MIMIC-IV Study on Acute Pancreatitis Readmission and Gastrointestinal Hemorrhage

2,004 patients with acute pancreatitis were enrolled from the MIMIC-IV database, with 445 readmission records. The study by Yazhen Zhan, last updated in May 2026, identified gastrointestinal hemorrhage as the complication most strongly associated with readmission. It found that a combination of low alanine aminotransferase and high red blood cell distribution width was associated with the highest odds of gastrointestinal hemorrhage.

Use Cases

• Predicting gastrointestinal hemorrhage risk in acute pancreatitis patients based on age, ALT, and RDW.
• Analyzing the association between specific laboratory values and hospital readmission for acute pancreatitis.
• Feature selection for clinical prediction models using LASSO regression and machine learning algorithms like XGBoost.

Strengths

• Dataset size of 2,004 patient records with 445 readmission events provides a substantive cohort for analysis.
• Analysis employed multiple machine learning algorithms (XGBoost, random forest, K-nearest neighbor) for robust feature selection.
• Clear association identified between a specific biomarker pattern (low-ALT + high-RDW) and gastrointestinal hemorrhage risk.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the underlying data is unknown, which may limit suitability assessment.
• The 15.9 KB file size suggests the provided document contains summary results, not the full raw patient-level dataset.

Provenance

Source

MIMIC-IV database

Collection Method

Retrospective study using data downloaded from the MIMIC-IV clinical database.

Freshness

Last updated 2026-05-28 04:52:44