This study documented the cost of a successful community-level integrated prevention campaign for HIV, malaria, and diarrhea in rural Kenya. We used these data to project the cost of achieving further scaled-up coverage of these interventions, with specified anticipated efficiencies in resource use and in unit costs. We project that participants could be provided with a high performance water filter, a long-lasting insecticide-impregnated bed net, HIV counseling and testing, health education regarding the use of these commodities, and 60 condoms at an estimated cost of $31.98 per person.
This estimate is consistent with subsequent experience with IPC implementation in Western Kenya. In a smaller campaign in Kisii and Kisumu (6 sites and 10,206 individuals reached), the total estimated cost per person was $31.08, slightly below the SUR estimate. Planning and set-up costs were $1.19 per person, as compared with $1.07 estimated in SUR. For a planned provincial level implementation, we received a price quote for social mobilization equivalent to $0.45 per person expected to be served by the campaign, lower than our SUR estimate ($0.83).
Compared with prior reports of the cost of providing these services, the integrated campaign approach appears to be efficient. HIV counseling and testing in Africa has had cost estimates ranging widely: $101 in South Africa, $26 in Kenya, $16 in Kenya, $7-11 in South Africa, $3-$100 (median $10.50) in Uganda, $8-$19 in Uganda, and $3-6 in Tanzania. Of these, only one is for programs that achieve high (>90%) community coverage. Our estimate of $9.92 for the SUR integrated campaign is toward the low end of this range, and includes 60 condoms valued at $1.40, as well as CD4 testing for those found to be HIV-positive. A formal comparison across studies is difficult, because varied timing affects both the technology (which has dropped in price), as well as rising prices for consistent inputs such as labor. The door-to-door VCT Uganda cost of $8.29  was for 2007, and thus to be financially comparable to our 2008 estimate would be inflated by 6%i, to $8.79.
Long-lasting insecticide-impregnated bed nets have had cost estimates (including delivery) in a narrower range: $6 in Togo in 2004, $9-27 including insecticide retreatments in several African settings, and $8 in Tanzania in 2006. The cost we estimated for a scaled-up campaign, $6.27, is in the lower end of this range, using a long-lasting bed net.
Water filters have been estimated to cost $3 per person-year of use. Transforming the cost of $15.79 per campaign participant (allowing for 2.5 participants per household, 7.73 individuals per household, and two years' benefit) yields an estimated $2.55 per person-year, slightly below the average for filters. Other strategies (e.g., chlorination) are less expensive, but may be less effective, as they require more frequent replenishment of supplies. Solar disinfection, though also low cost and effective, has limited evidence of scalability.
A likely explanation for the low delivery cost is apparent economy of scope in providing all the interventions at once. Seventeen percent of costs were associated with the general campaign infrastructure and general personnel, as compared with 83% for the intervention-specific components (e.g., commodities and training on their use, as well as HIV counseling). If we treat these 17% of costs as fixed, then the incremental costs of adding second and third interventions - i.e., the interventions that "integrate" - are decreased by 17% to $5.20 for malaria, $13.11 for diarrhea, and $8.23 for HIV. Thus, the first intervention incurs the full cost of the general campaign structure, and subsequent interventions require only the incremental costing specific to that disease. In other words, the second and third disease programs build efficiently on the base of the general campaign structure established for the first disease program.
We believe that our findings may apply to other rural areas with a similar mix of disease burden from HIV, diarrhea, and malaria and a similar level of service need for HIV testing and care, clean water, and protection from insect vectors. This might include much of Sub-Saharan Africa and many parts of Asia.
Our analysis has important limitations. First, our estimate of the cost of scaled-up replication (SUR) is based on our best assessment of potential reductions in resource intensity and unit costs, albeit supported by subsequent IPC activities. We could not find a validated method for this analysis. As shown in the sensitivity analyses, key uncertainties are associated with a 4-10% variation in results. Empirical verification with a scaled-up implementation is imperative. This is especially true for different country settings, where political, social, and health care contexts may affect IPC logistics and participation.
Second, we examined the cost of using specific commodities, such as the LifeStraw® water filter. Filters were an integral component of social mobilization for the campaign, serving as an incentive to achieve high VCT coverage. We cannot predict participation with a different approach.
Third, our analysis did not examine the effect of the IPC. We have separate publications under review on commodity uptake and use, and on cost-effectiveness, as well as two encouraging qualitative assessments of reactions to IPC participation [24, 25]. A post-campaign survey found commodity use levels of 60% - 90% (unpublished data, Grabowksy, M). Sustained use (e.g., 2-3 years) is essential to benefit, though we note that studies of the magnitude of health effects of commodity distribution are typically conducted at the community level, not for individuals with confirmed use. Commodity use should be encouraged by the substantial on-site health education component. Further, the Ministry of Health was a key partner in implementation, and referrals for HIV care were coordinated with local providers.
Fourth, it is possible that a more efficient, larger campaign might compromise quality. Key features of campaign quality include: ability to attract participants; acceptability of the participant experience; effective training in the use of nets and filters; HIV counseling and testing up to usual standards; and commodity use in the community. However, we believe that there is reason to be optimistic. As noted above, service delivery staffing in the SUR is at levels used in the campaign. Manager staffing is above levels found adequate in the latter parts of the campaign and in a subsequent smaller campaign. Ultimately, both efficiency and quality must be directly assessed in a campaign implemented on a larger scale, e.g., provincial.
Various managerial strategies may be employed to achieve efficient scale-up. A modular approach, in which several geographically defined population groups are targeted each week, appears reasonable to us, and is reflected in our SUR analysis. This approach builds on the initial implementation, which was in effect one module. The number and size of modules can be adjusted according to the availability of HIV testing counselors (likely the limiting resource) and the desired speed of large-scale implementation.
Financial incentives are another tool. For example, productivity-based payment (a fixed amount per campaign participant) may encourage continued or increased efficiency. Quality and financial controls may be especially important under such a system. For example, if the average counselor time per individual receiving VCT is shortened, direct supervision and counseling records must assure that key risk reduction messages are delivered and that HIV-infected individuals, who appear to offer the best opportunity for significant risk reduction, are fully and effectively counseled.