In Kigali, 28 INAs included 10 men and 18 women, with 5 religious group affiliates, 8 health-care providers, 5 NGO/CBO members, and 10 private sector representatives. In Lusaka, 33 INAs included 20 men and 13 women, of whom 10 were affiliated with religious groups, 8 were health-care providers, 6 worked with NGO/CBO, and 9 were from the private sector. The mean age of INAs was 40 years in both cities (range 20–65) and they lived in the catchment area of the CVCT centres. Forty-one of the 61 INAs (67%) received HIV counselling and testing prior to or during the training period, of whom 26 were tested with their spouses.
Invitations to attend CVCT
Invitations were distributed to 9,900 couples (2,680 in Rwanda, 7,220 in Zambia) in a 4 month period. The average ages of invited men and women in each country were similar (Rwanda: Men = 35.6 years, Women = 29.6 years; Zambia: Men = 34.5 years, Women = 28.1 years). Couples had been married or cohabiting for a mean of 8.0 years (Rwanda = 7.8 years, Zambia = 8.0 years), and had an average of 2.5 children (Rwanda = 2.5, Zambia = 2.4).
Table presents the contextual characteristics of CVCT invitations given by INAs, stratified by category and city. Invitations in Rwanda and Zambia were similar in that 56% were given to someone known to the INA, including professional or social contacts, friends, or family members. Rwandan INAs were more likely than Zambian counterparts to invite both members of the couple together (33% vs. 25%) rather than the man or woman alone. In both cities, 60% of invitations were delivered at the client's home, with the remainder given at the INA home, workplace or clinic, or in the community (including church, market, social gathering, or other). Zambian INAs were more likely to deliver invitations accompanied by their spouse (12% of invitations) or another INA (8%), while Rwandan INAs tended to work alone (95% of invitations). Invitations in Rwanda had more often been preceded by a public endorsement of CVCT than those in Zambia (11% vs. 6%).
Contextual characteristics of CVCT invitations given by INAs in four categories in Rwanda and Zambia
The relationships between invitees and INAs varied across categories in the two cities. In Rwanda, religious and private sector INAs were more likely to invite people they knew (68% and 64%, Table ) than CBO/NGO and health sector (45% and 49%) INAs, while religious and private sector Zambian INAs distributed half or more of their invitations to people they had just met (50% and 56%). Invitations from Rwandan health care INAs were the most likely to be given to couples (40%) compared with those from other INAs in Rwanda (25–31%), while religious and private sector INAs were more likely to target couples in Zambia (28%–29%) than their NGO/CBO (23%) or health sector (19%) counterparts. Compared to other categories, NGO/CBO INAs delivered a comparatively larger proportion of their invitations in their own homes (26% in Rwanda, 11% in Zambia), as did religious INAs in Rwanda (18%). Fewer than 6% of invitations delivered by the other INA country-categories were delivered in the INA's home. Workplace invitations were favoured by health care INAs in Zambia (46%) and religious INAs in Rwanda (34%), and these two groups were also more likely than their in-country counterparts to have invitations preceded by a public endorsement (14% and 24%, respectively).
Response to invitations
Of 9,900 couples who received invitations, 1,411 (14.3%) requested CVCT. The response rates, or percent of invited couples who came for testing, were much higher in Rwanda, where 721 of the 2,680 (26.9%) invited couples came in to be tested, than in Zambia, where only 690 of the 7,220 (9.6%) couples were tested. These rates were significantly different (t (9,839) = 6.55, p < .0001). Table details the number of invitations distributed, the number of invited couples who received CVCT, and the response rate for each INA, grouped by country and category and ordered by decreasing number of couples tested. Zambian INAs distributed more invitations (mean 219, S.D. 93) than Rwandan INAs (mean 96, S.D. 99), with greater variation among Rwandan INAs. Multiple regression analyses within each country separately revealed that the number of invitations distributed did not vary significantly by INA category, gender, age, marital status, or whether the INA had received HIV testing.
Number of invitations issued and couples tested by INA and category for Rwanda and Zambia
Although Rwandan INAs distributed fewer invitations, they had a higher response rate. As a result, the average number of couples tested per INA was not significantly different in the two countries (26 in Rwanda, 21 in Zambia). The main determinant of INA performance in Rwanda was the number of invitations distributed, as response rates were > = 20% for 22/28 INAs. In Zambia, only 2/33 INAs had response rates > = 20% and only 12/33 exceeded response rates of 10%. In Zambia, all INAs distributed a large number of invitations but had fewer invitees present for CVCT services. In both countries, there were a few 'high performing' individuals, who had both high numbers of invitations and a high response rate.
Predictors of successful invitations
Table summarizes the effects of univariate HLM analyses predicting response rates in each country separately. For categorical variables (e.g., location of invitation, INA category), dummy-coded vectors were entered as a set with the last category serving as a reference for the other categories. None of the INA-level measures were predictive of successful invitations in either country.
Summary of univariate effects from HLM analyses for predicting CVCT response rates in Rwanda and Zambia
Among invited couples, older age in men was marginally predictive of seeking testing in Rwanda, while older age in men and women and longer cohabitation were significant predictors in Zambia. In both countries, invitations delivered in the INA home were strongly predictive of a successful outcome, and workplace invitations also had a comparatively good yield. In Zambia, invitations delivered at the couples' home were also more likely to result in couples seeking testing compared to the "community" reference group. Invitations delivered to couples were associated with the greatest likelihood of CVCT, and those given to men were more likely to result in testing than those given to women alone. INAs had more success inviting people they knew, compared with people they had just met. Invitations given by the INA accompanied by their spouse or another INA were not associated with a different response rate than those given by the INA alone (not shown). In both countries, invitations delivered after a public endorsement of CVCT were associated with a higher response rate. This association was not significant for Zambia, but remained of interest with an exploratory p-value < 0.10.
Predictor variables with significant univariate effects (p < .05 in either country) were examined in the context of multi-predictor HLM analysis. Because years cohabiting, number of children, age of the man, and age of the woman were all highly correlated with each other (all r's ≥ 60), only the age of the man was included in order to minimize potential multicollinearity. None of the INA-level predictors achieved statistical significance in the univariate analysis, so these variables were excluded from the multi-predictor models.
Table presents the results of the multi-predictor analysis and confirms that invitations given in the INA home, workplace, and (in Zambia only) the couples' home remained independent predictors of CVCT. Similarly, invitations to couples and public endorsement remained predictive in the multivariate model. Invitations to family and to friends and social or professional contacts were independently predictive of success in Zambia but did not remain significant in Rwanda when controlled for the other variables measured.
Summary of effects from multi-predictor HLM analyses for predicting CVCT response rates- Rwanda and Zambia