Dataset Information

Development and Validation of Machine Learning Models for Real-Time Mortality Prediction in Critically Ill Patients With Sepsis-Associated Acute Kidney Injury.

ABSTRACT:

Background

Sepsis-associated acute kidney injury (SA-AKI) is common in critically ill patients, which is associated with significantly increased mortality. Existing mortality prediction tools showed insufficient predictive power or failed to reflect patients' dynamic clinical evolution. Therefore, the study aimed to develop and validate machine learning-based models for real-time mortality prediction in critically ill patients with SA-AKI.

Methods

The multi-center retrospective study included patients from two distinct databases. A total of 12,132 SA-AKI patients from the Medical Information Mart for Intensive Care IV (MIMIC-IV) were randomly allocated to the training, validation, and internal test sets. An additional 3,741 patients from the eICU Collaborative Research Database (eICU-CRD) served as an external test set. For every 12 h during the ICU stays, the state-of-the-art eXtreme Gradient Boosting (XGBoost) algorithm was used to predict the risk of in-hospital death in the following 48, 72, and 120 h and in the first 28 days after ICU admission. Area under the receiver operating characteristic curves (AUCs) were calculated to evaluate the models' performance.

Results

The XGBoost models, based on routine clinical variables updated every 12 h, showed better performance in mortality prediction than the SOFA score and SAPS-II. The AUCs of the XGBoost models for mortality over different time periods ranged from 0.848 to 0.804 in the internal test set and from 0.818 to 0.748 in the external test set. The shapley additive explanation method provided interpretability for the XGBoost models, which improved the understanding of the association between the predictor variables and future mortality.

Conclusions

The interpretable machine learning XGBoost models showed promising performance in real-time mortality prediction in critically ill patients with SA-AKI, which are useful tools for early identification of high-risk patients and timely clinical interventions.

SUBMITTER: Luo XQ

PROVIDER: S-EPMC9240603 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Background: Sepsis-induced coagulopathy (SIC) denotes an increased mortality rate and poorer prognosis in septic patients. Objectives: Our study aimed to develop and validate machine-learning models to dynamically predict the risk of SIC in critically ill patients with sepsis. Methods: Machine-learning models were developed and validated based on two public databases named Medical Information Mart for Intensive Care (MIMIC)-IV and the eICU Collaborative Research Database (eICU-CRD). Dynamic prediction of SIC involved an evaluation of the risk of SIC each day after the diagnosis of sepsis using 15 predictive models. The best model was selected based on its accuracy and area under the receiver operating characteristic curve (AUC), followed by fine-grained hyperparameter adjustment using the Bayesian Optimization Algorithm. A compact model was developed, based on 15 features selected according to their importance and clinical availability. These two models were compared with Logistic Regression and SIC scores in terms of SIC prediction. Results: Of 11,362 patients in MIMIC-IV included in the final cohort, a total of 6,744 (59%) patients developed SIC during sepsis. The model named Categorical Boosting (CatBoost) had the greatest AUC in our study (0.869; 95% CI: 0.850-0.886). Coagulation profile and renal function indicators were the most important features for predicting SIC. A compact model was developed with an AUC of 0.854 (95% CI: 0.832-0.872), while the AUCs of Logistic Regression and SIC scores were 0.746 (95% CI: 0.735-0.755) and 0.709 (95% CI: 0.687-0.733), respectively. A cohort of 35,252 septic patients in eICU-CRD was analyzed. The AUCs of the full and the compact models in the external validation were 0.842 (95% CI: 0.837-0.846) and 0.803 (95% CI: 0.798-0.809), respectively, which were still larger than those of Logistic Regression (0.660; 95% CI: 0.653-0.667) and SIC scores (0.752; 95% CI: 0.747-0.757). Prediction results were illustrated by SHapley Additive exPlanations (SHAP) values, which made our models clinically interpretable. Conclusions: We developed two models which were able to dynamically predict the risk of SIC in septic patients better than conventional Logistic Regression and SIC scores.

Project description:Sepsis is defined by life-threatening organ dysfunction during infection and is the leading cause of death in hospitals. During sepsis, there is a high risk that new onset of atrial fibrillation (AF) can occur, which is associated with significant morbidity and mortality. Consequently, early prediction of AF during sepsis would allow testing of interventions in the intensive care unit (ICU) to prevent AF and its severe complications. In this paper, we present a novel automated AF prediction algorithm for critically ill sepsis patients using electrocardiogram (ECG) signals. From the heart rate signal collected from 5-min ECG, feature extraction is performed using the traditional time, frequency, and nonlinear domain methods. Moreover, variable frequency complex demodulation and tunable Q-factor wavelet-transform-based time-frequency methods are applied to extract novel features from the heart rate signal. Using a selected feature subset, several machine learning classifiers, including support vector machine (SVM) and random forest (RF), were trained using only the 2001 Computers in Cardiology data set. For testing the proposed method, 50 critically ill ICU subjects from the Medical Information Mart for Intensive Care (MIMIC) III database were used in this study. Using distinct and independent testing data from MIMIC III, the SVM achieved 80% sensitivity, 100% specificity, 90% accuracy, 100% positive predictive value, and 83.33% negative predictive value for predicting AF immediately prior to the onset of AF, while the RF achieved 88% AF prediction accuracy. When we analyzed how much in advance we can predict AF events in critically ill sepsis patients, the algorithm achieved 80% accuracy for predicting AF events 10 min early. Our algorithm outperformed a state-of-the-art method for predicting AF in ICU patients, further demonstrating the efficacy of our proposed method. The annotations of patients' AF transition information will be made publicly available for other investigators. Our algorithm to predict AF onset is applicable for any ECG modality including patch electrodes and wearables, including Holter, loop recorder, and implantable devices.

Project description:Background: Systemic inflammation is a whole body reaction that can have an infection-positive (i.e. sepsis) or infection-negative origin. It is important to distinguish between septic and non-septic presentations early and reliably, because this has significant therapeutic implications for critically ill patients. We hypothesized that a molecular classifier based on a small number of RNAs expressed in peripheral blood could be discovered that would: 1) determine which patients with systemic inflammation had sepsis; 2) be robust across independent patient cohorts; 3) be insensitive to disease severity; and 4) provide diagnostic utility. The overall goal of this study was to identify and validate such a molecular classifier. Methods and Findings: We conducted an observational, non-interventional study of adult patients recruited from tertiary intensive care units (ICU). Biomarker discovery was conducted with an Australian cohort (n = 105) consisting of sepsis patients and post -surgical patients with infection-negative systemic inflammation. Using this cohort, a four-gene classifier consisting of a combination of CEACAM4, LAMP1, PLA2G7 and PLAC8 RNA biomarkers was identified. This classifier, designated SeptiCyteÂ® Lab, was externally validated using RT-qPCR and receiver operating characteristic (ROC) curve analysis in five cohorts (n = 345) from the Netherlands. Cohort 1 (n=59) consisted of unambiguous septic cases and infection-negative systemic inflammation controls; SeptiCyteÂ® Lab gave an area under curve (AUC) of 0.96 (95% CI: 0.91-1.00). ROC analysis of a more heterogeneous group of patients (Cohorts 2-5; 249 patients after excluding 37 patients with infection likelihood possible) gave an AUC of 0.89 (95% CI: 0.85-0.93). Disease severity, as measured by Sequential Organ Failure Assessment (SOFA) score or the Acute Physiology and Chronic Health Evaluation (APACHE) IV score, was not a significant confounding variable. The diagnostic utility o f SeptiCyteÂ® Lab was evaluated by comparison to various clinical and laboratory parameters that would be available to a clinician within 24 hours of ICU admission. SeptiCyteÂ® Lab was significantly better at differentiating sepsis from infection-negative systemic inflammation than all tested parameters, both singly and in various logistic combinations. SeptiCyteÂ® Lab more than halved the diagnostic error rate compared to PCT in all tested cohorts or cohort combinations. Conclusions: SeptiCyteÂ® Lab is a rapid molecular assay that may be clinically useful in the management of ICU patients with systemic inflammation. SIRS and Sepsis ICU patients, admission samples Retrospective, mutli-site sutdy using retrospective physician adjudication as a comparator

Project description:Objective: It is unclear whether the host response of gram-positive sepsis differs from gram-negative sepsis at a transcriptome level. Using microarray technology, we compared the gene-expression profiles of gram-positive sepsis and gram-negative sepsis in critically ill patients. Design: A prospective cross-sectional study. Setting: A 20-bed general intensive care unit of a tertiary referral hospital. Patients: Seventy-two patients admitted to the intensive care unit. Interventions: Intravenous blood was collected for leukocyte separation and RNA extraction. Microarray experiements were then performed examing the expression level of 19,232 genes in each sample. Measurements and Main Results: There was no difference in the expression profile between gram-positive and gram-negative sepsis. The finding remained unchanged even when genes with lower expression level were included or after statistical stringency was lowered. There were, however, ninety-four genes differentially expressed between sepsis and control patients. These genes included those involved in immune regulation, inflammation and mitochondrial function. Hierarchical cluster analysis confirmed that the difference in gene expression profile existed between sepsis and control patients, but not between gram-positive and gram-negative patients. Conclusion: Gram-positive and gram-negative sepsis share a common host response at a transcriptome level. These findings support the hypothesis that the septic response is non-specific and is designed to provide a more general response that can be elicited by a wide range of different micro-organisms. The study included seventy-two critically ill patients admitted to the intensive care unit (ICU) of Nepean Hospital, Sydney, Australia. Of these, fifty-five patients were diagnosed to have sepsis, as confirmed by microbiological culture. The remaining seventeen patients did not have sepsis and were therefore used as controls. The study was approved by the hospital ethics committee and informed consent was obtained from all patients or their relatives. Patient Samples. Whole blood was taken from each patient on admission to ICU. Neutrophils were separated from whole blood using density-gradient separation with Ficoll-PaqueP P(Amersham). Subsequent neutrophil RNA extraction was performed using guanidinium thiocyanate (Ambion). Microarray Experiment. The neutrophil RNA was converted to cDNA, fluorescently labeled and hybridized to its complimentary sequences on the microarray (Invitrogen). The fluorescent signals on each micrroarray were captured using the GenePix 4000B laser scanner (Axon Instruments). Expression level of each gene was represented by the intensity of its fluorescent signal. Data Extraction. All signal intensity values were processed using background-subtraction method. Prior to analysis, all values were log-transformed and normalized by fitting a print-tip group Lowess curve. Normalization minimizes bias due to dye chemistry, signal intensity or location of a gene on the array. It ensures the detection of genes that are truly differentially expressed, instead of those caused by experimental artifacts or variation in the hybridization process. After normalization, genes that had more than 50% of data missing were removed. We then selected genes that had at least 80% of the data showing two-fold changes from the geneâs median values. After filtering, 1617 genes were available for further analysis.

Project description:BackgroundSepsis and acute kidney injury (AKI) are common severe diseases in the intensive care unit (ICU). This study aimed to estimate the attributable mortality of AKI among critically ill patients with sepsis and to assess whether AKI was an independent risk factor for 30-day mortality.MethodsThe information we used was derived from a multicenter prospective cohort study conducted in 18 Chinese ICUs, focusing on septic patients post ICU admission. The patients were categorized into two groups: those who developed AKI (AKI group) within seven days following a sepsis diagnosis and those who did not develop AKI (non-AKI group). Using propensity score matching (PSM), patients were matched 1:1 as AKI and non-AKI groups. We then calculated the mortality rate attributable to AKI in septic patients. Furthermore, a survival analysis was conducted comparing the matched AKI and non-AKI septic patients. The primary outcome of interest was the 30-day mortality rate following the diagnosis of sepsis.ResultsOut of the 2175 eligible septic patients, 61.7% developed AKI. After the application of PSM, a total of 784 septic patients who developed AKI were matched in a 1:1 ratio with 784 septic patients who did not develop AKI. The overall 30-day attributable mortality of AKI was 6.6% (95% CI 2.3 ∼ 10.9%, p = 0.002). A subgroup analysis revealed that the 30-day attributable mortality rates for stage 1, stage 2, and stage 3 AKI were 0.6% (95% CI -5.9 ∼ 7.2%, p = 0.846), 4.7% (95% CI -3.1 ∼ 12.4%, p = 0.221) and 16.8% (95% CI 8.1 ∼ 25.2%, p < 0.001), respectively. Particularly noteworthy was that stage 3 AKI emerged as an independent risk factor for 30-day mortality, possessing an adjusted hazard ratio of 1.80 (95% CI 1.31 ∼ 2.47, p < 0.001).ConclusionsThe overall 30-day attributable mortality of AKI among critically ill patients with sepsis was 6.6%. Stage 3 AKI had the most significant contribution to 30-day mortality, while stage 1 and stage 2 AKI did not increase excess mortality.

Project description:BackgroundMechanical Ventilation (MV) is a complex and central treatment process in the care of critically ill patients. It influences acid-base balance and can also cause prognostically relevant biotrauma by generating forces and liberating reactive oxygen species, negatively affecting outcomes. In this work we evaluate the use of a Recurrent Neural Network (RNN) modelling to predict outcomes of mechanically ventilated patients, using standard mechanical ventilation parameters.MethodsWe performed our analysis on VENTILA dataset, an observational, prospective, international, multi-centre study, performed to investigate the effect of baseline characteristics and management changes over time on the all-cause mortality rate in mechanically ventilated patients in ICU. Our cohort includes 12,596 adult patients older than 18, associated with 12,755 distinct admissions in ICUs across 37 countries and receiving invasive and non-invasive mechanical ventilation. We carry out four different analysis. Initially we select typical mechanical ventilation parameters and evaluate the machine learning model on both, the overall cohort and a subgroup of patients admitted with respiratory disorders. Furthermore, we carry out sensitivity analysis to evaluate whether inclusion of variables related to the function of other organs, improve the predictive performance of the model for both the overall cohort as well as the subgroup of patients with respiratory disorders.ResultsPredictive performance of RNN-based model was higher with Area Under the Receiver Operating Characteristic (ROC) Curve (AUC) of 0.72 (± 0.01) and Average Precision (AP) of 0.57 (± 0.01) in comparison to RF and LR for the overall patient dataset. Higher predictive performance was recorded in the subgroup of patients admitted with respiratory disorders with AUC of 0.75 (± 0.02) and AP of 0.65 (± 0.03). Inclusion of function of other organs further improved the performance to AUC of 0.79 (± 0.01) and AP 0.68 (± 0.02) for the overall patient dataset and AUC of 0.79 (± 0.01) and AP 0.72 (± 0.02) for the subgroup with respiratory disorders.ConclusionThe RNN-based model demonstrated better performance than RF and LR in patients in mechanical ventilation and its subgroup admitted with respiratory disorders. Clinical studies are needed to evaluate whether it impacts decision-making and patient outcomes.Trial registrationNCT02731898 ( https://clinicaltrials.gov/ct2/show/NCT02731898 ), prospectively registered on April 8, 2016.