The Role of Health Networks in Disseminating Knowledge about Infant HIV Testing in Rural Uganda: Population-Based Sociocentric Network Study

Table 4 Correlates of ego knowledge of infant HIV testing at 6 weeks and health network composition and structure (among non-isolated individuals)

	Outcome: Ego knows about infant HIV testing at weeks of age
	aOR	95% CI	p-value
*Health network composition*
Number of alters who know about infant HIV testing at 6 weeks (n)	0.99	[0.90–1.09]	0.85
High trust in network (0/1)	1.03	[0.78–1.37]	0.84
Daily contact with many alters (0/1)	1.31*	[1.00–1.71]	0.05
At least one female in network (0/1)	1.14	[0.82–1.58]	0.45
At least one PLHIV in network (0/1)	1.01	[0.69–1.49]	0.94
At least one friend in network (0/1)	1.22	[0.91–1.62]	0.18
At least one alter with some primary education or more (0/1)	0.75	[0.50–1.13]	0.17
At least one alter with children in network (0/1)	1.13	[0.72–1.76]	0.59
Mean age of alters (n)	1.01	[0.99–1.03]	0.31
*Health network structure*
Indegree	0.98	[0.91–1.05]	0.51
Outdegree	0.97	[0.87–1.09]	0.65
High closeness (harmonic centrality) (0/1)	1.69	[0.86–3.34]	0.13
High eigenvector centrality (0/1)	0.90	[0.45–1.80]	0.77
High authority (0/1)	1.81**	[1.18–2.77]	0.01
High hub (0/1)	1.25^	[0.98–1.59]	0.07
Constraint	0.84	[0.49–1.44]	0.54

Each row represents results from separate generalized linear models with exponentiated coefficients for adjusted odds ratios. Models use cluster robust standard errors clustered by village. Covariate adjustment includes ego sex, self-reported HIV status, age, marital status, educational level, asset index, biological children alive, village fixed effects and research assistant fixed effects. Trust measure is binary identifying those with average trust in alters greater than or equal to median trust in network. Frequency of contact is binary identifying those who see a high number of alters daily, great than or equal to median number of alters seen daily. Other composition measures are also binary identify at least one female, at least one PLHIV, at least one friend, and at least one alter with children. Measures for eigenvector centrality, authority and hub are binary measures identifying observations in the top 10th percentile. Closeness (harmonic centrality) is also a binary measure identifying observations above the lowest 10th percentile (given left skewness of data). ^ p < 0.10, * p < 0.05, ** p < 0.01, *** p < 0.001; aOR = Adjusted Odds Ratio, CI = Confidence Interval, PLHIV = people living with HIV