In this study examining the costs and outputs of 14 I-TECH sponsored HIV treatment case management sites in Ethiopia, we found that the I-TECH Ethiopia program delivered 4,589 patient-quarters of services, served 5,056 patients and 1,995 successful exits at a cost of $167,457 over 12 months, or $36 per patient-quarter, $33 per patient served, and $84 per successful exit. Unit costs varied inversely with scale across sites. These results suggest that higher services volume may be an important determinant of higher efficiency for CM. However, there are limitations to what program managers can do to increase ART and CM uptake, as this depends importantly on HIV prevalence, population density, and access issues which may be outside of their control. However, we also found strong negative correlation between client-staff ratios and unit costs. This indicates that observed economies of scale are not only due to the spreading of fixed costs over larger services volumes, but that higher service volumes are associated with higher staff productivity. How the relationship between volume and productivity is mediated is not clear from our data. Personnel costs are conventionally classified as variable costs. However, a case manager may in the short run be a fixed cost in that "one" is the indivisible lowest unit that can be assigned even to a low-volume site. It is possible that efficiency would be increased, even absent greater volume if, at sites with higher costs and lower volumes, case managers were able to divide their time between CM and other HIV-related services. Alternatively, if there are sites with under-utilized case manger staff time, it might be possible to intensify CM services to clients and potentially generate a larger number of successful exits with existing staff levels. Some degree of flexibility and adaptability of the CM model to local conditions may be essential for efficient service delivery. Indeed, at some sites case managers have started to assume limited responsibility for other services to HIV inpatients.
A significant limitation of this analysis is that it was not always possible to quantify the portion of administrative staff time and other overhead resources that would have been needed in the absence of the CM program, i.e., the incremental opportunity cost of the overhead inputs. We estimated the average cost of general overhead inputs by allocating them to CM according to the proportion that CM service providers constitute of all I-TECH service providers supported by these overhead functions. A second limitation is that the definition of "successful exits" from the program was not operationalized consistently across the study sites, nor was it linked to quantified improvements in adherence. However, the high correlations between service volumes and "successful exits," and between the unit costs of these two outputs measures, suggest that despite variation across sites in its precise operational definition, the number of "successful exits" appears to be an outcome predominantly determined by the extent of services. Thus, varied definitions may still represent aspects of the same underlying process.
Cost analyses have a variety of important management and budget projection uses. However, since they do not evaluate health outcomes, they do not provide the guidance for program funding or resource allocation decisions conferred by cost-effectiveness analyses. A definitive analysis of the cost-effectiveness of improving adherence would entail a comparative assessment of an array of effective measures to improve adherence. These include eliminating ART co-payments (in settings that have them), payments for drugs to treat opportunistic infections, improved personnel training, and providing meals and reimbursing for transportation for participants [11
]. They might also include use of electronic devices for medication use monitoring, other economic incentives, and cell phone-based supervision [15
]. A recent review of interventions to increase adherence to ART in sub-Saharan Africa found that some methods appear to have benefit but that these benefits may be limited and temporary, or be confined to specific settings or the precise intervention content. In addition many of the results depended on observational studies rather than controlled trials [16
]. Thus, completion of a technically sophisticated analysis may be years away. Pending funding for randomized, multi-arm trials that document the costs and benefits of a range of adherence promotion strategies, a first step would be a cost-effectiveness analysis of a CM program versus a standard ART program without any adherence intervention. Lacking strong evidence on which candidate methods are likely to be most cost-effective, program managers should also be encouraged to try methods they deem appropriate for their respective settings, and to document the costs and resulting changes in adherence or health status.
Absent a cost-effectiveness assessment, it is nevertheless useful to place our unit cost estimates in context. ART has recently been estimated to cost $643 per patient-year in Ethiopia [17
]. Discounting at 3% per annum, and assuming 12 ART-years, the lifetime cost of ART is $6,400. The CM program thus adds 0.52% ($33/$6,400) to the lifetime cost of treatment for those patients receiving CM. Let us assume, perhaps pessimistically, that the CM program has no effect on 90% of those classified as having had a "successful exit" and that the remaining 10% are returned to normal adherence. The $840 ($84/10%) per CM client restored to normal adherence would thus add 13.1% to the lifetime cost of ART ($840/$6,400) per successful exit; or the equivalent of an additional 16 months of ART ($840/$643) for the patients receiving CM. This rough calculation does not account for changes in treatment costs resulting from greater adherence, or for the potential reduction in transmission to partners including the transmission of drug-resistant strains of HIV.
The evaluation of incremental cost-effectiveness may not form a sufficient basis for deciding whether to invest in CM services. Once patients have been accepted for treatment, there is an ethical obligation to make reasonable efforts to ensure that they benefit from those services. Yet, this perspective begs the question of what constitutes a "reasonable" additional effort, especially in resource-constrained countries such as Ethiopia. Program managers will not invest unlimited resources in a small number of poorly adhering patients, at the expense of expanded access to ART or less intensive but successful CM efforts. Ultimately, what is deemed "reasonable" depends on a wide range of factors that may be best decided by local program managers. Nevertheless, this decision should be informed by an understanding of the cost-effectiveness of additional investments in CM. Finally, the variation in unit costs reported here suggest that there may be opportunities to identify staffing patterns that increase overall program efficiency, thus freeing increasingly constrained HIV resources with no additional investment.