Machine Learning-Based Prediction of Malnutrition in Surgical In-Patients: A Validation Pilot Study
Diether Kramer1,2, Stefanie Jauk1,2, Sai Veeranki1,2
1Steiermärkische Krankenanstaltengesellschaft m.b.H. (KAGes), Graz, Austria.
Studies in Health Technology and Informatics
|April 29, 2024
Summary
Automated malnutrition screening using machine learning (ML) shows promise for hospitalised patients. This approach could reduce workload and improve patient outcomes by replacing traditional manual assessments.
Area of Science:
- Medical Informatics
- Clinical Nutrition
- Machine Learning Applications
Background:
- Malnutrition in hospitalised patients is linked to severe complications, poorer outcomes, and extended stays.
- Current screening relies on scores requiring manual assessment, increasing healthcare professionals' workload.
Purpose of the Study:
- To validate a machine learning (ML)-based approach for automated malnutrition prediction in hospitalised individuals.
- To assess the efficacy of ML in identifying malnutrition compared to traditional methods.
Main Methods:
- A prospective study involving 159 surgical in-patients.
- Dietitian assessments of malnutrition were compared against ML-based predictions made on admission evening.
Main Results:
- The ML model demonstrated a prospective validation accuracy of 83.0%.
- The model achieved an Area Under the Receiver Operating Characteristic curve (AUROC) of 0.833.
Conclusions:
- Automated malnutrition screening using ML is a viable alternative to manual tools.
- This pilot study suggests potential for ML to streamline hospital malnutrition detection.
