Sepsis Prognosis Data with RDW and ePVS Biomarkers for Deep Learning

A dataset of 73 sepsis patient records from an emergency department, used to construct a multilayer perceptron model for predicting 28-day mortality. The model was validated on independent MIMIC-III and MIMIC-IV datasets, achieving an accuracy of up to 92.0% and an AUC of 0.812. The data was authored by Qi Yun Gan and last updated in April 2026.

Use Cases

• Training deep learning models for sepsis mortality prediction based on biomarker data.
• Benchmarking prognostic models against reported accuracy, sensitivity, and specificity metrics.
• Investigating the relationship between red cell distribution width (RDW) and estimated plasma volume status (ePVS) with patient outcomes.
• Validating clinical AI models using the referenced MIMIC-III and MIMIC-IV datasets.

Strengths

• Model performance metrics are provided, including an accuracy of 92.0% and an AUC of 0.812.
• Data includes two specific biomarkers (RDW and ePVS) identified as significant predictors.
• The model underwent internal and external validation using large, independent MIMIC datasets.

Limitations

• The dataset is very small, containing only 73 patient records.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

figshare

Collection Method

Data from 73 sepsis patients in an emergency department, with validation on MIMIC-III and MIMIC-IV datasets.

Time Range

The study period is not specified.

Freshness

Last updated 2026-04-16 05:16:35; freshness should be verified.

Geography

Geographic coverage is not specified.