4, each 5 mmHg increase in MAP was associated with a 15% decrease in 28-day mortality risk (HR: 0.85, 95% CI: 0.79-0.91, p < 0.05) and a 14% decrease in in-hospital mortality risk (HR: 0.86, 95% CI: 0.80-0.93, p < 0.05). However, for MAP ≥ 81.5 mmHg, there was no significant association between MAP and mortality risk (P > 0.05). Subgroup analyses based on age, congestive heart failure, use of vasoactive drugs, and acute kidney injury showed consistent results across different subgroups.Subsequent analysis of SAE patients with septic shock also showed results similar to those of the original cohort.However, for comatose SAE patients (GCS ≤ 8), there was a negative correlation between MAP and the risk of 28-day and in-hospital mortality when MAP was < 81.5 mmHg, but a positive correlation when MAP was ≥ 81.5 mmHg in adjusted models 2 and 4.

CONCLUSION

There is a nonlinear relationship between MAP and the risk of 28-day and in-hospital mortality in SAE patients. The optimal MAP target for SAE patients in clinical practice appears to be 81.5 mmHg.

Enhancing predictions with a stacking ensemble model for ICU mortality risk in patients with sepsis-associated encephalopathy

OBJECTIVE

We identified predictive factors and developed a novel machine learning (ML) model for predicting mortality risk in patients with sepsis-associated encephalopathy (SAE).

METHODS

In this retrospective cohort study, data from the Medical Information Mart for Intensive Care IV (MIMIC-IV) and eICU Collaborative Research Database were used for model development and external validation. The primary outcome was the in-hospital mortality rate among patients with SAE; the observed in-hospital mortality rate was 14.74% (MIMIC IV: 1112, eICU: 594). Using the least absolute shrinkage and selection operator (LASSO), we built nine ML models and a stacking ensemble model and determined the optimal model based on the area under the receiver operating characteristic curve (AUC). We used the Shapley additive explanations (SHAP) algorithm to determine the optimal model.

RESULTS

The study included 9943 patients. LASSO identified 15 variables. The stacking ensemble model achieved the highest AUC on the test set (0.807) and 0.671 on external validation. SHAP analysis highlighted Glasgow Coma Scale (GCS) and age as key variables. The model (https://sic1.shinyapps.io/SSAAEE/) can predict in-hospital mortality risk for patients with SAE.

CONCLUSIONS

We developed a stacked ensemble model with enhanced generalization capabilities using novel data to predict mortality risk in patients with SAE.

The role of the peripheral system dysfunction in the pathogenesis of sepsis-associated encephalopathy

Sepsis is a condition that greatly impacts the brain, leading to neurological dysfunction and heightened mortality rates, making it one of the primary organs affected. Injury to the central nervous system can be attributed to dysfunction of various organs throughout the entire body and imbalances within the peripheral immune system. Furthermore, central nervous system injury can create a vicious circle with infection-induced peripheral immune disorders. We collate the pathogenesis of septic encephalopathy, which involves microglial activation, programmed cell death, mitochondrial dysfunction, endoplasmic reticulum stress, neurotransmitter imbalance, and blood-brain barrier disruption. We also spotlight the effects of intestinal flora and its metabolites, enterocyte-derived exosomes, cholinergic anti-inflammatory pathway, peripheral T cells and their cytokines on septic encephalopathy.

Interpretable Machine Learning Approach for Predicting 30-Day Mortality of Critical Ill Patients with Pulmonary Embolism and Heart Failure: A Retrospective Study

BACKGROUND

Pulmonary embolism (PE) patients combined with heart failure (HF) have been reported to have a high short-term mortality. However, few studies have developed predictive tools of 30-day mortality for these patients in intensive care unit (ICU). This study aimed to construct and validate a machine learning (ML) model to predict 30-day mortality for PE patients combined with HF in ICU.

METHODS

We enrolled patients with PE combined with HF in the Medical Information Mart for Intensive Care Database (MIMIC) and developed six ML models after feature selection. Further, eICU Collaborative Research Database (eICU-CRD) was utilized for external vali- dation. The area under curves (AUC), calibration curves, decision curve analysis (DCA), net reclassification improvement (NRI), and integrated discrimination improvement (IDI) were performed to evaluate the prediction performance. Shapley additive explanation (SHAP) was performed to enhance the interpretability of our models.

RESULTS

A total of 472 PE patients combined with HF were included. We developed six ML models by the 13 selected features. After internal validation, the Support Vector Ma- chine (SVM) model performed best with an AUC of 0.835, a superior calibration degree, and a wider risk threshold (from 0% to 90%) for obtaining clinical benefit, which also outperformed traditional mortality risk evaluation systems,as evaluated by NRI and IDI. The SVM model was still reliable after external validation. SHAP was performed to explain the model. Moreover, an online application was developed for further clinical use.

CONCLUSION

This study developed a potential tool for identify short-term mortality risk to guide clinical decision making for PE patients combined with HF in the ICU. The SHAP method also helped clinicians to better understand the model.

Association between Red Blood Cell Distribution Width and Short-Term Mortality in Patients with Paralytic Intestinal Obstruction: Retrospective Data Analysis Based on the MIMIC-III Database

Objective

Elevated red cell distribution (RDW) has been reported to be associated with mortality in patients with acute pancreatitis and cholecystitis admitted to the intensive care unit (ICU). However, evidence for the relationship between RDW and paralytic intestinal obstruction is lacking. Therefore, the article aims to investigate the relationship between RDW and 28-day mortality of the patients with paralytic intestinal obstruction. Patients and Methods. This is a single-center retrospective study. Based on a particular screening criterion, 773 patients with paralytic intestinal obstruction were selected from the Medical Information Mart for Intensive Care III (MIMIC-III) database. Indicators of the first 24 h into the ICU were used to analyze the relationship between RDW and 28-day death from paralytic intestinal obstruction by Kaplan-Meier (K-M) analysis, logistic regression analysis, and stratification analysis.

Results

The curve fitting exhibited a nonlinear relationship. The K-M curve showed that groups with higher RDW values had lower survival rates. The logistic regression analysis revealed that RDW increased with 28-day mortality in patients with paralytic intestinal obstruction in the fully adjusted model. In the fully adjusted model, OR value and 95% CI from the second to the third quantiles compared to the first quartile (reference group) were 1.89 (1.04, 3.44) and 3.29 (1.82, 5.93), respectively. The results of stratified analysis of each layer had the same trend as those of regression analysis, and the interaction results were not significant.

Conclusion

Elevated RDW was associated with increased 28-day mortality from paralytic intestinal obstruction in the ICU. This study can help to further explore the relationship between RDW and death in patients with paralytic intestinal obstruction.

Development of a machine learning-based prediction model for sepsis-associated delirium in the intensive care unit

Septic patients in the intensive care unit (ICU) often develop sepsis-associated delirium (SAD), which is strongly associated with poor prognosis. The aim of this study is to develop a machine learning-based model for the early prediction of SAD. Patient data were extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database and the eICU Collaborative Research Database (eICU-CRD). The MIMIC-IV data were divided into a training set and an internal validation set, while the eICU-CRD data served as an external validation set. Feature variables were selected using least absolute shrinkage and selection operator regression, and prediction models were built using logistic regression, support vector machines, decision trees, random forests, extreme gradient boosting (XGBoost), k-nearest neighbors and naive Bayes methods. The performance of the models was evaluated in the validation set. The model was also applied to a group of patients who were not assessed or could not be assessed for delirium. The MIMIC-IV and eICU-CRD databases included 14,620 and 1723 patients, respectively, with a median time to diagnosis of SAD of 24 and 30 h. Compared with Non-SAD patients, SAD patients had higher 28-days ICU mortality rates and longer ICU stays. Among the models compared, the XGBoost model had the best performance and was selected as the final model (internal validation area under the receiver operating characteristic curves (AUROC) = 0.793, external validation AUROC = 0.701). The XGBoost model outperformed other models in predicting SAD. The establishment of this predictive model allows for earlier prediction of SAD compared to traditional delirium assessments and is applicable to patients who are difficult to assess with traditional methods.