Machine learning for predicting cognitive deficits using auditory and demographic factors

Area of Science:

• Neuroscience
• Infectious Diseases
• Machine Learning

Background:

• Cognitive dysfunction is a concern in people living with HIV (PLWH), particularly in low-resource settings.
• Early detection of cognitive impairment is crucial for timely intervention.
• The predictive value of auditory assessments for neurocognitive deficits, beyond demographics, requires further investigation.

Purpose of the Study:

• To utilize machine learning to predict neurocognitive deficits.
• To evaluate the efficacy of auditory tests and demographic factors in predicting cognitive impairment.

Main Methods:

• A longitudinal study was conducted in Dar es Salaam, Tanzania, involving 478 participants (349 PLWH, 129 HIV-negative).
• Machine learning algorithms were trained and tested using auditory variables and demographic data.
• Neurocognitive dysfunction was defined as a score <26 on the Kiswahili Montreal Cognitive Assessment.

Main Results:

• Machine learning models incorporating auditory variables demonstrated superior prediction of neurocognitive deficits.
• Gaussian and kernel naïve Bayes classifiers showed high predictive performance, with AUC values of 0.91 and 0.87, respectively.
• Models without auditory features performed significantly worse (p < .001), highlighting the incremental predictive value of auditory assessments.

Conclusions:

• Auditory variables significantly enhance the prediction of cognitive function.
• Easy-to-administer auditory tests can serve as objective predictors of neurocognitive performance, suitable for global application.
• Further research into machine learning algorithms for predicting cognitive outcomes is warranted.