Machine Learning Enhances the Efficiency of Cognitive Screenings for Primary Care

Area of Science:

• Gerontology
• Computational Neuroscience
• Primary Care Medicine

Background:

• Primary care settings require efficient cognitive screening tools.
• The Montreal Cognitive Assessment (MoCA) is a widely used clinician-administered instrument.
• A brief, self-administered computerized assessment protocol was developed for cognitive screening.

Purpose of the Study:

• To evaluate the convergence validity of a brief, self-administered computerized cognitive assessment.
• To compare the computerized test's performance against the Montreal Cognitive Assessment (MoCA).
• To assess the potential of machine learning algorithms in analyzing computerized cognitive test data.

Main Methods:

• 206 participants completed both the MoCA and the computerized test.
• Three machine learning algorithms (Support Vector Machine, Random Forest, Gradient Boosting Trees) were trained.
• Synthetic Minority Oversampling Technique (SMOTE) was used to address class imbalance.

Main Results:

• Gradient Boosting Trees achieved high accuracy (0.81) and area under the curve (0.81).
• K-means clustering identified three cognitive categories: unimpaired, mildly impaired, and moderately impaired.
• The computerized test showed stronger correlation with age in unimpaired individuals than the MoCA.

Conclusions:

• The computerized cognitive screening tool demonstrates strong validity and potential for primary care.
• Machine learning effectively classifies cognitive impairment based on computerized test performance.
• Further research should enhance the computerized test's sensitivity across cognitive domains while maintaining efficiency.