The Application of Machine Learning in Clustering Borderline Mild Cognitive Impairment among Aging Thai People Living with HIV

Abstract

Many people living with HIV (PLWH) have cognitive impairment. Details of cognitive impairment subtypes are lacking. Unsupervised machine learning (ML) can reveal hidden subgroups within heterogeneous data. The study aimed to determine clusters of aging Thai PLWH with borderline cognitive impairment using unsupervised ML. HIV-NAT 207 study enrolled Thai PLWH aged ≥50 years. Cognitive performance was evaluated by the Thai-validated Montreal Cognitive Assessment (MoCA). This study included participants who scored between 23 and 27. The score of each cognitive domain served as cluster variables for the K-means algorithm. Among 340 PLWH, 177 (52.1%) scored between 23 and 27. Median age was 54 (IQR = 51-58) years, 118 (66.7%) were male, median CD4 was 620 (IQR = 489-795) cells/µL, and 170 (96.1%) were virally suppressed. K-means cluster demonstrated five clusters of all participants: 22.0% cluster 1 (marked memory with mild language impairment), 25.4% cluster 2 (mild visuospatial/executive function-language-memory impairment), 19.2% cluster 3 (moderate abstraction with mild visuospatial/executive function-language-memory impairment), 18.6% cluster 4 (marked language with mild memory impairment), 14.7% cluster 5 (marked language-abstraction impairment). A longitudinal study is warranted to identify differences in clinical significance and prognosis between each cluster.