Abstract Body

Social network information is more accessible than sexual network information, and may capture HIV transmission risk. Social networks may also reveal the formation of peer relationships between HIV+ individuals, which are important for treatment support. To investigate the potential for social networks to capture these dynamics, we linked named social contacts to population-wide baseline HIV testing data from Kenyan communities in the intervention arm of the SEARCH HIV Test and Treat Study (NCT01864603).

During a census enumeration, 15,028 adult (age ≥15) residents of 3 communities named up to 6 friends in each of five social domains (health, money, emotional, food, and free time); 85% of residents were tested for HIV. Named contacts outside the household were matched to enumerated residents to construct community-wide social networks (Figure). Targeted maximum likelihood was used to estimate the relative risk of HIV associated with exposure to an HIV+ cross-gender or same-gender contact, adjusting for demographics, circumcision, alcohol use, contraception, mental health, work productivity, mobility and household wealth. Confidence intervals were adjusted for multiple comparisons across social domains.

HIV prevalence was 16%; 12% in all men, 19.5% in all women; 1% in young (<25) men, and 9.8% in young women. Men with an HIV+ female contact in any domain were at increased risk of HIV (aRR:1.5; 95%CI:1.1,2.0). Women with an HIV+ male contact in any domain (aRR:1.4; 95%CI:1.1,1.8) or in the emotional domain (aRR:1.8; 95%CI:1.2,2.7) were at increased risk of HIV. Women’s increased risk was amplified if the HIV+ male contact was >10 years older (aRR:1.6; 95%CI:1.1,2.5). Among young women, higher risk was associated with having an older HIV+ man in the free time (aRR:7.9; 95%CI:3.2,19.4), food (aRR:4.1; 95%CI:1.4,11.7), health (aRR:3.5; 95%CI:1.2,10.1), or any domain (aRR:3.4; 95%CI:1.6,7.1). Women with an HIV+ female contact in the health domain (aRR:1.6; 95%CI:1.2,2.1) were also more likely to be HIV+.

In this cross-sectional analysis of social network predictors of HIV risk in rural Kenya, exposure to HIV+ contacts within easily assessed social networks significantly predicted HIV risk and supported both known sexual transmission dynamics, such as those between young women and older men, and HIV+ peer relationships. These data may be useful for designing prevention and treatment support interventions for at risk populations.