Early predicting 30-day mortality in sepsis in MIMIC-III by an artificial neural networks model

Su, Yingjie; Guo, Cuirong; Zhou, Shifang; Li, Changluo; Ding, Ning

doi:10.1186/s40001-022-00925-3

Cited by 10 publications

(8 citation statements)

References 40 publications

(38 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Further analysis revealed generally higher levels of HOTTIP in deaths than in survivors. Furthermore, elevated HOTTIP, as well as APACHE II and SOFA scores [ 33 , 34 ], predicted the appearance of death in patients with sepsis and were both independent predictors of death in the patients with sepsis. However, HOTTIP has its unique advantages over the two scores in terms of predictive value, firstly given that HOTTIP, as an ncRNA, may already be resentful in early changes in the disease and its prediction may be earlier than the two scores.…”

Section: Discussionmentioning

confidence: 99%

LncRNA HOTTIP as a diagnostic biomarker for acute respiratory distress syndrome in patients with sepsis and to predict the short-term clinical outcome: a case-control study

Shi,

Zhu,

et al. 2024

BMC Anesthesiol

View full text Add to dashboard Cite

Background The present research aims to investigate the clinical diagnostic value of LncRNA HOXA distal transcript antisense RNA (HOTTIP) in acute respiratory distress syndrome (ARDS) of sepsis and its predictive significance for mortality. Methods One hundred eighteenth patients with sepsis and 96 healthy individuals were enrolled. RT-qPCR to examine HOTTIP levels. The incidence of ARDS and death was recorded. The diagnostic significance of HOTTIP in sepsis ARDS was examined using ROC and logistic regression analysis. The correlation between HOTTIP and disease severity was evaluated using Pearson’s coefficients. Kaplan-Meier analysis and COX regression were employed to examine the predictive significance of mortality. Validation of HOTTIP target miRNA by dual-luciferase assay. Results HOTTIP was persistently up-regulated in patients with ARDS sepsis than in patients without ARDS patients (P < 0.05). HOTTIP was a risk factor for the development of ARDS, which could be diagnosed in ARDS patients from non-ARDS patients (AUC = 0.847). Both the SOFA score (r = 0.6793) and the APACHE II score (r = 0.6384) were positively correlated with the HOTTIP levels. Furthermore, serum HOTTIP was an independent predictor of short-term mortality (HR = 4.813. 95%CI: 1.471–15.750, P = 0.009) and noticeably predicted the occurrence of short-term death (log rank = 0.020). miR-574-5p, a target miRNA for HOTTIP, was reduced in patients with sepsis ARDS and negatively correlated with HOTTIP. Conclusions The presence of HOTTIP serves as a diagnostic biomarker for the occurrence of ARDS, exhibits correlation with disease severity, and provides predictive value of short-term mortality in sepsis patients. HOTTIP may be involved in ARDS progression by targeting miR-574-5p.

show abstract

Section: Discussionmentioning

confidence: 99%

LncRNA HOTTIP as a diagnostic biomarker for acute respiratory distress syndrome in patients with sepsis and to predict the short-term clinical outcome: a case-control study

Shi,

Zhu,

et al. 2024

BMC Anesthesiol

View full text Add to dashboard Cite

show abstract

“…Several methods have been concurrently developed to predict mortality for ICU patients with Sepsis [13,16]. However, to the best of our knowledge there is only one research focusing on prediction of mortality for ICU patients of Sepsis-3 using MIMIC-III database [15].…”

Section: Discussion Existing Model Compilation Summarymentioning

confidence: 99%

Prediction of 30-Day Mortality for ICU Patients with Sepsis-3

Yu,

Ashrafi,

et al. 2024

Preprint

View full text Add to dashboard Cite

Background There has emerged an increasing demand for advanced methodologies aimed at augmenting our comprehension and prognostication of illnesses. This study is distinctly centered on tackling the complexity of Sepsis, an immediate bodily reaction to infection. Our objective is to refine the early identification and mortality forecasting for patients diagnosed under the Sepsis-3 criteria, with the overarching aim of enhancing the allocation of hospital resources. Methods In this study, we introduced a Machine Learning (ML) framework aimed at predicting the 30-day mortality rate among Intensive Care Unit (ICU) patients diagnosed with Sepsis-3. Leveraging the Medical Information Mart for Intensive Care III (MIMIC-III) database, we systematically identified eligible patients using advanced big data extraction tools such as Snowflake. Additionally, we employed decision tree models to ascertain the importance of various features and conducted entropy analyses across decision nodes to refine feature selection. Collaborating with esteemed clinical experts, we curated a list of 30 relevant features. Moreover, we used the Light Gradient Boosting Machine (LightGBM) model due to its gradient boosting architecture and computational efficiency. Results The study comprised a cohort of 9118 patients diagnosed with Sepsis-3. Through our meticulous preprocessing techniques, we observed a marked enhancement in both the Area Under the Curve (AUC) and accuracy metrics. The LightGBM model yielded an impressive AUC of 0.983, with a 95% confidence interval [0.980-0.990]. Moreover, it exhibited a commendable accuracy of 0.966 and an F1-score of 0.910. Notably, LightGBM showcased a substantial 6% enhancement over our best baseline model and a significant 14% enhancement over the best existing literature. These noteworthy advancements can be attributed to several factors: (I) the incorporation of a novel and pivotal feature in our model, Hospital Length of Stay (HOSP_LOS), which has not been included in previous literature; (II) the inherent strengths of LightGBM's gradient boosting architecture, enabling robust predictions even with high-dimensional data, while maintaining computational efficiency, as evidenced by its learning curve. Conclusions The introduced preprocessing methodology not only led to a substantial reduction in the number of relevant features compared to the best existing literature, thereby alleviating computational complexities, but also enabled the identification of a crucial feature previously ignored in existing literature. Through the integration of these pivotal features and meticulous parameter tuning, our proposed model achieved remarkable predictive power, with its learning curve demonstrating its capacity for generalization to unseen data. This underscores the potential of ML as indispensable tools in the dynamic environment of the ICU. Employing our model stands to streamline resource allocation within ICUs, offering clinicians greater efficiency and tailored interventions for patients afflicted with Sepsis-3. Keywords: Sepsis-3 prediction, Machine Learning, Entropy Analysis, Gradient Boosting Machine Model

show abstract

“…By incorporating diverse information sources, the model enhances healthcare professionals' predictive abilities, fostering a more comprehensive understanding of patient outcomes. [14][15][16][17][18][19][20][21].…”

Section: Introduction：mentioning

confidence: 99%

Development and Validation of a Machine learning Model for the Prediction of Short-Term Mortality among Patients with Sepsis: A Retrospective Cohort Study

Hu,

Ming,

Zhu

et al. 2024

Preprint

View full text Add to dashboard Cite

Background: Sepsis, a severe condition arising from the body's response to infection, leading to organ damage, is a major contributor to mortality in intensive care units. This study assesses and compares the predictive performance of six machine learning algorithms in forecasting in-patient mortality among sepsis patients. Results: There were 6,455 sepsis patients (5,165 for training and 1,290 for internal validation) in Medical Information Mart for Intensive Care IV v2.0 and 411 sepsis patients in eICU Collaborative Research Database v2.0 who met the inclusion criteria. The median age of the patients was 64 years and 61 years respectively. 42 features were retained for model building. The most important features for predicting mortality were urine output, Glasgow Coma Scale score, urine output, average body temperature, age, and lactate. The random forest algorithm performed the best on both the imbalanced and balanced datasets, achieving an AUC of 78.6% and 78.5%, respectively, on the external test set. Conclusions: This study showcases the applicability of machine learning for outcome prediction in sepsis patients, employing internal and multicenter external validation datasets. The application of the SHAP method enhances model interpretability by providing insights into global explanations, individual explanations, and the interactions between features.

show abstract

Early predicting 30-day mortality in sepsis in MIMIC-III by an artificial neural networks model

Cited by 10 publications

References 40 publications

LncRNA HOTTIP as a diagnostic biomarker for acute respiratory distress syndrome in patients with sepsis and to predict the short-term clinical outcome: a case-control study

LncRNA HOTTIP as a diagnostic biomarker for acute respiratory distress syndrome in patients with sepsis and to predict the short-term clinical outcome: a case-control study

Prediction of 30-Day Mortality for ICU Patients with Sepsis-3

Development and Validation of a Machine learning Model for the Prediction of Short-Term Mortality among Patients with Sepsis: A Retrospective Cohort Study

Contact Info

Product

Resources

About