Background: Early and accurate identification of sepsis patients with high risk of in-hospital death can help physicians in intensive care units (ICUs) make optimal clinical decisions. This study aimed to develop machine learning-based tools to predict the risk of hospital death of patients with sepsis in ICUs. Methods: The source database used for model development and validation is the medical information mart for intensive care (MIMIC) III. We identified adult sepsis patients using the new sepsis definition Sepsis-3. A total of 86 predictor variables consisting of demographics, laboratory tests and comorbidities were used. We employed the least absolute shrinkage and selection operator (LASSO), random forest (RF), gradient boosting machine (GBM) and the traditional logistic regression (LR) method to develop prediction models. In addition, the prediction performance of the four developed models was evaluated and compared with that of an existent scoring tool – simplified acute physiology score (SAPS) II – using five different performance measures: the area under the receiver operating characteristic curve (AUROC), Brier score, sensitivity, specificity and calibration plot. Results: The records of 16,688 sepsis patients in MIMIC III were used for model training and test. Amongst them, 2949 (17.7%) patients had in-hospital death. The average AUROCs of the LASSO, RF, GBM, LR and SAPS II models were 0.829, 0.829, 0.845, 0.833 and 0.77, respectively. The Brier scores of the LASSO, RF, GBM, LR and SAPS II models were 0.108, 0.109, 0.104, 0.107 and 0.146, respectively. The calibration plots showed that the GBM, LASSO and LR models had good calibration; the RF model underestimated high-risk patients; and SAPS II had the poorest calibration. Conclusion: The machine learning-based models developed in this study had good prediction performance. Amongst them, the GBM model showed the best performance in predicting the risk of in-hospital death. It has the potential to assist physicians in the ICU to perform appropriate clinical interventions for critically ill sepsis patients and thus may help improve the prognoses of sepsis patients in the ICU.

We propose to employ evidential reasoning (ER) rule to construct a clinical decision support system (CDSS) to aid physicians to predict the probability of intensive care unit (ICU) admission and in-hospital death for trauma patients once they arrive at a hospital. A generalized Bayesian rule is used to mine evidence from historical data. Evidence is profiled using a format of belief distribution, where the belief degrees of different trauma outcomes are assigned with derived probabilities linked to the corresponding outcomes. Inputs to the CDSS are clinical data of a patient, and output from the system is predicted belief degree of severe trauma, including ICU admission and in-hospital death. The inner logic of the CDSS is that pieces of evidence that match the clinical data of a patient are identified from the evidence base first, and then the ER rule-based evidence aggregation mechanism is utilized to combine the matched evidences to arrive at a prediction. The reliability, weight, and interdependence of clinical evidence are taken into account. Moreover, an evidence weight training module is constructed. The ER rule-based prediction model has superior performance compared with logistic regression and artificial neural network models. An innovative and pragmatic ER rule-based CDSS for trauma outcome prediction is contributed by this article. In the era of big data, this CDSS helps predict patient outcomes based on historical data and helps physicians in emergency departments make proper trauma management decisions.

This study aimed to construct machine learning (ML) models for predicting prolonged length of stay (pLOS) in intensive care units (ICU) among general ICU patients. A multicenter database called eICU (Collaborative Research Database) was used for model derivation and internal validation, and the Medical Information Mart for Intensive Care (MIMIC) III database was used for external validation. We used four different ML methods (random forest, support vector machine, deep learning, and gradient boosting decision tree (GBDT)) to develop prediction models. The prediction performance of the four models were compared with the customized simplified acute physiology score (SAPS) II. The area under the receiver operation characteristic curve (AUROC), area under the precision-recall curve (AUPRC), estimated calibration index (ECI), and Brier score were used to measure performance. In internal validation, the GBDT model achieved the best overall performance (Brier score, 0.164), discrimination (AUROC, 0.742; AUPRC, 0.537), and calibration (ECI, 8.224). In external validation, the GBDT model also achieved the best overall performance (Brier score, 0.166), discrimination (AUROC, 0.747; AUPRC, 0.536), and calibration (ECI, 8.294). External validation showed that the calibration curve of the GBDT model was an optimal fit, and four ML models outperformed the customized SAPS II model. The GBDT-based pLOS-ICU prediction model had the best prediction performance among the five models on both internal and external datasets. Furthermore, it has the potential to assist ICU physicians to identify patients with pLOS-ICU risk and provide appropriate clinical interventions to improve patient outcomes.

Objectives: Acute kidney injury (AKI) is common in intensive care unit (ICU) patients and is associated with high mortality, prolonged ICU stays, and increased costs. Early prediction is crucial for timely intervention and improved outcomes. Various prediction models, including machine learning, deep learning, and dynamic prediction frameworks, have been developed, but their modeling approaches, data utilization, and clinical applicability require further investigation. This review comprehensively assesses the modeling methods, data utilization strategies, and clinical applicability of AKI prediction models in the ICU, identifies current challenges, and proposes future research directions. Materials and Methods: A systematic search was conducted in PubMed, Embase, Scopus, Web of Science, IEEE Xplore, and ACM Digital Library up to December 12, 2024. Studies were included if they reported AKI prediction models using ICU-specific data, included at least 2 predictors, and evaluated model performance. Extracted data included study characteristics, model details, data sources, performance metrics, and validation methods. The risk of bias was assessed using PROBAST (Prediction Model Risk of Bias Assessment Tool), and the reporting quality was evaluated using the TRIPOD (Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis) guideline. Results: From 1305 screened studies, 47 met the inclusion criteria. Models ranged from machine learning to advanced deep learning techniques. Only 14 studies conducted external validation. Most studies (n = 44) had a high risk of bias, particularly in generalizability and clinical applicability. Discussion: Although AI models have shown promise in predicting AKI in ICU settings, key challenges remain. These include limited external validation, lack of dynamic modeling, insufficient interpretability, and poor consideration of clinical integration. Different study designs, prediction windows, and data sources also hinder model comparability. Conclusions: Future research should prioritize dynamic, interpretable, and externally validated models. These efforts are critical to bridge the gap between model development and clinical implementation and to enhance the real-world applicability of AI in AKI prediction.

Background: The relationship between kidney function and hearing loss has long been recognized, but evidence of this association mostly comes from small observational studies or other populations. The aim of this study is to explore the association between reduced kidney function and hearing loss in a large population-based study among the middle-aged and elderly Chinese. Methods: Data collected from the Chinese Health and Retirement Longitudinal Study (CHARLS) in 2015 were used for analysis. A cross-sectional study was conducted among 12,508 participants aged 45 years and older. Hearing loss, the outcome of this study, was defined according to interviewees’ responses to three survey questions related to hearing in the CHARLS. Estimated glomerular filtration rate (eGFR) was employed to assess kidney function, and participants were classified into three categories based on eGFR: ≥90, 60–89 and < 60 mL/min/1.73 m². Multivariable logistic regression was employed to adjust for potential confounders, including demographics, health-related behaviors, and cardiovascular risk factors. Results: The overall prevalence of self-reported hearing loss in the study population was 23.6%. Compared with participants with eGFR ≥90 mL/min/1.73 m², participants with eGFR of 60–89 mL/min/1.73 m² (odds ratio [OR]: 1.11, 95% confidence interval [CI]: 1.00–1.23) and eGFR < 60 mL/min/1.73 m² (OR: 1.25, 95% CI: 1.04–1.49) showed increased risk of hearing loss after adjusting for potential confounders. Conclusions: Reduced kidney function is independently associated with hearing loss. Testing for hearing should be included in the integrated management of patients with chronic kidney disease.