Fig. 3.

Effect of scaling up HB-HCT activities on the average cost at two different values of contact quality (fitted values estimated in Model 4). Legend: The figure plots the fitted values of average cost (with 90% confidence intervals, CI) estimated in Model 4 across the range of the scale variable at two different values of contact quality (62.8% and 82.8%). As the mean contact quality was 0.728 (i.e., on average 72.8% of all contacts eligible for an HIV test were eventually tested for HIV), the two values considered are 10 percentage points below and above the sample average. For instance, for a contact quality of 62.8%, increasing the scale from 100 to 200 contacts in a cluster-month would decrease the average cost per contact from $21.9 to $17.8 (p < 0.001). On the other hand, the average cost would decrease from $30.6 to $26.6 (p < 0.001) for the same increase in scale but a contact quality of 82.8%. Although both increases in the number of contacts would lead to significant economies of scale, the decrease in the average cost would be lower for a higher contact quality (−13.1% versus −18.7% when 62.8% of the contacts eligible for an HIV test were eventually tested, p < 0.10).