Artificial Intelligence-Based Hospital Malnutrition Screening: Validation of a Novel Machine Learning Model

Area of Science:

• Clinical Informatics
• Artificial Intelligence in Healthcare
• Predictive Modeling

Background:

• In-hospital malnutrition is a significant issue, leading to increased morbidity, mortality, and costs.
• Current malnutrition screening methods are often underutilized or ineffective.
• Artificial intelligence (AI) presents a potential solution for improved malnutrition risk identification.

Purpose of the Study:

• To evaluate an AI-based malnutrition screening model in a large, diverse inpatient population.
• To compare the AI model's performance against a standard clinician-delivered malnutrition screening tool.

Main Methods:

• A gradient-boosted decision tree model was developed using electronic medical record data from 106,449 patients.
• A large language model (LLM) was incorporated for feature extraction.
• The model's performance was evaluated against discharge-coded and dietitian-recorded malnutrition.

Main Results:

• The AI model achieved an area under the receiver operating curve of 0.92 on day one, increasing to 0.95 when considering maximum predicted risk.
• The model significantly outperformed a nurse-administered Malnutrition Screening Tool (MST).
• Patients identified by the AI model showed higher risks of readmission and mortality compared to those identified by the MST.

Conclusions:

• The study validates the use of a novel AI model for predicting in-hospital malnutrition.
• This AI tool demonstrates superior performance in identifying at-risk patients.
• The findings support the integration of AI into clinical workflows for malnutrition management.