FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W4387249062> ?p ?o ?g. }

• W4387249062 endingPage "A1803" @default.
• W4387249062 startingPage "A1803" @default.
• W4387249062 abstract "SESSION TITLE: Sepsis: Novel Identification and Treatment SESSION TYPE: Rapid Fire Original Inv PRESENTED ON: 10/10/2023 12:00 pm - 12:45 pm PURPOSE: Sepsis is a serious medical condition caused by a dysregulated host response to infection that can lead to organ failure. It is a highly heterogeneous disease that is challenging to treat, often resulting in poor outcomes. Clustering sepsis can help classify patients and potentially improve the outcomes. This study aims to develop a clustering model that does not require the imputation and aggregation of time-series data, making it more applicable for real-world clinical settings. METHODS: The study used the MIMIC-IV dataset, extracting two datasets: the pretrain dataset and the sepsis dataset. The pretrain dataset included all intensive care unit admissions within the first 25 hours, while the sepsis dataset included a -23 to 0-hour window before the sepsis time point. The top 100 time-varying features that were most important for predicting 28-day mortality were selected using logistic regression, along with demographics. The modified transformer model based on triplet embedding — time, feature, and value — was employed to overcome sparsity and irregular time intervals by treating time series as a set of observation triplets. After obtaining the status embedding vector, the K-means clustering model was applied to cluster the sepsis patients based on immune status. Survival analysis was conducted among the clusters, with the log-rank test used to compare the survival curve of each cluster. Clinical variables were also compared among the clusters. RESULTS: We clustered a total of 25,955 sepsis patients based on their status embedding vectors, resulting in five distinct clusters using the K-means algorithm. Cluster 5 was found to have the poorest prognosis, with the highest 28-day mortality rate of 30% and exhibited worse organ-related variables and other clinical features. In contrast, Cluster 1 had the most favorable prognosis, exhibiting the lowest 28-day mortality rate among all clusters. Furthermore, we analyzed changes in clinical features over time within each cluster. We found that the trends in organ-specific features, such as temperature, neutrophil count, lymphocyte count, mean blood pressure, liver function, kidney function, and coagulation index, varied among the clusters. Survival analysis of sepsis clusters showed significant differences between the clusters, with all models being significantly different each others (p < 0.0001). CONCLUSIONS: The clustering model developed in this study for sepsis without relying on the imputation and aggregation of time-series data provides a practical and applicable tool for real-world clinical settings. The potential of clustering analysis may lead to the development of more effective treatments tailored to specific patient subgroups, ultimately improving patient outcomes. CLINICAL IMPLICATIONS: This study aimed to cluster the sepsis patients without imputing or aggregating patient time-series data, making the methodology particularly useful in the medical field where medical data exhibits irregular frequencies and sparse distributions. It is acknowledged that sepsis patients’ immune status is heterogeneous. Future studies should investigate the association between clinical sepsis clustering and immune status of the patient. Clinical feature clustering may provide valuable assistance in personalized therapy for septic patients. DISCLOSURES: No relevant relationships by Kyungsoo Chung No relevant relationships by Sujung Park No relevant relationships by JUHYE SHIN No relevant relationships by MinDong Sung" @default.
• W4387249062 created "2023-10-03" @default.
• W4387249062 creator A5052235938 @default.
• W4387249062 creator A5059262053 @default.
• W4387249062 creator A5061232210 @default.
• W4387249062 creator A5087269096 @default.
• W4387249062 date "2023-10-01" @default.
• W4387249062 modified "2023-10-03" @default.
• W4387249062 title "DEVELOPING A CLUSTERING MODEL FOR SEPSIS WITHOUT IMPUTATION AND AGGREGATION OF TIME-SERIES DATA" @default.
• W4387249062 doi "https://doi.org/10.1016/j.chest.2023.07.1243" @default.
• W4387249062 hasPublicationYear "2023" @default.
• W4387249062 type Work @default.
• W4387249062 citedByCount "0" @default.
• W4387249062 crossrefType "journal-article" @default.
• W4387249062 hasAuthorship W4387249062A5052235938 @default.
• W4387249062 hasAuthorship W4387249062A5059262053 @default.
• W4387249062 hasAuthorship W4387249062A5061232210 @default.
• W4387249062 hasAuthorship W4387249062A5087269096 @default.
• W4387249062 hasBestOaLocation W43872490621 @default.
• W4387249062 hasConcept C105795698 @default.
• W4387249062 hasConcept C119857082 @default.
• W4387249062 hasConcept C124101348 @default.
• W4387249062 hasConcept C126322002 @default.
• W4387249062 hasConcept C151956035 @default.
• W4387249062 hasConcept C154945302 @default.
• W4387249062 hasConcept C2776376669 @default.
• W4387249062 hasConcept C2778384902 @default.
• W4387249062 hasConcept C33923547 @default.
• W4387249062 hasConcept C41008148 @default.
• W4387249062 hasConcept C58041806 @default.
• W4387249062 hasConcept C71924100 @default.
• W4387249062 hasConcept C73555534 @default.
• W4387249062 hasConcept C9357733 @default.
• W4387249062 hasConceptScore W4387249062C105795698 @default.
• W4387249062 hasConceptScore W4387249062C119857082 @default.
• W4387249062 hasConceptScore W4387249062C124101348 @default.
• W4387249062 hasConceptScore W4387249062C126322002 @default.
• W4387249062 hasConceptScore W4387249062C151956035 @default.
• W4387249062 hasConceptScore W4387249062C154945302 @default.
• W4387249062 hasConceptScore W4387249062C2776376669 @default.
• W4387249062 hasConceptScore W4387249062C2778384902 @default.
• W4387249062 hasConceptScore W4387249062C33923547 @default.
• W4387249062 hasConceptScore W4387249062C41008148 @default.
• W4387249062 hasConceptScore W4387249062C58041806 @default.
• W4387249062 hasConceptScore W4387249062C71924100 @default.
• W4387249062 hasConceptScore W4387249062C73555534 @default.
• W4387249062 hasConceptScore W4387249062C9357733 @default.
• W4387249062 hasIssue "4" @default.
• W4387249062 hasLocation W43872490621 @default.
• W4387249062 hasOpenAccess W4387249062 @default.
• W4387249062 hasPrimaryLocation W43872490621 @default.
• W4387249062 hasRelatedWork W2107950767 @default.
• W4387249062 hasRelatedWork W2126631032 @default.
• W4387249062 hasRelatedWork W2167133299 @default.
• W4387249062 hasRelatedWork W3025379803 @default.
• W4387249062 hasRelatedWork W3161652179 @default.
• W4387249062 hasRelatedWork W4285727872 @default.
• W4387249062 hasRelatedWork W4319653815 @default.
• W4387249062 hasRelatedWork W4321488553 @default.
• W4387249062 hasRelatedWork W4381149614 @default.
• W4387249062 hasRelatedWork W2566086483 @default.
• W4387249062 hasVolume "164" @default.
• W4387249062 isParatext "false" @default.
• W4387249062 isRetracted "false" @default.
• W4387249062 workType "article" @default.