Identification of Clinical Phenotypes Among People with HIV Using Electronic Health Record Data

Area of Science:

• Public Health
• Clinical Medicine
• Data Science

Background:

• Loss to follow-up (LTFU) impacts nearly half of people with HIV (PWH), hindering effective care.
• Electronic health records (EHR) offer rich data for understanding patient heterogeneity in HIV management.

Purpose of the Study:

• To identify distinct clinical phenotypes among PWH using EHR data.
• To assess the association between these phenotypes and LTFU from HIV care.

Main Methods:

• Latent Class Analysis of 4,316 visits from 849 adults at an urban HIV clinic (2017-2020).
• Extracted demographic, social history, laboratory, diagnosis, and clinical note features.
• Logistic regression analyzed associations with LTFU and high viral load.

Main Results:

• Six distinct patient subgroups (phenotypes) were identified.
• Phenotype 2 (younger men, new patients) showed the highest LTFU risk (OR 1.57) and elevated viral load (OR 1.74).
• Phenotype 4 (White/Hispanic, fewer substance use mentions) had increased LTFU risk (OR 1.39) but decreased unsuppressed viral load risk (OR 0.65).

Conclusions:

• Six clinically relevant PWH phenotypes were identified in an urban HIV clinic setting.
• Two phenotypes were significantly associated with increased risk of LTFU.
• Findings can guide tailored interventions to improve retention in HIV care.