eICU Collaborative Research Database Research Articles

Machine-learning-derived sepsis bundle of care.

Compliance to the Surviving Sepsis Campaign (SSC) guidelines is limited. This is known to be associated with increased mortality. The aim of this retrospective cohort study was to identify among the SCC guidelines the optimal bundle of recommendations that minimize 28-day mortality. We used a training cohort to identify, using a least absolute shrinkage and selection operator penalized machine learning model, this bundle. Patients with sepsis/septic shock admitted to the intensive care unit (ICU) were extracted from two US databases, the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database (training and internal validation cohorts) and the eICU Collaborative Research Database (eICU-CRD) (external validation cohort). In the validation cohorts, we defined a bundle group that includes patients who were treated with at least all the recommendations selected in our bundle and a no-bundle group that includes patients in whom at least one recommendation from our bundle was omitted. All-cause 28-day mortality was the primary outcome measure. A total of 42,735 patients were included. Six recommendations (antimicrobials, balanced crystalloid, insulin therapy, corticosteroids, vasopressin, and bicarbonate therapy) were identified from the training cohort to be included in our bundle. In the propensity score-(PS)-matched internal validation cohort, the bundle group was associated with a lower mortality (OR 0.41 [0.33-0.53]; p < 0.001) compared to the no-bundle group. This was confirmed in the PS-matched external validation cohort (OR 0.75 [0.60-0.94]; p 0.02). Our bundle of six recommendations is associated with a dramatic reduction in mortality in sepsis and septic shock. This bundle needs to be evaluated prospectively.

Association between serum calcium level and in-hospital mortality in patients with acute myocardial infarction: a retrospective cohort study

The association between serum calcium levels and the prognosis of patients with acute myocardial infarction (AMI) remains controversial. This study aimed to explore the association between serum calcium and in-hospital mortality in patients with AMI. The data of this study were extracted from the Philips eICU Collaborative Research Database. A total of 7284 patients were eventually enrolled in this study, of which 799 (10.97%) died during hospitalization. For each patient, serum calcium, corrected to albumin, was calculated and categorized into four groups: Q1 ≤ 8.5, Q2 8.5–9.5, Q3 9.5–10.5, and Q4 > 10.5 mg/dL. Multivariate analysis demonstrated that corrected sCa was an independent predictor of in-hospital death (Q2 vs. Q1, OR 0.5, 95% CI 0.4–0.7, P < 0.001; Q3 vs. Q1, OR 0.8, 95% CI 0.6–1.0, P = 0.035; Q4 vs. Q1, OR 1.6, 95% CI 1.1–2.3, P = 0.008). The association remained stable in the fully adjusted model. A significant U-shaped association between corrected serum calcium and in-hospital mortality was observed in piecewise linear regression model (Corrected sCa < 9.4 mg/dL, OR 0.8, 95% CI 0.7–0.9, P < 0.001; corrected sCa > 9.4 mg/dL, OR 1.5, 95% CI 1.3–1.8, P < 0.001). In conclusion, both decreased and increased corrected serum calcium is associated with increased in-hospital mortality in patients with AMI, and patients may have the lowest risk of in-hospital death when corrected serum calcium is 9.4 mg/dL (2.35 mmol/L).

Machine learning models to predict in-hospital mortality in septic patients with diabetes.

Sepsis is a leading cause of morbidity and mortality in hospitalized patients. Up to now, there are no well-established longitudinal networks from molecular mechanisms to clinical phenotypes in sepsis. Adding to the problem, about one of the five patients presented with diabetes. For this subgroup, management is difficult, and prognosis is difficult to evaluate. From the three databases, a total of 7,001 patients were enrolled on the basis of sepsis-3 standard and diabetes diagnosis. Input variable selection is based on the result of correlation analysis in a handpicking way, and 53 variables were left. A total of 5,727 records were collected from Medical Information Mart for Intensive Care database and randomly split into a training set and an internal validation set at a ratio of 7:3. Then, logistic regression with lasso regularization, Bayes logistic regression, decision tree, random forest, and XGBoost were conducted to build the predictive model by using training set. Then, the models were tested by the internal validation set. The data from eICU Collaborative Research Database (n = 815) and dtChina critical care database (n = 459) were used to test the model performance as the external validation set. In the internal validation set, the accuracy values of logistic regression with lasso regularization, Bayes logistic regression, decision tree, random forest, and XGBoost were 0.878, 0.883, 0.865, 0.883, and 0.882, respectively. Likewise, in the external validation set 1, lasso regularization = 0.879, Bayes logistic regression = 0.877, decision tree = 0.865, random forest = 0.886, and XGBoost = 0.875. In the external validation set 2, lasso regularization= 0.715, Bayes logistic regression = 0.745, decision tree = 0.763, random forest = 0.760, and XGBoost = 0.699. The top three models for internal validation set were Bayes logistic regression, random forest, and XGBoost, whereas the top three models for external validation set 1 were random forest, logistic regression, and Bayes logistic regression. In addition, the top three models for the external validation set 2 were decision tree, random forest, and Bayes logistic regression. Random forest model performed well with the training and three validation sets. The most important features are age, albumin, and lactate.

A machine learning-based prediction model for in-hospital mortality among critically ill patients with hip fracture: An internal and external validated study

IntroductionFew studies have investigated the in-hospital mortality among critically ill patients with hip fracture. This study aimed to develop and validate a model to estimate the risk of in-hospital mortality among critically ill patients with hip fracture. MethodsFor this study, data from the Medical Information Mart for Intensive Care III (MIMIC-III) Database and electronic Intensive Care Unit (eICU) Collaborative Research Database were evaluated. Enrolled patients (n=391) in the MIMIC-III database were divided into a training (2/3, n=260) and a validation (1/3, n=131) group at random. Using machine learning algorithms such as random forest, gradient boosting machine, decision tree, and eXGBoosting machine approach, the training group was utilized to train and optimize models. The validation group was used to internally validate models and the optimal model could be obtained in terms of discrimination (area under the receiver operating characteristic curve, AUROC) and calibration (calibration curve). External validation was done in the eICU Collaborative Research Database (n=165). To encourage practical use of the model, a web-based calculator was developed according to the eXGBoosting machine approach. ResultsThe in-hospital death rate was 13.81% (54/391) in the MIMIC-III database and 10.91% (18/165) in the eICU Collaborative Research Database. Age, gender, anemia, mechanical ventilation, cardiac arrest, and chronic airway obstruction were the six model parameters which were identified using the Least Absolute Shrinkage and Selection Operator (LASSO) method combined with 10-fold cross-validation. The model established using the eXGBoosting machine approach showed the highest area under curve (AUC) value (0.797, 95% CI: 0.696-0.898) and the best calibrating ability, with a calibration slope of 0.999 and intercept of -0.019. External validation also revealed favorable discrimination (AUC: 0.715, 95% CI: 0.566-0.864; accuracy: 0.788) and calibration (calibration slope: 0.805) in the eICU Collaborative Research Database. The web-based calculator could be available at https://doctorwangsj-webcalculator-main-yw69yd.streamlitapp.com/. ConclusionThe model has the potential to be a pragmatic risk prediction tool that is able to identify hip fracture patients who are at a high risk of in-hospital mortality in ICU settings, guide patient risk counseling, and simplify prognosis bench-marking by controlling for baseline risk.

Predictive value of the hemoglobin-albumin-lymphocyte-platelet (HALP) index for ICU mortality in patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD).

The combined index of hemoglobin, albumin, lymphocyte, and platelet (HALP) is a novel indicator reflecting systemic inflammation and nutritional status. To explore the relationship between HALP score and ICU mortality risk in patients with acute exacerbations of chronic obstructive pulmonary disease (AECOPD). A total of 1533 AECOPD patients from the eICU Collaborative Research Database (eICU-CRD) between 2014 and 2015 were included in this retrospective cohort study. Univariate and multivariate Cox proportional hazards models were utilized to investigate the association of HALP score, platelet-to-lymphocyte ratio (PLR) score, and lymphocyte-to-monocyte ratio (LMR) score with the ICU mortality risk in patients with AECOPD. Stratified analyses were performed based on patients' ICU admission type, body mass index (BMI), and Acute Physiology, Age and Chronic Health Evaluation IV (APACHE IV) score. Of these 1533 AECOPD patients, 123 (8.00%) patients died in the ICU. Low HALP score [hazard ratio (HR) = 1.69; 95% confidence interval (CI) 1.14-2.53] and low LMR score (HR = 1.60; 95% CI 1.07-2.39) were associated with an increased ICU mortality risk in patients with AECOPD after adjusting for all confounders. Stratified analyses indicated that low HALP score were still associated with a higher ICU mortality risk in patients admitted to ICU by emergency (HR = 1.81; 95% CI 1.11-2.96), obese patients (HR = 2.81; 95% CI 1.29-6.10), and patients with low APACHE scores (HR = 2.87; 95% CI 1.75-4.69). Low HALP score was associated with an increased risk of ICU mortality in patients with AECOPD, suggesting that the HALP score may be a novel prognostic predictor in patients with AECOPD.

Development and deployment of interpretable machine-learning model for predicting in-hospital mortality in elderly patients with acute kidney disease

Background Acute kidney injury (AKI) is more likely to develop in the elderly admitted to the intensive care unit (ICU). Acute kidney disease (AKD) affects ∼45% of patients with AKI and increases short-term mortality. However, there are no studies on the prognosis of AKD in the elderly. Methods Data from 2666 elderly patients with AKD in the Medical Information Mart for Intensive Care IV were used for model development and 535 in the eICU Collaborative Research Database for external validation. Based on 5 machine learning algorithms, 33 noninvasive parameters were extracted as features for modeling. Results In-hospital mortality of AKD in the elderly was 29.6% and 31.8% in development and validation cohorts, respectively. The comprehensive best-performing algorithm was the support vector machine (SVM), and a simplified online application included only 10 features employing SVM (AUC: 0.810 and 0.776 in the training and external validation cohorts, respectively) was deployed. Model interpretation by SHapley Additive exPlanation (SHAP) values revealed that the difference (AKD day – ICU day) in sequential organ failure assessment (delta SOFA), Glasgow coma scale (GCS), delta GCS, delta peripheral oxygen saturation (SpO2), and SOFA were the top five features associated with prognosis. The optimal target was determined by SHAP values from partial dependence plots. Conclusions A web-based tool was externally validated and deployed to predict the early prognosis of AKD in the elderly based on readily available noninvasive parameters, assisting clinicians in intervening with precision and purpose to save lives to the greatest extent.

Machine-learning models for prediction of sepsis patients mortality

ObjectivesSepsis is an infection-caused syndrome, that leads to life-threatening organ damage. We aim to develop machine learning models with large-scale data to predict sepsis patients’ mortality. Designwe extracted sepsis patients from two databases, Medical Information Mart for Intensive Care IV (MIMIC-IV) as a train set and Philips eICU Collaborative Research Database as a test set. SettingICUs in multicenter hospitals in the USA during 2012–2019. Patients or participantsA total of 21,680 sepsis-3 patients are included in the study, in which, 3771 patients were dead and 17,909 survived during hospitalization, respectively. InterventionsNo interventions. Main variables of interestBasic information, examination items during hospitalization and some medication and treatment information are incorporated into analyzed. Seven different models were built with a Support vector machine, Decision Tree Classifier, Random Forest, Gradients Boosting, Multiple Layer Perception, Xgboost, light Gradients Boosting to predict dead or live during hospitalization. ResultsAlgorithms with an AUC value in the test set of the top three: light GBM, GBM, Xgboost. Considering the performance of the training set and the test set, the light GBM model performs best, and then the parameters of the model were adjusted, after that the AUC value was 0.99 in the train set, 0.96 in the test set, respectively. ConclusionsModels built with light GBM algorithm from real-world sepsis patients from electronic health records accurately predict whether sepsis patients are dead and can be incorporated into clinical decision tools to enhance the prognosis of the patient and prevent adverse outcomes.

Association between body mass index change and mortality in critically ill patients: A retrospective observational study

ObjectivePrevious studies have emphasized the association between baseline body mass index (BMI) and mortality in patients during a stay in the intensive care unit (ICU). However, to our knowledge, few studies have focused on BMI change during an ICU stay. The aim of this study was to explore the prognostic value of BMI change during ICU hospitalization. MethodsThis was a multicenter, retrospective cohort study with data extracted from the eICU Collaborative Research Database. Logistic regression models were used to explore the relationship between BMI change and mortality in ICU patients. BMI change was calculated as follows: {[discharge ICU weight (kg) – admission ICU weight (kg)] / height (m)2]}.Interaction and subgroup analyses were conducted for patients grouped with baseline BMI on ICU admission (≥30 versus 25–29.9 versus <25 kg/m2), Acute Physiology and Chronic Health Evaluation (APACHE) IV score (<53 versus ≥53), and ICU length of stay (≥3 versus <3 d). ResultsCompared with those with weight loss (n = 17 134), patients with weight gain during ICU hospitalization (n = 17 436) were associated with higher hospital mortality (odds ratio [OR], 1.251; 95% confidence interval [CI], 1.155–1.356; P < 0.001) and ICU mortality (OR, 1.360; 95% CI, 1.227–1.506; P < 0.001) after multivariable adjustment. The associations remained robust in patients with different baseline BMI levels and were especially remarkable among those with higher APACHE IV score and the longer ICU stay. ConclusionsThe present study exposed the potential hazard of increasing BMI for hospital and ICU mortalities during ICU hospitalization and indicating that patients in the ICU may benefit from a more balanced nutritional strategy.

Conservative oxygen therapy in critically ill and perioperative period of patients with sepsis-associated encephalopathy.

Sepsis-associated encephalopathy (SAE) patients in the intensive care unit (ICU) and perioperative period are administrated supplemental oxygen. However, the correlation between oxygenation status with SAE and the target for oxygen therapy remains unclear. This study aimed to examine the relationship between oxygen therapy and SAE patients. Patients diagnosed with sepsis 3.0 in the intensive care unit (ICU) were enrolled. The data were collected from the Medical Information Mart for Intensive Care IV (MIMIC IV) database and the eICU Collaborative Research Database (eICU-CRD) database. The generalized additive models were adopted to estimate the oxygen therapy targets in SAE patients. The results were confirmed by multivariate Logistic, propensity score analysis, inversion probability-weighting, doubly robust model, and multivariate COX analyses. Survival was analyzed by the Kaplan-Meier method. A total of 10055 patients from eICU-CRD and 1685 from MIMIC IV were included. The incidence of SAE patients was 58.43%. The range of PaO2 (97-339) mmHg, PaO2/FiO2 (189-619), and SPO2≥93% may reduce the incidence of SAE, which were verified by multivariable Logistic regression, propensity score analysis, inversion probability-weighting, and doubly robust model estimation in MIMIC IV database and eICU database. The range of PaO2/FiO2 (189-619) and SPO2≥93% may reduce the hospital mortality of SAE were verified by multivariable COX regression. SAE patients in ICU, including perioperative period, require conservative oxygen therapy. We should maintain SPO2≥93%, PaO2 (97-339) mmHg and PaO2/FiO2 (189-619) in SAE patients.

Lactate-to-albumin ratio is associated with in-hospital mortality in patients with spontaneous subarachnoid hemorrhage and a nomogram model construction.

Subarachnoid hemorrhage (SAH) is a severe hemorrhagic stroke with high mortality. However, there is a lack of clinical tools for predicting in-hospital mortality in clinical practice. LAR is a novel clinical marker that has demonstrated prognostic significance in a variety of diseases. Critically ill patients diagnosed and SAH with their data in the Medical Information Mart for Intensive Care-IV (MIMIC-IV) database and the eICU Collaborative Research Database (eICU-CRD) were included in our study. Multivariate logistic regression was utilized to establish the nomogram. A total of 244 patients with spontaneous SAH in the MIMIC-IV database were eligible for the study as a training set, and 83 patients in eICU-CRD were included for external validation. Data on clinical characteristics, laboratory parameters and outcomes were collected. Univariate and multivariate logistic regression analysis identified age (OR: 1.042, P-value: 0.003), LAR (OR: 2.592, P-value: 0.011), anion gap (OR: 1.134, P-value: 0.036) and APSIII (OR: 1.028, P-value: < 0.001) as independent predictors of in-hospital mortality and we developed a nomogram model based on these factors. The nomogram model incorporated with LAR, APSIII, age and anion gap demonstrated great discrimination and clinical utility both in the training set (accuracy: 77.5%, AUC: 0.811) and validation set (accuracy: 75.9%, AUC: 0.822). LAR is closely associated with increased in-hospital mortality of patients with spontaneous SAH, which could serve as a novel clinical marker. The nomogram model combined with LAR, APSIII, age, and anion gap presents good predictive performance and clinical practicability.

Interpretable machine learning for 28-day all-cause in-hospital mortality prediction in critically ill patients with heart failure combined with hypertension: A retrospective cohort study based on medical information mart for intensive care database-IV and eICU databases.

BackgroundHeart failure (HF) combined with hypertension is an extremely important cause of in-hospital mortality, especially for the intensive care unit (ICU) patients. However, under intense working pressure, the medical staff are easily overwhelmed by the large number of clinical signals generated in the ICU, which may lead to treatment delay, sub-optimal care, or even wrong clinical decisions. Individual risk stratification is an essential strategy for managing ICU patients with HF combined with hypertension. Artificial intelligence, especially machine learning (ML), can develop superior models to predict the prognosis of these patients. This study aimed to develop a machine learning method to predict the 28-day mortality for ICU patients with HF combined with hypertension.MethodsWe enrolled all critically ill patients with HF combined with hypertension in the Medical Information Mart for IntensiveCare Database-IV (MIMIC-IV, v.1.4) and the eICU Collaborative Research Database (eICU-CRD) from 2008 to 2019. Subsequently, MIMIC-IV was divided into training cohort and testing cohort in an 8:2 ratio, and eICU-CRD was designated as the external validation cohort. The least absolute shrinkage and selection operator (LASSO) Cox regression with internal tenfold cross-validation was used for data dimension reduction and identifying the most valuable predictive features for 28-day mortality. Based on its accuracy and area under the curve (AUC), the best model in the validation cohort was selected. In addition, we utilized the Shapley Additive Explanations (SHAP) method to highlight the importance of model features, analyze the impact of individual features on model output, and visualize an individual’s Shapley values.ResultsA total of 3,458 and 6582 patients with HF combined with hypertension in MIMIC-IV and eICU-CRD were included. The patients, including 1,756 males, had a median (Q1, Q3) age of 75 (65, 84) years. After selection, 22 out of a total of 58 clinical parameters were extracted to develop the machine-learning models. Among four constructed models, the Neural Networks (NN) model performed the best predictive performance with an AUC of 0.764 and 0.674 in the test cohort and external validation cohort, respectively. In addition, a simplified model including seven variables was built based on NN, which also had good predictive performance (AUC: 0.741). Feature importance analysis showed that age, mechanical ventilation (MECHVENT), chloride, bun, anion gap, paraplegia, rdw (RDW), hyperlipidemia, peripheral capillary oxygen saturation (SpO2), respiratory rate, cerebrovascular disease, heart rate, white blood cell (WBC), international normalized ratio (INR), mean corpuscular hemoglobin concentration (MCHC), glucose, AIDS, mean corpuscular volume (MCV), N-terminal pro-brain natriuretic peptide (Npro. BNP), calcium, renal replacement therapy (RRT), and partial thromboplastin time (PTT) were the top 22 features of the NN model with the greatest impact. Finally, after hyperparameter optimization, SHAP plots were employed to make the NN-based model interpretable with an analytical description of how the constructed model visualizes the prediction of death.ConclusionWe developed a predictive model to predict the 28-day mortality for ICU patients with HF combined with hypertension, which proved superior to the traditional logistic regression analysis. The SHAP method enables machine learning models to be more interpretable, thereby helping clinicians to better understand the reasoning behind the outcome and assess in-hospital outcomes for critically ill patients.

High resolution data modifies intensive care unit dialysis outcome predictions as compared with low resolution administrative data set.

High resolution clinical databases from electronic health records are increasingly being used in the field of health data science. Compared to traditional administrative databases and disease registries, these newer highly granular clinical datasets offer several advantages, including availability of detailed clinical information for machine learning and the ability to adjust for potential confounders in statistical models. The purpose of this study is to compare the analysis of the same clinical research question using an administrative database and an electronic health record database. The Nationwide Inpatient Sample (NIS) was used for the low-resolution model, and the eICU Collaborative Research Database (eICU) was used for the high-resolution model. A parallel cohort of patients admitted to the intensive care unit (ICU) with sepsis and requiring mechanical ventilation was extracted from each database. The primary outcome was mortality and the exposure of interest was the use of dialysis. In the low resolution model, after controlling for the covariates that are available, dialysis use was associated with an increased mortality (eICU: OR 2.07, 95% CI 1.75-2.44, p<0.01; NIS: OR 1.40, 95% CI 1.36-1.45, p<0.01). In the high-resolution model, after the addition of the clinical covariates, the harmful effect of dialysis on mortality was no longer significant (OR 1.04, 95% 0.85-1.28, p = 0.64). The results of this experiment show that the addition of high resolution clinical variables to statistical models significantly improves the ability to control for important confounders that are not available in administrative datasets. This suggests that the results from prior studies using low resolution data may be inaccurate and may need to be repeated using detailed clinical data.

A Machine Learning Approach for the Prediction of Severe Acute Kidney Injury Following Traumatic Brain Injury.

Acute kidney injury (AKI), a prevalent non-neurological complication following traumatic brain injury (TBI), is a major clinical issue with an unfavorable prognosis. This study aimed to develop and validate machine learning models to predict severe AKI (stage 3 or greater) incidence in patients with TBI. A retrospective cohort study was conducted by using two public databases: the Medical Information Mart for Intensive Care IV (MIMIC)-IV and the eICU Collaborative Research Database (eICU-CRD). Recursive feature elimination was used to select candidate predictors obtained within 24h of intensive care unit admission. The area under the curve and decision curve analysis curves were used to determine the discriminatory ability. On the other hand, the calibration curve was employed to evaluate the calibrated performance of the newly developed machine learning models. In the MIMIC-IV database, there were 808 patients diagnosed with moderate and severe TBI (msTBI) (msTBI is defined as Glasgow Coma Score < 12). Of these, 60 (7.43%) patients experienced severe AKI. External validation in the eICU-CRD indicated that the random forest (RF) model had the highest area under the curve of 0.819 (95% confidence interval 0.783-0.851). Furthermore, in the calibration curve, the RF model was well calibrated (P = 0.795). In this study, the RF model demonstrated better discrimination in predicting severe AKI than other models. An online calculator could facilitate its application, potentially improving the early detection of severe AKI and subsequently improving the clinical outcomes among patients with msTBI.

Lactate/Albumin Ratio Could be a Prognostic Indicator for Patients with Post-operative Intestinal Obstruction

Objective: This study aimed to investigate the value of lactate/albumin ratio in prediction of mortality in intensive care unit (ICU) patients with post-operative intestinal obstruction. Method: A retrospective analysis was conducted with patient data from the eICU Collaborative Research Database (eICU-CRD). According to the outcome, the patients with post-operative intestinal obstruction who were admitted to ICU within 24 hours after surgery were separated into two groups: survivors (233 cases) and non-survivors (35 cases). Their clinical characteristics and scoring data were collected. Logistic regression analysis was used to evaluate the risk factors for death, and these risk factors were further included for the construction of the receiver operating characteristic curve (ROC) to evaluate the predictive value of death for these patients. Results: In-hospital mortality for patients admitted to ICU with post-operative intestinal obstruction was 13.1% (35/268). The level of lactate/albumin ratio was significantly higher in non-survivors than in survivors (1.36±1.54 versus 0.70±0.64; P&lt;0.001). Logistic regression analysis showed that the lactate/albumin ratio (OR=0.667, 95%CI: 1.328-2.485, P=0.001) could predict in-hospital mortality independently for the patients in ICU with post-operative intestinal obstruction. Further analysis showed that the area under the ROC curve (AUC) value of lactate/albumin ratio level was 0.681. Conclusions: These data suggested that the lactate/albumin ratio has potential predictive value for mortality in ICU patients with post-operative intestinal obstruction.

Identification of key predictors of hospital mortality in critically ill patients with embolic stroke using machine learning.

Embolic stroke (ES) is characterized by high morbidity and mortality. Its mortality predictors remain unclear. The present study aimed to use machine learning (ML) to identify the key predictors of mortality for ES patients in the intensive care unit (ICU). Data were extracted from two large ICU databases: Medical Information Mart for Intensive Care (MIMIC)-IV for training and internal validation, and eICU Collaborative Research Database (eICU-CRD) for external validation. We developed predictive models of ES mortality based on 15 ML algorithms. We relied on the synthetic minority oversampling technique (SMOTE) to address class imbalance. Our main performance metric was area under the receiver operating characteristic (AUROC). We adopted recursive feature elimination (RFE) for feature selection. We assessed model performance using three disease-severity scoring systems as benchmarks. Of the 1566 and 207 ES patients enrolled in the two databases, there were 173 (15.70%), 73 (15.57%), and 36 (17.39%) hospital mortality in the training, internal validation, and external validation cohort, respectively. The random forest (RF) model had the largest AUROC (0.806) in the internal validation phase and was chosen as the best model. The AUROC of the RF compact (RF-COM) model containing the top six features identified by RFE was 0.795. In the external validation phase, the AUROC of the RF model was 0.838, and the RF-COM model was 0.830, outperforming other models. Our findings suggest that the RF model was the best model and the top six predictors of ES hospital mortality were Glasgow Coma Scale, white blood cell, blood urea nitrogen, bicarbonate, age, and mechanical ventilation.

No Association of Early Postoperative Heart Rate With Outcomes After Coronary Artery Bypass Grafting.

Elevated perioperative heart rate potentially causes perioperative myocardial injury because of imbalance in oxygen supply and demand. However, large multicenter studies evaluating early postoperative heart rate and major adverse cardiac and cerebrovascular events (MACCEs) are lacking. To assess the associations of 4 postoperative heart rate assessment methods with in-hospital MACCEs after elective coronary artery bypass grafting (CABG). Using data from the eICU Collaborative Research Database in the United States from 2014 to 2015, the study evaluated postoperative heart rate measured during hospitalization within 24 hours after intensive care unit admission. Four heart rate assessment methods were evaluated: maximum heart rate, duration above heart rate 100/min, area above heart rate 100/min, and time-weighted average heart rate. The outcome was in-hospital MACCEs, defined as a composite of in-hospital death, myocardial infarction, angina, arrhythmia, heart failure, stroke, cardiac arrest, or repeat revascularization. Among 2585 patients, the crude rate of in-hospital MACCEs was 6.2%. In multivariable logistic regression analysis, the adjusted odds ratios (95% CI) for in-hospital MAC-CEs assessed by maximum heart rate in each heart rate category (beats per minute: >100-110, >110-120, >120-130, and >130) were 1.43 (0.95-2.15), 0.98 (0.56-1.64), 1.47 (0.76-2.69), and 1.71 (0.80-3.35), respectively. Similarly, none of the other 3 methods were associated with MACCEs. More research is needed to assess the usefulness of heart rate measurement in patients after CABG.

Association between serum chloride levels with mortality in critically ill patients with acute kidney injury: An observational multicenter study employing the eICU database.

ObjectiveThe effect of the serum chloride (Cl) level on mortality in critically ill patients with acute kidney injury (AKI) remains unknown. We sought an association between mortality and serum Cl.MethodsWe identified AKI patients in the eICU Collaborative Research Database from 2014 to 2015 at 208 US hospitals. The outcomes included in-hospital and intensive care unit (ICU) mortality. Time-varying covariates Cox regression models and the Kaplan-Meier (K-M) curves were used to assess the association between serum Cl levels and mortality. Multivariable adjusted restricted cubic spline models were used to analyze the potential nonlinear relationship between mortality and serum Cl.ResultsIn total, 4,234 AKI patients were included in the study. Compared with normochloremia (98≤chloride<108mEq/L), hypochloremia (Cl<98mEq/L) was associated with mortality (adjusted hazard ratio [HR] for in-hospital mortality 1.46, 95% confidence interval [CI] 1.20–1.80, P = 0.0003; adjusted HR for ICU mortality 1.37, 95% CI 1.05–1.80, P = 0.0187). Hyperchloremia showed no significant difference in mortality compared to normochloremia (adjusted HR for in-hospital mortality 0.89, 95% CI 0.76–1.04, P = 0.1438; adjusted HR for ICU mortality 0.87, 95% CI 0.72–1.06, P = 0.1712). Smoothing curves revealed continuous non-linear associations between serum Cl levels and mortality. The K-M curve showed that patients with hypochloremia presented with a lower survival rate.ConclusionsLower serum Cl levels after ICU admission was associated with increased in-hospital and ICU mortality in critically ill patients with AKI. The results should be verified in well-designed prospective studies.

Association of longitudinal platelet count trajectory with ICU mortality: A multi-cohort study.

ObjectivePlatelet (PLT) engages in immune and inflammatory responses, all of which are related to the prognosis of critically ill patients. Although thrombocytopenia at ICU admission contributes to in-hospital mortality, PLT is repeatedly measured during ICU hospitalization and the role of longitudinal PLT trajectory remains unclear. We aimed to identify dynamic PLT trajectory patterns and evaluate their relationships with mortality risk and thrombocytopenia.MethodsWe adopted a three-phase, multi-cohort study strategy. Firstly, longitudinal PLT trajectory patterns within the first four ICU days and their associations with 28-day survival were tested in the eICU Collaborative Research Database (eICU-CRD) and independently validated in the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Secondly, the relationships among PLT trajectory patterns, thrombocytopenia, and 28-day mortality were explored and validated. Finally, a Mortality GRade system for ICU dynamically monitoring patients (Mortality-GRID) was developed to quantify the mortality risk based on longitudinal PLT, which was further validated in the Molecular Epidemiology of Acute Respiratory Distress Syndrome (MEARDS) cohort.ResultsA total of 35,332 ICU patients were included from three cohorts. Trajectory analysis clustered patients into ascending (AS), stable (ST), or descending (DS) PLT patterns. DS patients with high baseline PLT decline quickly, resulting in poor prognosis. AS patients have low baseline PLT but recover quickly, favoring a better prognosis. ST patients maintain low PLT, having a moderate prognosis in between (HR ST vs AS = 1.26, 95% CI: 1.14–1.38, P = 6.15 × 10−6; HR DS vs AS = 1.58, 95% CI: 1.40–1.79, P = 1.41 × 10−13). The associations remained significant in patients without thrombocytopenia during the entire ICU hospitalization and were robust in sensitivity analyses and stratification analyses. Further, the trajectory pattern was a warning sign of thrombocytopenia, which mediated 27.2% of the effects of the PLT trajectory on 28-day mortality (HR indirect = 1.11, 95% CI: 1.06–1.17, P = 9.80 × 10−6). Mortality-GRID well predicts mortality risk, which is in high consistency with that directly estimated in MEARDS (r = 0.98, P = 1.30 × 10−23).ConclusionLongitudinal PLT trajectory is a complementary predictor to baseline PLT for patient survival, even in patients without risk of thrombocytopenia. Mortality-GRID could identify patients at high mortality risk.

Developing an ensemble machine learning model for early prediction of sepsis-associated acute kidney injury

SummarySepsis-associated acute kidney injury (S-AKI) is very common and early prediction is beneficial. This study aiming to develop an accurate ensemble model to predict the risk of S-AKI based on easily available clinical information. Patients with sepsis from the United States (US) database Medical Information Mart for Intensive Care-IV were used as a modeling cohort to predict the occurrence of AKI by combining Support Vector Machine, Random Forest, Neural Network, and Extreme Gradient Boost as four first-level learners via stacking algorithm. The external validation databases were the eICU Collaborative Research Database from US and Critical Care Database comprising infection patients at Zigong Fourth People’s Hospital from China, whose AUROC values for the ensemble model 48–12 h before the onset of AKI were 0.774–0.788 and 0.756–0.813, respectively. In this study, an ensemble model for early prediction of S-AKI onset was developed and it demonstrated good performance in multicenter external datasets.

A nomogram to predict prolonged stay of obesity patients with sepsis in ICU: Relevancy for predictive, personalized, preventive, and participatory healthcare strategies.

ObjectiveIn an era of increasingly expensive intensive care costs, it is essential to evaluate early whether the length of stay (LOS) in the intensive care unit (ICU) of obesity patients with sepsis will be prolonged. On the one hand, it can reduce costs; on the other hand, it can reduce nosocomial infection. Therefore, this study aimed to verify whether ICU prolonged LOS was significantly associated with poor prognosis poor in obesity patients with sepsis and develop a simple prediction model to personalize the risk of ICU prolonged LOS for obesity patients with sepsis.MethodIn total, 14,483 patients from the eICU Collaborative Research Database were randomized to the training set (3,606 patients) and validation set (1,600 patients). The potential predictors of ICU prolonged LOS among various factors were identified using logistic regression analysis. For internal and external validation, a nomogram was developed and performed.ResultsICU prolonged LOS was defined as the third quartile of ICU LOS or more for all sepsis patients and demonstrated to be significantly associated with the mortality in ICU by logistic regression analysis. When entering the ICU, seven independent risk factors were identified: maximum white blood cell, minimum white blood cell, use of ventilation, Glasgow Coma Scale, minimum albumin, maximum respiratory rate, and minimum red blood cell distribution width. In the internal validation set, the area under the curve was 0.73, while in the external validation set, it was 0.78. The calibration curves showed that this model predicted probability due to actually observed probability. Furthermore, the decision curve analysis and clinical impact curve showed that the nomogram had a high clinical net benefit.ConclusionIn obesity patients with sepsis, we created a novel nomogram to predict the risk of ICU prolonged LOS. This prediction model is accurate and reliable, and it can assist patients and clinicians in determining prognosis and making clinical decisions.