A machine learning model for predicting acute respiratory distress syndrome risk in patients with sepsis using circulating immune cell parameters: a retrospective study

BMC Infectious Diseases, Journal Year: 2025, Volume and Issue: 25(1)

Published: April 21, 2025

Acute respiratory distress syndrome (ARDS) is a severe complication associated with high mortality rate in patients sepsis. Early identification of sepsis at risk developing ARDS crucial for timely intervention, optimization treatment strategies, and improvement clinical outcomes. However, traditional prediction methods are often insufficient. This study aimed to develop machine learning (ML) model predict the using circulating immune cell parameters other physiological data. Clinical data from 10,559 were obtained MIMIC-IV database. Principal component analysis (PCA) was used dimensionality reduction comprehensively evaluate models' predictive capabilities, we several ML algorithms, including decision trees, k-nearest neighbors (KNN), logistic regression, naive Bayes, random forests, neural networks, XGBoost, support vector machines (SVM) risk. The performance assessed area under receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, F1 score. Shapley additive explanations (SHAP) interpret contribution individual features predictions. Among all models, XGBoost showed best an AUC 0.764. Feature importance revealed that mean arterial pressure, monocyte count, neutrophil pH, platelet count key predictors SHAP provided further information on how these contributed model's predictions, aiding interpretability potential applications. accurately predicted could be useful tool early high-risk intervention; however, validation integration into practice required.

Language: Английский

A Machine Learning Approach for Predicting Maternal Health Risks in Lower-Middle-Income Countries Using Sparse Data and Vital Signs

Future Internet, Journal Year: 2025, Volume and Issue: 17(5), P. 190 - 190

Published: April 22, 2025

Background/Objectives: According to the World Health Organization, maternal mortality rates remain a critical public health issue, with 94% of deaths occurring in low- and middle-income countries (LMICs), where reached 430 per 100,000 live births 2020 compared 13 high-income countries. Despite this difference, only few studies have investigated whether sparse data features such as vital signs can effectively predict risks. This study addresses gap by evaluating predictive capability sign using machine learning models trained on dataset 1014 pregnant women from rural Bangladesh. Methods: developed multiple containing age, blood pressure, temperature, heart rate, glucose The models’ performance were evaluated regular, random stratified sampling techniques. Additionally, we stacking ensemble model combining methods evaluate accuracy. Results: A key contribution is developing combined sampling, an approach not previously considered risk prediction. achieved highest accuracy (87.2%), outperforming CatBoost (84.7%), XGBoost (84.2%), forest (81.3%) decision trees (80.3%) without sampling. Conclusions: Observations our demonstrate feasibility for prediction algorithms. By focusing resource-constrained settings, show that offers convenient accessible solution improve prenatal care reduce LMICs.

Language: Английский