Phenotyping Obstructive Sleep Apnea Patients: A First Approach to Cluster Visualization.

Area of Science:

• Sleep Medicine
• Data Science in Healthcare
• Medical Informatics

Background:

• Obstructive sleep apnea (OSA) presents diverse phenotypes, complicating diagnosis.
• Objective patient stratification is needed for effective OSA management.
• Current diagnostic approaches may not fully capture OSA heterogeneity.

Purpose of the Study:

• To apply k-modes clustering for identifying distinct patient groups within obstructive sleep apnea.
• To stratify OSA patients using demographic, clinical, and comorbidity data.
• To explore objective methods for recognizing OSA patient subtypes.

Main Methods:

• K-modes clustering algorithm applied to a dataset of 318 OSA patients.
• Utilized 41 characterization variables including demographics, physical exam, and clinical history.
• Missing data imputed using k-nearest neighbors (k-NN); chi-square test performed.

Main Results:

• Three distinct patient clusters were identified based on 13 selected variables.
• Cluster 1: middle-aged men; Cluster 2: middle-aged women; Cluster 3: oldest men.
• Cluster 3 showed highest weight and increased neck circumference; OSA severity predominantly mild.

Conclusions:

• Objective phenotyping of obstructive sleep apnea patients is achievable using clustering techniques.
• Identified patient clusters based on age, sex, and physical characteristics.
• Enhanced data visualization may improve recognition of OSA subtypes and pathology.