Machine Learning Prediction of Sepsis in Paralytic Ileus Patients

579 adult paralytic ileus patient records from the MIMIC-IV database (2008–2022) and an external hospital cohort, used to develop a sepsis prediction model. The dataset includes seven clinical predictors identified through feature selection, such as pneumonia and red cell distribution width. It was created by Qingzhou Song and last updated on June 2, 2026.

Use Cases

• Training a logistic regression model for sepsis risk prediction based on seven clinical variables.
• Evaluating feature importance for sepsis using SHAP analysis on predictors like pneumonia and RDW.
• Benchmarking machine learning model performance (AUC, calibration) for clinical prediction tasks.
• Conducting external validation of a predictive model using an independent hospital cohort.

Strengths

• Includes 579 patient records with a defined outcome (29.78% developed sepsis).
• Model was validated on an independent external cohort from a cancer hospital.
• Feature selection was performed using LASSO regression and the Boruta algorithm combined with clinical knowledge.
• The final logistic regression model achieved an AUC of 0.715 in external validation.

Limitations

• Row count for the underlying data is unknown, which may limit suitability assessment.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 228.8 KB, indicating limited scope.

Provenance

Source

MIMIC-IV database (2008–2022) and Guangxi Medical University Cancer Hospital.

Collection Method

Retrospective cohort study; predictors extracted within the first 12 hours after admission.

Time Range

2008–2022

Freshness

Last updated 2026-06-02 04:21:17; freshness should be verified.

Geography

Likely United States (MIMIC-IV) and China (external validation cohort).