Prediction Model For In-hospital Mortality Research Articles

Validation and refinement of a predictive nomogram using artificial intelligence: assessing in-hospital mortality in patients with large hemispheric cerebral infarction.

Large Hemispheric Infarction (LHI) poses significant mortality and morbidity risks, necessitating predictive models for in-hospital mortality. Previous studies have explored LHI progression to malignant cerebral edema (MCE) but have not comprehensively addressed in-hospital mortality risk, especially in non-decompressive hemicraniectomy (DHC) patients. Demographic, clinical, risk factor, and laboratory data were gathered. The population was randomly divided into Development and Validation Groups at a 3:1 ratio, with no statistically significant differences observed. Variable selection utilized the Bonferroni-corrected Boruta technique (p < 0.01). Logistic Regression retained essential variables, leading to the development of a nomogram. ROC and DCA curves were generated, and calibration was conducted based on the Validation Group. This study included 314 patients with acute anterior-circulating LHI, with 29.6% in the Death group (n = 93). Significant variables, including Glasgow Coma Score, Collateral Score, NLR, Ventilation, Non-MCA territorial involvement, and Midline Shift, were identified through the Boruta algorithm. The final Logistic Regression model led to a nomogram creation, exhibiting excellent discriminative capacity. Calibration curves in the Validation Group showed a high degree of conformity with actual observations. DCA curve analysis indicated substantial clinical net benefit within the 5 to 85% threshold range. We have utilized NIHSS score, Collateral Score, NLR, mechanical ventilation, non-MCA territorial involvement, and midline shift to develop a highly accurate, user-friendly nomogram for predicting in-hospital mortality in LHI patients. This nomogram serves as valuable reference material for future studies on LHI patient prognosis and mortality prevention, while addressing previous research limitations.

DERIVATION OF A MORTALITY PREDICTION MODEL IN CRITICAL CARE PATIENTS WITH CIRRHOSIS AND SEPSIS.

Objective : The aim of the study is to develop a predictive model for in-hospital mortality in critically ill patients with cirrhosis and sepsis, using clinical and laboratory data. Design : This is a retrospective cohort study. Setting: Medical and mixed intensive care units (ICUs) of a tertiary medical center. Patients : Cirrhotic adults were admitted with sepsis to the ICUs from January of 2007 to May of 2017. Interventions : None. Measurements and Main Results : Of 2,595 ICU admissions of patients with cirrhosis, 277 with first ICU admission for sepsis were included in the analysis, and 37% died in the hospital. Patients who stayed in the ICU for at least 6 h (n = 275) were considered for the multivariate model. Ten-fold cross-validation was used to estimate best parameter values and model performance, and the final model was chosen as the model maximizing area under the receiver-operating characteristic curve. Variables in order of impact were Acute Physiology and Chronic Health Evaluation (APACHE) III score, initial serum lactate, conjugated bilirubin, serum creatinine, model for end-stage liver disease score, age, body mass index, and serum hemoglobin. The final best model from cross-validation presented an area under the receiver operator characteristic curve (AUC) of 0.75, using a cut-point of 50% estimated probability, sensitivity and specificity were 0.46 and 0.90, respectively, with positive predictive value of 0.72 and negative predictive value of 0.74. These results were similar to the APACHE III only model (AUC = 0.74, sensitivity = 0.43, specificity = 0.89, positive predictive value = 0.69, negative predictive value = 0.73). Conclusion : The combination of initial serum lactate level, conjugated bilirubin, initial serum creatinine, model for end-stage liver disease score, age, body mass index, and serum hemoglobin did not yield meaningful improvement in the AUC and did not provide advantage over the APACHE III score for the prediction of in-hospital mortality in critically ill patients with cirrhosis and sepsis.

Development and evaluation of regression tree models for predicting in-hospital mortality of a national registry of COVID-19 patients over six pandemic surges.

Objective prognostic information is essential for good clinical decision making. In case of unknown diseases, scarcity of evidence and limited tacit knowledge prevent obtaining this information. Prediction models can be useful, but need to be not only evaluated on how well they predict, but also how stable these models are under fast changing circumstances with respect to development of the disease and the corresponding clinical response. This study aims to provide interpretable and actionable insights, particularly for clinicians. We developed and evaluated two regression tree predictive models for in-hospital mortality of COVID-19 patient at admission and 24 hours (24 h) after admission, using a national registry. We performed a retrospective analysis of observational routinely collected data. Two regression tree models were developed for admission and 24 h after admission. The complexity of the trees was managed via cross validation to prevent overfitting. The predictive ability of the model was assessed via bootstrapping using the Area under the Receiver-Operating-Characteristic curve, Brier score and calibration curves. The tree models were assessed on the stability of their probabilities and predictive ability, on the selected variables, and compared to a full-fledged logistic regression model that uses variable selection and variable transformations using splines. Participants included COVID-19 patients from all ICUs participating in the Dutch National Intensive Care Evaluation (NICE) registry, who were admitted at the ICU between February 27, 2020, and November 23, 2021. From the NICE registry, we included concerned demographic data, minimum and maximum values of physiological data in the first 24 h of ICU admission and diagnoses (reason for admission as well as comorbidities) for model development. The main outcome measure was in-hospital mortality. We additionally analysed the Length-of-Stay (LoS) per patient subgroup per survival status. A total of 13,369 confirmed COVID-19 patients from 70 ICUs were included (with mortality rate of 28%). The optimism-corrected AUROC of the admission tree (with seven paths) was 0.72 (95% CI: 0.71-0.74) and of the 24 h tree (with 11 paths) was 0.74 (0.74-0.77). Both regression trees yielded good calibration and variable selection for both trees was stable. Patient subgroups comprising the tree paths had comparable survival probabilities as the full-fledged logistic regression model, survival probabilities were stable over six COVID-19 surges, and subgroups were shown to have added predictive value over the individual patient variables. We developed and evaluated regression trees, which operate at par with a carefully crafted logistic regression model. The trees consist of homogenous subgroups of patients that are described by simple interpretable constraints on patient characteristics thereby facilitating shared decision-making.

Creating an In-hospital mortality prediction model for Indian patients with UA/ NSTEMI using Machine learning-based techniques

Creating an In-hospital mortality prediction model for Indian patients with UA/ NSTEMI using Machine learning-based techniques

2956. Increased Microbiota Alpha Diversity is Associated with Decreased Hospital Mortality in Medical Intensive Care Unit Patients

Abstract Background Characteristics of the gut microbiome may be useful to predict resistance to colonization and expansion of pathogenic organisms such as C. difficile. In this study, we evaluated whether inclusion of microbiome characteristics improved a clinical prediction model for in-hospital mortality. Methods Rectal or fecal swab samples were collected daily from medical intensive care unit (ICU) patients from three hospitals between 2017-2019 and underwent 16S rRNA gene sequencing of the V4 region. Patient admissions with ≥1 rectal swab collected &amp;lt; 48hrs from ICU admission were randomly selected 2:1 for calibration and validation analyses. Independent variables included Shannon Index, Microbiome Health Index, Community Type (CT) and relative abundance (RA) of specific bacterial taxa based on 3% sequence difference or class-level classification. The RA of each bacterial taxon was range-standardized (95%) and scaled to analyze a 10% relative difference within each taxon’s range. Logistic regression models identified microbiome characteristics at admission that were associated with in-hospital mortality. Fixed effects were included in all models to account for microbiome differences between facilities. Results There were 1365 patient admissions (910 calibration, 455 validation) analyzed (Table), of which 134 (96 calibration, 38 validation) died during hospitalization. Increase in Shannon Index by 1 [OR 0.66 (0.51-0.84), p=0.001], 10% relative increase in the RA of Bacilli [OR 1.15 (1.08-1.22), p&amp;lt; 0.001] and 10% relative increase in the RA of Enterococcus [OR 1.10 (1.04-1.14), p=0.001] were associated with in-hospital mortality in calibration models after controlling for Charlson Comorbidity Index [OR 1.23 (1.15-1.32), p&amp;lt; 0.001]. Increase in Shannon Index by 1 [OR 0.88 (0.57-1.38), p=0.581] had the same direction and similar odds in the validation cohort (Figure), supporting the findings of the calibration model. Conclusion In this secondary analysis of a large, multicenter, prospective medical ICU cohort, we found that higher alpha diversity of the gut microbiome was associated with decreased odds of in-hospital mortality. Disclosures Michael Z. David, MD PhD, Covance: Grant/Research Support|GSK: Advisor/Consultant|GSK: Grant/Research Support Erik R. Dubberke, MD, MSPH, Abbott: Advisor/Consultant|AstraZeneca: Advisor/Consultant|Ferring Pharmaceuticals: Advisor/Consultant|Ferring Pharmaceuticals: Grant/Research Support|Merck and Co.: Advisor/Consultant|Pfizer: Advisor/Consultant|Pfizer: Grant/Research Support|Seres Therapeutics: Advisor/Consultant|Summit: Advisor/Consultant|Theriva Biologics: Grant/Research Support Vincent B. Young, MD/PhD, ASM: Senior Editor|Debiopharm: Advisor/Consultant|Vedanta Biosciences: Advisor/Consultant

In-Hospital Mortality Prediction Model for Critically Ill Older Adult Patients Transferred from the Emergency Department to the Intensive Care Unit.

Studies on the prognosis of critically ill older adult patients admitted to the emergency department (ED) but requiring immediate admission to the intensive care unit (ICU) remain limited. This study aimed to develop an in-hospital mortality prediction model for critically ill older adult patients transferred from the ED to the ICU. The training cohort was taken from the Medical Information Mart for Intensive Care IV (version 2.2) database, and the external validation cohort was taken from the Affiliated Dongyang Hospital of Wenzhou Medical University. In the training cohort, class balance was addressed using Random Over Sampling Examples (ROSE). Univariate and multivariate Cox regression analyses were performed to identify independent risk factors. These were then integrated into the predictive nomogram. In the validation cohort, the predictive performance of the nomogram was evaluated using the area under the curve (AUC) of the receiver operating characteristic curve, calibration curve, clinical utility decision curve analysis (DCA), and clinical impact curve (CIC). In the ROSE-balanced training cohort, univariate and multivariate Cox regression analysis identified that age, sex, Glasgow coma scale score, malignant cancer, sepsis, use of mechanical ventilation, use of vasoactive agents, white blood cells, potassium, and creatinine were independent predictors of in-hospital mortality in critically ill older adult patients, and were included in the nomogram. The nomogram showed good predictive performance in the ROSE-balanced training cohort (AUC [95% confidence interval]: 0.792 [0.783-0.801]) and validation cohort (AUC [95% confidence interval]: 0.780 [0.727-0.834]). The calibration curves were well-fitted. DCA and CIC demonstrated that the nomogram has good clinical application value. This study developed a predictive model for early prediction of in-hospital mortality in critically ill older adult patients transferred from the ED to the ICU, which was validated by external data and has good predictive performance.

Prediction model for in-hospital mortality in patients at high altitudes with ARDS due to COVID-19.

The diagnosis of acute respiratory distress syndrome (ARDS) includes the ratio of pressure arterial oxygen and inspired oxygen fraction (P/F) ≤ 300, which is often adjusted in locations more than 1,000 meters above sea level (masl) due to hypobaric hypoxemia. The main objective of this study was to develop a prediction model for in-hospital mortality among patients with ARDS due to coronavirus disease 2019 (COVID-19) (C-ARDS) at 2,600 masl with easily available variables at patient admission and to compare its discrimination capacity with a second model using the P/F adjusted for this high altitude. This study was an analysis of data from patients with C-ARDS treated between March 2020 and July 2021 in a university hospital located in the city of Bogotá, Colombia, at 2,600 masl. Demographic and laboratory data were extracted from electronic records. For the prediction model, univariate analyses were performed to screen variables with p <0.25. Then, these variables were automatically selected with a backward stepwise approach with a significance level of 0.1. The interaction terms and fractional polynomials were also examined in the final model. Multiple imputation procedures and bootstraps were used to obtain the coefficients with the best external validation. In addition, total adjustment of the model and logistic regression diagnostics were performed. The same methodology was used to develop a second model with the P/F adjusted for altitude. Finally, the areas under the curve (AUCs) of the receiver operating characteristic (ROC) curves of the two models were compared. A total of 2,210 subjects were included in the final analysis. The final model included 11 variables without interaction terms or nonlinear functions. The coefficients are presented excluding influential observations. The final equation for the model fit was g(x) = age(0.04819)+weight(0.00653)+height(-0.01856)+haemoglobin(-0.0916)+platelet count(-0.003614)+ creatinine(0.0958)+lactate dehydrogenase(0.001589)+sodium(-0.02298)+potassium(0.1574)+systolic pressure(-0.00308)+if moderate ARDS(0.628)+if severe ARDS(1.379), and the probability of in-hospital death was p (x) = e g (x)/(1+ e g (x)). The AUC of the ROC curve was 0.7601 (95% confidence interval (CI) 0.74-0, 78). The second model with the adjusted P/F presented an AUC of 0.754 (95% CI 0.73-0.77). No statistically significant difference was found between the AUC curves (p value = 0.6795). This study presents a prediction model for patients with C-ARDS at 2,600 masl with easily available admission variables for early stratification of in-hospital mortality risk. Adjusting the P/F for 2,600 masl did not improve the predictive capacity of the model. We do not recommend adjusting the P/F for altitude.

Development and validation of a prediction model for in-hospital death in patients with heart failure and atrial fibrillation

BackgroundTo develop a prediction model for in-hospital mortality of patients with heart failure (HF) and atrial fibrillation (AF).MethodsThis cohort study extracted the data of 10,236 patients with HF and AF upon intensive care unit (ICU) from the Medical Information Mart for Intensive Care (MIMIC). The subjects from MIMIC-IV were divided into the training set to construct the prediction model, and the testing set to verify the performance of the model. The samples from MIMIC-III database and eICU-CRD were included as the internal and external validation set to further validate the predictive value of the model, respectively. Univariate and multivariable Logistic regression analyses were used to explore predictors for in-hospital death in patients with HF and AF. The receiver operator characteristic (ROC), calibration curves and the decision curve analysis (DCA) curves were plotted to evaluate the predictive values of the model.ResultsThe mean survival time of participants from MIMIC-III was 11.29 ± 10.05 days and the mean survival time of participants from MIMIC-IV was 10.56 ± 9.19 days. Simplified acute physiology score (SAPSII), red blood cell distribution width (RDW), beta-blocker, race, respiratory rate, urine output, coronary artery bypass grafting (CABG), Charlson comorbidity index, renal replacement therapies (RRT), antiarrhythmic, age, and anticoagulation were predictors finally included in the prediction model. The AUC of our prediction model was 0.810 (95%CI: 0.791–0.828) in the training set, 0.757 (95%CI: 0.729–0.786) in the testing set, 0.792 (95%CI: 0.774–0.810) in the internal validation set, and 0.724 (95%CI: 0.687–0.762) in the external validation set. The calibration curves of revealed that the predictive probabilities of our model for the in-hospital death in patients with HF and AF deviated slightly from the ideal model. The DCA curves revealed that the use of our prediction model increased the net benefit than use no model.ConclusionThe prediction model had good discriminative ability, and might provide a tool to timely identify patients with HF complicated with AF who were at high risk of in-hospital mortality.

Validation study of EuroSCORE II for dialysis patients: A single-center study in Japan

BackgroundThe European System for Cardiac Operative Risk Evaluation (EuroSCORE) II is a predictive model for in-hospital mortality after cardiac surgery. Although it has good performance among the general population undergoing cardiac surgery, it has not been validated among dialysis patients, who have a higher rate of mortality after cardiac surgery. This study aimed to evaluate the performance of the model in predicting in-hospital mortality in maintenance dialysis patients undergoing cardiac surgery. MethodsThis retrospective, single-center study included adult patients on maintenance dialysis who underwent open cardiac surgery at our institution. Calibration performance of EuroSCORE II for in-hospital death was determined based on the comparison between expected and observed mortalities for low- (EuroSCORE II <4 %), intermediate- (4–8 %), and high-risk (>8 %) groups. The area under receiver operating characteristic curve (AUROC) was investigated to determine the model's discrimination performance. ResultsA total of 163 patients (male, 73.6 %; median age, 70 years; median dialysis vintage, 9 years; median EuroSCORE II, 3.3 %) were included. The mortality rate was 9.2 %. The observed mortality rates (vs. mean expected mortality) rates were 2.1 % (vs. 2.4 %), 7.5 % (vs. 5.5 %), and 34.5 % (vs. 21.1 %) in the low-, intermediate-, and high-risk groups, respectively. Its AUROC was 0.825 (95 % confidence interval, 0.711–0.940). ConclusionsAlthough EuroSCORE II model adequately estimated in-hospital mortality in the low-and intermediate-risk groups (EuroSCORE II <8 %), it underestimated in-hospital mortality in the high-risk group (EuroSCORE II >8 %) among maintenance dialysis patients. The discrimination performance of the model for in-hospital death was good among maintenance dialysis patients.

Development and Validation of an XGBoost-Algorithm-Powered Survival Model for Predicting In-Hospital Mortality Based on 545,388 Isolated Severe Traumatic Brain Injury Patients from the TQIP Database.

Traumatic brain injury (TBI) represents a significant global health issue; the traditional tools such as the Glasgow Coma Scale (GCS) and Abbreviated Injury Scale (AIS) which have been used for injury severity grading, struggle to capture outcomes after TBI. This paper aims to implement extreme gradient boosting (XGBoost), a powerful machine learning algorithm that combines the predictions of multiple weak models to create a strong predictive model with high accuracy and efficiency, in order to develop and validate a predictive model for in-hospital mortality in patients with isolated severe traumatic brain injury and to identify the most influential predictors. In total, 545,388 patients from the 2013-2021 American College of Surgeons Trauma Quality Improvement Program (TQIP) database were included in the current study, with 80% of the patients used for model training and 20% of the patients for the final model test. The primary outcome of the study was in-hospital mortality. Predictors were patients' demographics, admission status, as well as comorbidities, and clinical characteristics. Penalized Cox regression models were used to investigate the associations between the survival outcomes and the predictors and select the best predictors. An extreme gradient boosting (XGBoost)-powered Cox regression model was then used to predict the survival outcome. The performance of the models was evaluated using the Harrell's concordance index (C-index). The time-dependent area under the receiver operating characteristic curve (AUC) was used to evaluate the dynamic cumulative performance of the models. The importance of the predictors in the final prediction model was evaluated using the Shapley additive explanations (SHAP) value. On average, the final XGBoost-powered Cox regression model performed at an acceptable level for patients with a length of stay up to 250 days (mean time-dependent AUC = 0.713) in the test dataset. However, for patients with a length of stay between 20 and 213 days, the performance of the model was relatively poor (time-dependent AUC < 0.7). When limited to patients with a length of stay ≤20 days, which accounts for 95.4% of all the patients, the model achieved an excellent performance (mean time-dependent AUC = 0.813). When further limited to patients with a length of stay ≤5 days, which accounts for two-thirds of all the patients, the model achieved an outstanding performance (mean time-dependent AUC = 0.917). The XGBoost-powered Cox regression model can achieve an outstanding predictive ability for in-hospital mortality during the first 5 days, primarily based on the severity of the injury, the GCS on admission, and the patient's age. These variables continue to demonstrate an excellent predictive ability up to 20 days after admission, a period of care that accounts for over 95% of severe TBI patients. Past 20 days of care, other factors appear to be the primary drivers of in-hospital mortality, indicating a potential window of opportunity for improving outcomes.

A Prediction Model for In-Hospital Mortality of Acute Exacerbations of Chronic Obstructive Pulmonary Disease Patients Based on Red Cell Distribution Width-to-Platelet Ratio.

To explore the association between red cell distribution width (RDW)-to-platelet ratio (RPR) and in-hospital mortality of acute exacerbations of chronic obstructive pulmonary disease (AECOPD) patients and establish a prediction model based on RPR and other predictors. This cohort study included 1922 AECOPD patients aged ≥18 years in the Medical Information Mart for Intensive Care (MIMIC)-III and MIMIC-IV as well as 1738 AECOPD patients from eICU Collaborative Research Database (eICU-CRD). Possible confounding factors were screened out by univariate logistic regression, and multivariable logistic regression was applied to evaluate the association between RPR and in-hospital mortality of AECOPD patients. The area under the curve (AUC), calibration curve and decision curve analysis (DCA) curve were plotted to evaluate the predictive value of the model. The median follow-up time was 3.14 (1.87, 6.25) day. At the end of follow-up, there were 1660 patients survived and 262 subjects died. After adjusting for confounders, we found that Log (RPR×1000) was linked with elevated risk of in-hospital mortality of AECOPD patients [odds ratio (OR)=1.36, 95% confidence interval (CI): 1.01-1.84]. The prediction model was constructed using predictors including Log (RPR×1000), age, malignant cancer, atrial fibrillation, ventilation, renal failure, diastolic blood pressure (DBP), temperature, Glasgow Coma Scale (GCS) score, white blood cell (WBC), creatinine, blood urea nitrogen (BUN), hemoglobin, infectious diseases and anion gap. The AUC of the prediction model was 0.785 (95% CI: 0.751-0.820) in the training set, 0.721 (95% CI: 0.662-0.780) in the testing set, and 0.795 (95% CI: 0.762-0.827) in the validation set. RPR was associated with the in-hospital mortality of AECOPD patients. The prediction model for the in-hospital mortality of AECOPD patients based on RPR and other predictors presented good predictive performance, which might help the clinicians to quickly identify AECOPD patients at high risk of in-hospital mortality.

In-hospital mortality prediction model of heart failure patients using imbalanced registry data: A machine learning approach

In-hospital mortality prediction model of heart failure patients using imbalanced registry data: A machine learning approach

Construction of a predictive model for in-hospital mortality of sepsis patients in intensive care unit based on machine learning

To analyze the risk factors of in-hospital death in patients with sepsis in the intensive care unit (ICU) based on machine learning, and to construct a predictive model, and to explore the predictive value of the predictive model. The clinical data of patients with sepsis who were hospitalized in the ICU of the Affiliated Hospital of Jining Medical University from April 2015 to April 2021 were retrospectively analyzed,including demographic information, vital signs, complications, laboratory examination indicators, diagnosis, treatment, etc. Patients were divided into death group and survival group according to whether in-hospital death occurred. The cases in the dataset (70%) were randomly selected as the training set for building the model, and the remaining 30% of the cases were used as the validation set. Based on seven machine learning models including logistic regression (LR), K-nearest neighbor (KNN), support vector machine (SVM), decision tree (DT), random forest (RF), extreme gradient boosting (XGBoost) and artificial neural network (ANN), a prediction model for in-hospital mortality of sepsis patients was constructed. The receiver operator characteristic curve (ROC curve), calibration curve and decision curve analysis (DCA) were used to evaluate the predictive performance of the seven models from the aspects of identification, calibration and clinical application, respectively. In addition, the predictive model based on machine learning was compared with the sequential organ failure assessment (SOFA) and acute physiology and chronic health evaluation II (APACHE II) models. A total of 741 patients with sepsis were included, of which 390 were discharged after improvement, 351 died in hospital, and the in-hospital mortality was 47.4%. There were significant differences in gender, age, APACHE II score, SOFA score, Glasgow coma score (GCS), heart rate, oxygen index (PaO2/FiO2), mechanical ventilation ratio, mechanical ventilation time, proportion of norepinephrine (NE) used, maximum NE, lactic acid (Lac), activated partial thromboplastin time (APTT), albumin (ALB), serum creatinine (SCr), blood urea nitrogen (BUN), blood uric acid (BUA), pH value, base excess (BE), and K+ between the death group and the survival group. ROC curve analysis showed that the area under the curve (AUC) of RF, XGBoost, LR, ANN, DT, SVM, KNN models, SOFA score, and APACHE II score for predicting in-hospital mortality of sepsis patients were 0.871, 0.846, 0.751, 0.747, 0.677, 0.657, 0.555, 0.749 and 0.760, respectively. Among all the models, the RF model had the highest precision (0.750), accuracy (0.785), recall (0.773), and F1 score (0.761), and best discrimination. The calibration curve showed that the RF model performed best among the seven machine learning models. DCA curve showed that the RF model exhibited greater net benefit as well as threshold probability compared to other models, indicating that the RF model was the best model with good clinical utility. The machine learning model can be used as a reliable tool for predicting in-hospital mortality in sepsis patients. RF models has the best predictive performance, which is helpful for clinicians to identify high-risk patients and implement early intervention to reduce mortality.

Machine learning prediction of in-hospital mortality and external validation in patients with cardiogenic shock: the RESCUE score

Abstract Funding Acknowledgements Type of funding sources: None. Introduction Despite the introduction of mechanical circulatory support (MCS), short-term mortality rates in cardiogenic shock (CS) remain high. Prior CS risk scores have limited reliability because of selected derivation cohorts or inadequate validation. Purpose The aim of this study is to develop and validate externally the risk prediction model for in-hospital mortality in CS patients using machine learning (ML) algorithms. Methods 1247 patients of all-cause CS from the RESCUE registry (Retrospective and Prospective Observational Study to Investigate Clinical Outcomes and Efficacy of Left Ventricular Assist Device for Korean Patients with CS) between January 2014 and December 2018 were analyzed. Patients were randomly divided into two groups for training and testing at a ratio of 7 to 3. Risk prediction models were estimated with the logistic regression analysis and four ML algorithms: Least absolute shrinkage and selection operator analysis (LASSO), Support vector machine (SVM), Extreme Gradient Boosting (XGB), and Random Forest classifier. Internal validation with the RESCUE registry and external validation with the single center CS registry including 678 patients were performed. Results All four ML models made an agreement with the seven important variables associated with mortality as follows; age, vasoactive inotropic score, left ventricular ejection fraction, lactic acid, cardiac arrest at presentation, requiring continuous renal replacement therapy, and mechanical ventilator. The logistic regression analysis was performed with the risk factors which were derived by ML models. The logistic regression model named RESCUE score showed the predictive performance with an area under the curve (AUC) of 0.86 [95% CI 0.83-0.88] for in-hospital mortality. External validation showed excellent discrimination with an AUC of 0.80 [95% CI 0.76-0.84]. Decision curve analysis indicated that the nomogram conferred high clinical net benefit. Conclusion The RESCUE score using advanced machine learning techniques has excellent predictive performance for the in-hospital mortality of CS regardless of the underlying cause. This model may be useful as a tool to estimate the risk stratification of CS patients.

A prediction model for in-hospital mortality in intensive care unit patients with metastatic cancer.

To identify predictors for in-hospital mortality in patients with metastatic cancer in intensive care units (ICUs) and established a prediction model for in-hospital mortality in those patients. In this cohort study, the data of 2,462 patients with metastatic cancer in ICUs were extracted from the Medical Information Mart for Intensive Care III (MIMIC-III) database. Least absolute shrinkage and selection operator (LASSO) regression analysis was applied to identify the predictors for in-hospital mortality in metastatic cancer patients. Participants were randomly divided into the training set (n = 1,723) and the testing set (n = 739). Patients with metastatic cancer in ICUs from MIMIC-IV were used as the validation set (n = 1,726). The prediction model was constructed in the training set. The area under the curve (AUC), sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) were employed for measuring the predictive performance of the model. The predictive performance of the model was validated in the testing set and external validation was performed in the validation set. In total, 656 (26.65%) metastatic cancer patients were dead in hospital. Age, respiratory failure, the sequential organ failure assessment (SOFA) score, the Simplified Acute Physiology Score II (SAPS II) score, glucose, red cell distribution width (RDW) and lactate were predictors for the in-hospital mortality in patients with metastatic cancer in ICUs. The equation of the prediction model was ln(P/(1 + P)) = -5.9830 + 0.0174 × age + 1.3686 × respiratory failure + 0.0537 × SAPS II + 0.0312 × SOFA + 0.1278 × lactate - 0.0026 × glucose + 0.0772 × RDW. The AUCs of the prediction model was 0.797 (95% CI,0.776-0.825) in the training set, 0.778 (95% CI, 0.740-0.817) in the testing set and 0.811 (95% CI, 0.789-0.833) in the validation set. The predictive values of the model in lymphoma, myeloma, brain/spinal cord, lung, liver, peritoneum/pleura, enteroncus and other cancer populations were also assessed. The prediction model for in-hospital mortality in ICU patients with metastatic cancer exhibited good predictive ability, which might help identify patients with high risk of in-hospital death and provide timely interventions to those patients.

The prediction of in-hospital mortality in chronic kidney disease patients with coronary artery disease using machine learning models

ObjectiveChronic kidney disease (CKD) patients with coronary artery disease (CAD) in the intensive care unit (ICU) have higher in-hospital mortality and poorer prognosis than patients with either single condition. The objective of this study is to develop a novel model that can predict the in-hospital mortality of that kind of patient in the ICU using machine learning methods.MethodsData of CKD patients with CAD were extracted from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Boruta algorithm was conducted for the feature selection process. Eight machine learning algorithms, such as logistic regression (LR), random forest (RF), Decision Tree, K-nearest neighbors (KNN), Gradient Boosting Decision Tree Machine (GBDT), Support Vector Machine (SVM), Neural Network (NN), and Extreme Gradient Boosting (XGBoost), were conducted to construct the predictive model for in-hospital mortality and performance was evaluated by average precision (AP) and area under the receiver operating characteristic curve (AUC). Shapley Additive Explanations (SHAP) algorithm was applied to explain the model visually. Moreover, data from the Telehealth Intensive Care Unit Collaborative Research Database (eICU-CRD) were acquired as an external validation set.Results3590 and 1657 CKD patients with CAD were acquired from MIMIC-IV and eICU-CRD databases, respectively. A total of 78 variables were selected for the machine learning model development process. Comparatively, GBDT had the highest predictive performance according to the results of AUC (0.946) and AP (0.778). The SHAP method reveals the top 20 factors based on the importance ranking. In addition, GBDT had good predictive value and a certain degree of clinical value in the external validation according to the AUC (0.865), AP (0.672), decision curve analysis, and calibration curve.ConclusionMachine learning algorithms, especially GBDT, can be reliable tools for accurately predicting the in-hospital mortality risk for CKD patients with CAD in the ICU. This contributed to providing optimal resource allocation and reducing in-hospital mortality by tailoring precise management and implementation of early interventions.

Circadian rhythms of vital signs are associated with in-hospital mortality in critically ill patients: A retrospective observational study

ABSTRACT Vital signs have been widely used to assess the disease severity of patients, but there is still a lack of research on their circadian rhythms. The objective is to explore the circadian rhythms of vital signs in critically ill patients and establish an in-hospital mortality prediction model. Study patients from the recorded eICU Collaborative Research Database were included in the present analyses. The circadian rhythms of vital signs are analyzed in critically ill patients using the cosinor method. Logistic regression was used to screen independent predictors and establish a prediction model for in-hospital mortality by multivariate logistic regression analysis and to show in the nomogram. Internal validation is used to evaluate the prediction model by bootstrapping with 1000 resamples. A total of 29,448 patients were included in the current analyses. The Mesor, Amplitude, and Peak time of vital signs, such as heart rate (HR), temperature, respiration rate (RR), pulse oximetry-derived oxygen saturation (SpO2), and blood pressure (BP), were significant differences between survivors and non-survivors . Logistic regression analysis showed that Mesor, Amplitude, and Peak time of HR, RR, and SpO2 were independent predictors for in-hospital mortality in critically ill patients. The area under the curve (AUC) and c-index of the prediction model for the Medical intensive care unit (MICU) and Surgical intensive care unit (SICU) were 0.807 and 0.801, respectively. The Hosmer-Lemeshow test P-values were 0.076 and 0.085, respectively, demonstrating a good fit for the prediction model. The calibration curve and decision curve analysis (DCA) also demonstrated its accuracy and applicability. Internal validation assesses the consistency of the results. There were significant differences in the circadian rhythms of vital signs between survivors and non-survivors in critically ill patients. The prediction model established by the Mesor, Amplitude, and Peak time of HR, RR, and SpO2 combined with the Acute Physiology and Chronic Health Evaluation (APACHE) IV score has good predictive performance for in-hospital mortality and may eventually support clinical decision-making. Abbreviations: ICU: Intensive care unit; MICU: Medical intensive care unit; SICU: Surgical intensive care unit; HR: Heart rate; RR: Respiration rate; SpO2: Pulse oximetry-derived oxygen saturation; BP: Blood pressure; SBP: Systolic blood pressure; DBP: Diastolic blood pressure; APACHE: Acute Physiology and Chronic Health Evaluation; bpm: beats per min; BMI: Body mass index; OR: Odd ratio; CI: Confidential interval; IQR: Interquartile range; SD: Standard deviation; ROC: Receiver operating characteristic; AUC: area under the curve; DCA: Decision curve analysis; IRB: Institutional review board.

Development and Validation of a Risk Prediction Model for In-Hospital Mortality in Patients With Acute Upper Gastrointestinal Bleeding.

Acute upper gastrointestinal bleeding (UGIB) is a common life-threatening clinical emergency with a poor prognosis. The aim of this study was to develop a risk prediction model for in-hospital mortality in patients with UGIB. We performed a post hoc analysis of a publicly available retrospective clinical data. A total of 360 patients with UGIB were included in this study. The least absolute shrinkage and selection operator regression was used to screen predictors and a restricted cubic spline function was used to investigate the assumption of linear relationships between continuous predictors and the risk of in-hospital mortality. Backward stepwise selection with the Akaike information criterion was used to identify variables for the best prediction model. A nomogram was developed based on the results of the best prediction model. The receiver operating characteristic curve, GiViTI calibration plot, and decision curve analysis were used to evaluate the performance of the nomogram. The optimal prediction model consisting of 4 predictors: red cell distribution width (odds ratio [OR] = 8.44; 95% confidence interval [CI]: 1.77-89.10), platelet count (OR = 0.99; 95% CI: 0.99-1.00), pulse rate (OR = 1.03; 95% CI: 1.01-1.05), and SpO2 (OR = 0.92; 95% CI: 0.86-0.96). The nomogram model had good discrimination (area under the curve = 0.86, 95% CI: 0.78-0.95), calibration, and clinical usefulness. In this study, we developed a nomogram model for predicting death during hospitalization in patients with UGIB based on blood biomarkers and baseline vital signs at the time of admission. The model has good performance, allowing rapid risk stratification of patients with UGIB.

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.

PREDICTION MODEL FOR IN-HOSPITAL MORTALITY OF PATIENTS WITH HEART FAILURE BASED ON OPTUNA AND LIGHT GRADIENT BOOSTING MACHINE

Background and objective: Heart failure (HF) is a lethal public health problem in the field of cardiovascular diseases with high incidence, rehospitalization, and mortality rates. Therefore, the prediction of in-hospital mortality of patients with HF is of paramount significance in providing clinical information to doctors. To improve the accuracy of prediction, this study constructed a prediction model for in-hospital mortality of patients with HF based on machine learning algorithms. Methods: We obtained the medical data of 1901 patients with HF from a public database and performed preprocessing to extract 19 variables as inputs of the prediction model. A prediction model was constructed based on the light gradient boosting machine (LightGBM) and its performance was improved using the Optuna framework to optimize the LightGBM hyperparameters. To evaluate the proposed algorithm, five machine learning algorithms widely used in the field of biomedicine were selected for comparison: support vector machine, classification and regression tree, random forests, gradient boosting decision tree, and LightGBM. Further, we explained the proposed model based on Deep SHapley Additive exPlanations. We also quantified the importance of each variable and analyzed its correlation with the results. Results: The accuracy rate of Optuna–LightGBM was [Formula: see text], the precision rate was [Formula: see text], the recall rate was [Formula: see text], the [Formula: see text]-measure was [Formula: see text], and the area under the receiver operating characteristic curve was [Formula: see text]. The results show that this model outperformed other models on all evaluation indicators. Conclusions: The proposed method can be used to construct a prediction model for in-hospital mortality of patients with HF. Optuna–LightGBM can assist clinicians to quickly classify the high-risk patients with HF so that the clinicians can provide timely care and optimize hospital resources.