MUST-Plus: A Machine Learning Classifier That Improves Malnutrition Screening in Acute Care Facilities

Area of Science:

• Clinical Nutrition
• Health Informatics
• Machine Learning in Healthcare

Background:

• Hospital malnutrition is prevalent, under-diagnosed, and negatively impacts patient outcomes and healthcare costs.
• Existing screening tools like the Malnutrition Universal Screening Tool (MUST) have suboptimal predictive value.
• Accurate malnutrition screening is crucial for timely nutritional intervention.

Purpose of the Study:

• To develop a machine learning (ML)-based classifier, MUST-Plus, for more accurate malnutrition prediction.
• To enhance the predictive accuracy beyond conventional screening methods.
• To improve the early identification of at-risk patients.

Main Methods:

• A retrospective cohort of adult inpatient data (anthropometric, lab, clinical, demographics) was analyzed.
• A random forest model was trained using 10-fold cross-validation.
• Performance was evaluated against registered dietitian assessments as the gold standard and compared to MUST.

Main Results:

• Malnutrition was present in 13.3% of the test cohort.
• MUST-Plus achieved 73.07% sensitivity and 76.89% specificity.
• MUST-Plus demonstrated a 30% higher sensitivity, 6% higher specificity, and 17% increased AUC compared to MUST.

Conclusions:

• Machine learning-based MUST-Plus significantly outperforms the classic MUST in identifying malnutrition.
• MUST-Plus can enhance operational efficiency for registered dietitians through timely patient referrals.
• This tool offers a more accurate approach to malnutrition screening in hospital settings.