Supervised machine learning models for classifying common causes of dizziness.

Area of Science:

• Neurology
• Data Science
• Public Health

Background:

• Dizziness affects 16.8% of US adults annually.
• Accurate classification of dizziness subtypes is crucial for effective treatment.
• Common subtypes include vestibular migraine (VM), benign paroxysmal positional vertigo (BPPV), Ménière's disease (MD), persistent postural-perceptual dizziness (PPPD), superior canal dehiscence (SCD), and bilateral vestibular hypofunction (BVH).

Purpose of the Study:

• To identify key factors for classifying common dizziness types using a supervised machine learning (ML) platform.
• To leverage a national dataset for robust classification model development.

Main Methods:

• Utilized the 2016 National Health Interview Survey data (n=33,028).
• Generated case definitions for six dizziness subtypes based on clinical criteria and survey responses.
• Developed decision tree models using 136 variables (sociodemographics, comorbidities) to predict dizziness types.

Main Results:

• Machine learning models achieved high accuracy in classifying all six dizziness subtypes (sensitivity: 70-92%; specificity: 89-99%).
• Classification relied on functional limitation questions, such as falls and social activity modifications due to dizziness.
• Vestibular migraine (VM) was a prevalent subtype, affecting 4.0% of respondents.

Conclusions:

• Supervised ML models accurately predict dizziness subtypes using non-symptom-related functional limitation data.
• This demonstrates the potential of ML in analyzing large datasets for dizziness classification.
• Findings highlight the importance of functional impact in understanding and diagnosing dizziness.