Comparison between traditional logistic regression and machine learning for predicting mortality in adult sepsis patients

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 11

Published: Jan. 6, 2025

Sepsis is a life-threatening disease associated with high mortality rate, emphasizing the need for exploration of novel models to predict prognosis this patient population. This study compared performance traditional logistic regression and machine learning in predicting adult sepsis mortality. To develop an optimum model patients based on comparing methodology. Retrospective analysis was conducted 606 inpatients at our medical center between January 2020 December 2022, who were randomly divided into training validation sets 7:3 ratio. Traditional methods employed assess predictive ability sepsis. Univariate identified independent risk factors model, while Least Absolute Shrinkage Selection Operator (LASSO) facilitated variable shrinkage selection model. Among various models, which included Bagged Tree, Boost Decision LightGBM, Naïve Bayes, Nearest Neighbors, Support Vector Machine (SVM), Random Forest (RF), one maximum area under curve (AUC) chosen construction. Model comparison Sequential Organ Failure Assessment (SOFA) Acute Physiology Chronic Health Evaluation (APACHE) scores performed using receiver operating characteristic (ROC) curves, calibration decision (DCA) curves set. 17 variables, namely gender, history coronary heart (CHD), systolic pressure, white blood cell (WBC), neutrophil count (NEUT), lymphocyte (LYMP), lactic acid, neutrophil-to-lymphocyte ratio (NLR), red distribution width (RDW), interleukin-6 (IL-6), prothrombin time (PT), international normalized (INR), fibrinogen (FBI), D-dimer, aspartate aminotransferase (AST), total bilirubin (Tbil), lung infection. Significant differences (p < 0.05) survival non-survival groups observed these variables. Utilizing stepwise "backward" method, factors, including NLR, RDW, IL-6, PT, Tbil, identified. These then incorporated minimum Akaike Information Criterion (AIC) value (98.65). techniques also applied, RF demonstrating Area Under Curve 0.999, selected. LASSO regression, employing lambda.1SE criteria, NEUT, IL6, INR, Tbil as variables constructing validated through ten-fold cross-validation. For SOFA, APACHE scoring. Based deep principles, demonstrates advantages over prognosis. The holds significant potential clinical surveillance interventions enhance outcomes patients.

Language: Английский

---

A comprehensive survey of artificial intelligence adoption in European laboratory medicine: current utilization and prospects

Clinical Chemistry and Laboratory Medicine (CCLM), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 24, 2024

Abstract Background As the healthcare sector evolves, Artificial Intelligence’s (AI’s) potential to enhance laboratory medicine is increasingly recognized. However, adoption rates and attitudes towards AI across European laboratories have not been comprehensively analyzed. This study aims fill this gap by surveying professionals assess their current use of AI, digital infrastructure available, future implementations. Methods We conducted a methodical survey during October 2023, distributed via EFLM mailing lists. The explored six key areas: general characteristics, equipment, access health data, data management, advancements, personal perspectives. analyzed responses quantify integration identify barriers its adoption. Results From 426 initial responses, 195 were considered after excluding incomplete non-European entries. findings revealed limited engagement, with significant gaps in necessary training. Only 25.6 % reported ongoing projects. Major included inadequate tools, restricted comprehensive lack AI-related skills among personnel. Notably, substantial interest training was expressed, indicating demand for educational initiatives. Conclusions Despite recognized revolutionize enhancing diagnostic accuracy efficiency, face challenges. highlights critical need strategic investments programs improvements support Europe. Future efforts should focus on accessibility, upgrading technological expanding literacy professionals. In response, our working group plans develop make available online materials meet growing demand.

Language: Английский

---

Artificial intelligence in the clinical laboratory

Clinica Chimica Acta, Journal Year: 2024, Volume and Issue: 559, P. 119724 - 119724

Published: May 10, 2024

Language: Английский

---

Data flow in clinical laboratories: could metadata and peridata bridge the gap to new AI-based applications?

Clinical Chemistry and Laboratory Medicine (CCLM), Journal Year: 2024, Volume and Issue: unknown

Published: Oct. 5, 2024

In the last decades, clinical laboratories have significantly advanced their technological capabilities, through use of interconnected systems and software. Laboratory Information Systems (LIS), introduced in 1970s, transformed into sophisticated information technology (IT) components that integrate with various digital tools, enhancing data retrieval exchange. However, current capabilities LIS are not sufficient to rapidly save extensive data, generated during total testing process (TTP), beyond just test results. This opinion paper discusses qualitative types TTP proposing how divide laboratory-generated two categories, namely metadata peridata. Being both peridata derived from process, it is proposed first useful describe characteristics while second for interpretation Together standardizing preanalytical coding, subdivision or might enhance ML studies, also by facilitating adherence laboratory-derived Findability, Accessibility, Interoperability, Reusability (FAIR) principles. Finally, integrating can improve usability, support utility, advance AI model development healthcare, emphasizing need standardized management practices.

Language: Английский

---

Artificial Intelligence in Sepsis Management: An Overview for Clinicians

Journal of Clinical Medicine, Journal Year: 2025, Volume and Issue: 14(1), P. 286 - 286

Published: Jan. 6, 2025

Sepsis is one of the leading causes mortality in hospital settings, and early diagnosis a crucial challenge to improve clinical outcomes. Artificial intelligence (AI) emerging as valuable resource address this challenge, with numerous investigations exploring its application predict diagnose sepsis early, well personalizing treatment. Machine learning (ML) models are able use data collected from Electronic Health Records or continuous monitoring patients at risk hours before onset symptoms. Background/Objectives: Over past few decades, ML other AI tools have been explored extensively sepsis, developed for detection, diagnosis, prognosis, even real-time management treatment strategies. Methods: This review was conducted according SPIDER (Sample, Phenomenon Interest, Design, Evaluation, Research Type) framework define study methodology. A critical overview each paper by three different reviewers, selecting those that provided original comprehensive relevant specific topic contributed significantly conceptual practical discussed, without dwelling on technical aspects used. Results: total 194 articles were found; 28 selected. Articles categorized analyzed based their focus—early prediction, improvement sepsis. The scientific literature presents mixed outcomes; while some studies demonstrate improvements rates management, others highlight challenges, such high incidence false positives lack external validation. designed clinicians healthcare professionals, aims provide an reviewing main methodologies used assess effectiveness, limitations, future potential.

Language: Английский

---

Application of Machine Learning Algorithms in Improving the Performance of Autonomous Vehicles

Scientific Journal of Technology, Journal Year: 2025, Volume and Issue: 7(2), P. 118 - 124

Published: Feb. 21, 2025

With the rapid development of intelligent transportation systems, autonomous driving technology relying on machine learning algorithms has received widespread attention. Although made significant improvements, how to utilize advanced further enhance its performance remains a core issue. This study aims analyze role in enhancing vehicles and discuss such as deep neural networks reinforcement can effectively solve key technical bottlenecks. A series innovative strategies based have been proposed address challenges currently faced by technology, insufficient sensor perception accuracy, contradiction between safety efficiency route planning, real-time constraints decision-making control. The goal these is improve perception, planning operational reliability auto drive system.

Language: Английский

---

Monocyte distribution width (MDW) as a reliable diagnostic biomarker for sepsis in patients with HIV

Emerging Microbes & Infections, Journal Year: 2025, Volume and Issue: unknown

Published: March 17, 2025

Introduction: Sepsis is a leading cause of death among patients with HIV, but early diagnosis remains challenge. This study evaluates the diagnostic performance monocyte distribution width (MDW) in detecting sepsis HIV.

Language: Английский

---

A machine learning approach for assessing acute infection by erythrocyte sedimentation rate (ESR) kinetics

Clinica Chimica Acta, Journal Year: 2025, Volume and Issue: 574, P. 120308 - 120308

Published: April 22, 2025

The erythrocyte sedimentation rate (ESR) is a traditional marker of inflammation, valued for its simplicity and low cost but limited by unsatisfactory specificity sensitivity. This study evaluated the equivalence ESR measurements obtained from three automated analyzers compared to Westergren method. Furthermore, various machine learning (ML) techniques were employed assess usefulness early kinetics in inflammatory disease classification. A total 346 blood samples control, rheumatological, oncological, sepsis/acute status groups analyzed. was measured using TEST 1 (Alifax Spa, Padua, Italy), VESMATIC 5 (Diesse Diagnostica Senese Siena, CUBE 30 TOUCH Italy) analyzers, Early (within 20 min) with assessed. ML models [Gradient Boosting Machine (GBM), Support Vector (SVM), Naïve Bayes (NB), Neural Networks (NN) logistic regression (LR)] discriminating trained validated ESR, slopes, clinical data. second validation cohort control sepsis used validate LR models. Automated methods showed good agreement Westergren's results. Multivariate analyses identified significant associations between values (measured TOUCH) age (p = 0.025), gender < 0.001), and, overall, samples' group 0.001). Sedimentation slopes differed significantly across groups, particularly 12 min, cases showing distinct patterns. achieved moderate accuracy, GBM performing best (AUC 0.80). classification an AUC 0.884, high sensitivity (96.9 %) (74.2 %). In cohort, outperformed prior results, reaching 0.991 (95 % CI: 0.973-1.000), 95.2 100 %. Current technologies measurement well agree reference method provide robust results evaluating systemic infections. novelty this lies connecting states, identifying status. Future studies larger datasets are needed these approaches guide application.

Language: Английский

---

Algorithms for predicting COVID outcome using ready-to-use laboratorial and clinical data

Frontiers in Public Health, Journal Year: 2024, Volume and Issue: 12

Published: May 14, 2024

The pandemic caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is an emerging crisis affecting the public health system. clinical features of COVID-19 can range from asymptomatic state to and multiple organ dysfunction. Although some hematological biochemical parameters are altered during moderate COVID-19, there still a lack tools combine these predict outcome patient with COVID-19. Thus, this study aimed at employing patients diagnosed in order build machine learning algorithms for predicting COVID mortality or survival. Patients included had diagnosis SARS-CoV-2 infection confirmed RT-PCR measurements were performed three different time points upon hospital admission. Among evaluated, ones that stand out most important T1 point (urea, lymphocytes, glucose, basophils age), which could be possible biomarkers severity patients. This shows urea parameter best classifies rises over time, making it crucial analyte used outcome. In optimal medically interpretable prediction presented each point. It was found paramount variable all points. However, importance other variables changes point, demonstrating dynamic approach effective patient’s prediction. All all, use defining tool laboratory monitoring prediction, may bring benefits future pandemics newly reemerging variants concern.

Language: Английский

---

Aloe-emodin plus TIENAM ameliorate cecal ligation and puncture-induced sepsis in mice by attenuating inflammation and modulating microbiota

Frontiers in Microbiology, Journal Year: 2024, Volume and Issue: 15

Published: Dec. 12, 2024

Despite the high sepsis-associated mortality, effective and specific treatments remain limited. Using conventional antibiotics as TIENAM (imipenem cilastatin sodium for injection, TIE) is challenging due to increasing bacterial resistance, diminishing their efficacy leading adverse effects. We previously found that aloe-emodin (AE) exerts therapeutic effects on sepsis by reducing systemic inflammation regulating gut microbiota. Here, we investigated whether administering AE TIE post-sepsis onset, using a cecal ligation puncture (CLP)-induced model, extends survival improves physiological functions. Survival rates, inflammatory cytokines, tissue damage, immune cell populations, ascitic fluid microbiota, key signaling pathways were assessed. Combining significantly enhanced reduced load in septic mice, indicating potent antimicrobial properties. Moreover, substantial improvements rates of + TIE-treated mice (10% 60%) within 168 h observed relative CLP group. This combination therapy also effectively modulated marker (interleukin [IL]-6, IL-1β, tumor necrosis factor [TNF]-α) levels counts decreasing those B, NK, TNFR2+ T reg cells, while CD8+ cells; alleviated damage; peritoneal cavity; suppressed NF-κB pathway. altered cavity microbiota composition post-treatment, characterized pathogenic bacteria ( Bacteroides ) abundance. Our findings underscore potential treating sepsis, encourage further research possible clinical implementations surmount limitations amplify AE.

Language: Английский

---