Invitation-level predictors of testing among cohabiting couples included inviting the couple versus the woman/man alone (aOR=1.2); also couple (aOR=1.4) or woman (aOR=1.6) versus INA initiated contact was predictive.
Generalised estimating equation (GEE) methods evaluated the association between couple-level and invitation-level characteristics predictive of successful invitations.
GEE methods accounted for clustering of couple-level and invitation-level characteristics within individual INAs and INLs.
13 Similarly, the Copperbelt study described previously did not examine INA-level, couple-level or invitation-level predictors of success.
GEE analysis methods with an exchangeable correlation structure accounted for two-level clustering of couple-level and invitation-level characteristics within individual INAs and INLs.
Counts (percentages) for categorical variables and means (SD) for continuous variables were calculated for INL-level, INA-level, couple-level and invitation-level characteristics.
Since couple-level and invitation-level data are clustered at two levels, within-individual INAs and INLs, GEE methods accounted for non-independence of observations.
Results from a pilot study of promotions in Lusaka with 33 INAs (no INLs) showed that, while invitation-level predictors were similar to those found in this larger study, the small sample size did not allow simultaneous detection of INA-level, couple-level and invitation-level characteristics in hierarchical analysis.
Here, we not only demonstrated the feasibility of CVCT promotions using influential network agents and leaders (INAs and INLs) to promote CVCT, but also identified INA-level, couple-level and invitation-level predictors of CVCT uptake.
INA-level, couple-level and invitation-level variables significant (Bonferroni corrected p value=0.002) in univariate analyses were entered into a multivariate logistic regression model, and the variables were examined for multi-collinearity.
4 6 7 36 40 Here, we demonstrated not only the feasibility of CVCT promotions using INAs and INLs, but also identified practical INA-level, couple-level and invitation-level factors that were marginally though significantly predictive of CVCT uptake in these analyses.
