A Novel Machine Learning-Based Point-Score Model as a Non-Invasive Decision-Making Tool for Identifying Infected
Silvia Würstle1,2, Alexander Hapfelmeier3,4, Siranush Karapetyan4
1Department of Internal Medicine II, University Hospital rechts der Isar, School of Medicine, Technical University of Munich, 81675 Munich, Germany.
Antibiotics (Basel, Switzerland)
|November 24, 2022
Summary
This study developed a scoring system to identify patients with liver cirrhosis who do not need invasive abdominocentesis for infected ascites. The non-invasive model accurately rules out infection, improving patient care.
Area of Science:
- Hepatology
- Clinical Medicine
- Medical Informatics
Background:
- Infected ascites is a serious complication in patients with decompensated liver cirrhosis.
- Abdominocentesis is the standard diagnostic procedure but is invasive.
- Accurate non-invasive methods are needed to identify patients who do not require this procedure.
Purpose of the Study:
- To assess distinctive features of patients with infected ascites and liver cirrhosis.
- To develop a scoring system for accurate identification of patients not requiring abdominocentesis to rule out infected ascites.
- To create a non-invasive tool for clinical routine.
Main Methods:
- Machine learning analysis of 34 clinical, drug, and laboratory features from 700 episodes of decompensated liver cirrhosis patients.
- Lasso regression model to select 11 discriminatory features.
- Development of a point-score model and a simplified model with fast-to-collect features.
Main Results:
- The point-score model identified 11 key discriminatory features.
- High negative predictive values (98.1%, 97.0%, 94.6%) for ruling out infected ascites were achieved.
- A simplified model showed similar predictive values, suggesting clinical applicability.
Conclusions:
- The developed point-score model is a promising non-invasive approach to rule out infected ascites in patients with hydropic decompensated liver cirrhosis.
- The model demonstrates high negative predictive values, potentially reducing the need for abdominocentesis.
- Further external validation in a prospective study is required to confirm its utility in clinical practice.
