GI bleeding in patients with left ventricular assist device: endoscopic approach and prediction model using...

Area of Science:

• Cardiology
• Gastroenterology
• Medical Devices

Background:

• Gastrointestinal (GI) bleeding is a frequent complication in patients with left ventricular assist devices (LVADs).
• Optimal endoscopic management strategies and quantifiable bleeding risk remain incompletely understood.

Purpose of the Study:

• To analyze the endoscopic management of GI bleeding in LVAD patients.
• To identify clinical predictors of GI bleeding after LVAD implantation using machine learning.

Main Methods:

• Retrospective analysis of adult patients undergoing LVAD implantation (February 1996 - September 2021).
• Analysis of endoscopic procedures for GI bleeding and clinical data.
• Supervised machine learning model to predict GI bleeding risk.

Main Results:

• 132 of 557 (23.7%) LVAD patients experienced GI bleeds; gastric and small intestinal bleeds were most common.
• Video capsule endoscopy had the highest diagnostic yield (68.9%).
• Destination therapy, warfarin, and antiplatelet use were significant predictors of GI bleeding risk (total score ≥10 indicates high risk).

Conclusions:

• Push enteroscopy demonstrated comparable diagnostic yield and intervention rates to EGD and colonoscopy.
• A machine learning model identified key clinical variables for quantifying GI bleeding risk in LVAD patients.