Sub-district Costs and Efficiency of the Re Mmogo Pholong (“Together in Wellness”) Combination HIV/AIDS Prevention Intervention in the North West Province of South Africa

Background Re Mmogo Pholong (RMP) or “Together in Wellness”), was a combination prevention program to strengthen HIV prevention programming, community support mechanisms, community-based HIV testing, referral systems, and HIV prevention integration at the primary care level, thereby sustainably reducing HIV/AIDS transmission in the North West Province of South Africa. RMP included four overlapping components: situational analysis, community engagement and mobilization, community-based biomedical and behavioral prevention, and primary health care systems strengthening. In support of the PEPFAR country-ownership paradigm, we conducted costing analysis of the RMP combination HIV prevention program to determine data needed for potential transition of to local ownership. Methods We used standard costing methodology for this research. Results We found that cost per unit of output ranged from $63.93 (cost per person reached with individual or small group prevention interventions) to $4,344.88 (cost per health facility strengthened). The RMP intervention was primarily dependent on personnel costs. This was true regardless of the time period (Year 1 vs. Year 2) or activity (i.e. wellness days or events, primary health care strengthening, community engagement, and wellness clubs). Conclusions The development of labor-intensive rather than capital intensive interventions for low-income settings such as RMP was identified as being particularly important in treating and preventing HIV/AIDS and other health conditions in a sustainable manner. Costs were also observed to transition from international cost centers to in-country headquarters offices over time, in keeping with the transition of international to local responsibility required for sustainable PEPFAR initiatives. Such costing center evolution was also reflected by changes in the composition of the intervention, including (1) the redesign and re-deployment of service delivery sites according to local needs, uptake and implementation success and (2) the flexible and adaptable restructuring of intervention components in response to community needs.


The Role of HIV Counseling and Testing and Combination Prevention
Combination HIV/AIDS prevention approaches, including multi-level initiatives that combine community mobilization, counseling and testing, and post-test support with other health services (e.g. Khumalo  Province, South Africa. RMP included four overlapping components: situational analysis, community engagement and mobilization, community-based biomedical and behavioral prevention (i.e. wellness days and wellness clubs), and primary health care systems strengthening. This comprehensive, multi-level and holistic strategy also aimed to sustainably reduce HIV/AIDS incidence and prevalence via integration with longer-term structural health care system change in the form of improved access to (and quality of) referral systems to higher-level care.

Situational Analyses & Intervention Components
Situational analyses were conducted in both sub-districts in Year 1 (September 2011 to September 2012) (1) to understand the epidemic response in the community, (2) engage in community mapping, and (3) design programmatic interventions.
Community mobilization and engagement took place through the development of community working groups (CWGs), engaging the South African Department of Health (DoH) for community entry and report-back on activities, and community mobilization strategies based on raising community awareness and engagement around HIV/AIDS prevention, and was considered an essential element of the broader program.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint Related RMP service biomedical delivery was offered at "wellness days" and "wellness events", building on evidence of their cost-effectiveness in the South African setting (Moodley et al, 2016). These provided HCT, referred and linked HIVpositive persons to care and treatment, provided pregnancy testing and family planning options, conducted rapid syphilis testing in pregnant women, provided syndromic management for STs and symptomatic screening and referral for TB to promote safer sexual behaviors, and emphasized the importance of consistent and correct condom use through health education. Wellness days were considered as extensions of public health care services, and referrals and linkages to Primary Health Centers (PHCs) were a primary focus.
Of note, those who received HCT also received support and guidance regarding personalized behavioral risk reduction, including partner and family testing, partner reduction, and disclosure of HIV and STI status. Individuals testing HIV-positive were referred to wellness clubs for psycho-social support, coping mechanisms, awareness of behavior change recommendations, and life skills development (e.g. patient goal setting, household and medical budget management, and involvement with food gardens and nutritional security). In each of the program components, linkage and referral to HIV care and treatment was a primary focus, and prevention amongst people living with HIV/AIDS (PLHIV) was an emphasis. We aimed to build on related research exploring the costs and outcomes of HIV/AIDS interventions in South African rural areas which often face very different operating environments and expenses compared to their urban counterparts (Mbonigaba and Oumar, 2017).

Costing the RMP Program
A key feature of the sustainability, transferability, and effectiveness of global health intervention implementation and roll-out involves the understanding and quantification of key costs (and related resources) required for service delivery. The costing process may also help to inform broader policy decisions related to resource allocation across HIV/AIDS treatment and prevention programmes (Marseille & Khan 2002), as well as providing a more detailed understanding of the key cost . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint centers within program components. In the case of the RMP program, combined service delivery initiatives focused on provision of services via community engagement, wellness days, wellness clubs, and primary health center strengthening (the 'cost centers'), across personnel, transport, facilities, supplies, and equipment cost categories.
The stratification of costs according to such categories is also related to geographical regions of expenditure (e.g. intervention sub-districts). We categorized and review related costs according to these classifications and approaches for the first two fiscal years of the intervention. Such "efficiency comparisons" represent a useful tool for local and international global health program managers and policymakers to determine returns on program investments (Jamison et al 2006). The use of output information from intervention monitoring and evaluation (M&E) activities, in conjunction with costing data, also presented opportunities to assess the efficiency of the implementation of RMP.

Data Collection
Data for the costs incurred by the RMP program were obtained in several stages.

RMP Site Visit
A costing site visit was conducted in February 2014 to observe service delivery at a two-day "wellness day" event in Bakubung Village, Moses Kotane sub-district, in the . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint Bojanala Platinum District. This site visit allowed for (1) identification of any potential hidden costs not evident in the financial data and (2) a greater understanding of the operational needs of the intervention from a costing perspective.

Data Entry and Stratification
All transactions for in-country cost data collection were originally presented in South African Rand (ZAR). Periodical average exchange rates were used to convert transaction costs to US dollars for the two fiscal years. Within each year, we then categorized costs by (1) location (e.g. Seattle headquarters office, in-country central office, and implementation sub-district); (2) intervention activity (e.g. headquarters oversight expenditures, in-country management and office costs, wellness days, wellness clubs, implementation science, primary health center strengthening, and community stakeholder engagement); and (3) economic resource categories such as capital (e.g. large, single purchase, equipment costs); personnel (e.g. salaries, benefits, and per diems,); utilities (e.g. electricity, communications); transport (e.g. vehicle maintenance); meetings (e.g. conference fees); oversight and support (e.g. maintenance and repair, or information technology and computing costs) 1 ; space (e.g. office rental, conference facility rental); travel (e.g. air fare, shuttle service); medical supplies (e.g. HIV/AIDS testing kits); non-medical supplies (e.g. printing and copying or computer software); miscellaneous costs (e.g. freight and express, insurance, legal fees, and postage); and other costs (e.g. indirect costs related to administration).

Inclusion of Output Data to Inform Efficiency Comparisons
As part of broader project oversight efforts, output data recorded during Year 2 was collected from project managers at the headquarters and field levels. Output information was then linked with site-specific costs aggregated by economic costing category (e.g. personnel) and divided by activity category (e.g. wellness days). These output values were then combined with component, economic category, and sitespecific costs to generate a series of cost-efficiency ratios across sites. Shared costs or costs which could not be allocated to either sub-district were allocated proportionally by known site-specific costs. These results were, in turn, directly . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint associated with (1) key costing groups (e.g. wellness days) and (2) geographical areas of service delivery (e.g. the Moses Kotane and Naledi sub-districts). Efforts to include the widest range of intervention outputs were made in keeping with current recommended practices (e.g. Padian et al 2012), helping to identify those sub-districts that performed with, for example, higher levels of productivity or lower costs.
Outputs related to key intervention activity categories of (1) community stakeholder engagement; (2) wellness days; (3)  Third, costs were divided by standard economic costing resource category (e.g. personnel, space, supplies, and travel costs). Fourthly, costs were allocated by activity component (e.g. wellness days and wellness clubs). Separately, for RMP outputs, information was transferred from standard reporting formats (e.g. Excel . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint spreadsheets or tables embedded in Word and PDF documents) into a collective and inclusive Excel output database in which all outputs were quantified and divided into group (e.g. wellness clubs) and individual (e.g. number of HCT clients) level results.
Efficiency comparisons were then completed by developing activity-specific subdistrict-level costs and combining with key output categories. 2 Finally, where multiple outputs were generated by a single activity category (e.g. the number of HCT clients and number of individuals reached for the wellness day activity), total component costs were applied to all output categories. Over sixteen percent (16.6%) of costs were attributable to in-country management and office support, while wellness days accounted for 14.6% of activity costs. In turn, incountry management costs were primarily generated by personnel costs (65.8%), while wellness day costs were primarily generated by equipment costs (53.2%).

Year 1 Costs
When stratified by economic category, key cost centers included personnel, oversight and support (64.1% of total costs) and equipment (9.3%). Of personnel costs, 47.1% of all costs were attributable to University of Washington costs, 17.0% to in-country office management, and 16.4% to implementation science. Of note, medical supplies 2 Activity-specific costs for the Moses Kotane and Naledi sub-districts were generated via allocation of costs specifically assigned to each sub-district, as well as the proportional distribution of headquarters other cross-district costs.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made were almost equally attributable across wellness days (19.1%), implementation science (20.6%), and in-country management costs (19.8%). Finally, when stratified by economic resource category, costs were primarily generated by personnel, oversight and support (73.1%), travel (4.5%), and transport (2.6%).

Output Results
Output results for Year 2 of the intervention were comparable between sub-districts for all group (e.g. wellness days, post-test 'wellness' clubs) and individual (e.g. number of HCT clients) level output categories (Table 1).

Table 1: Costs and Outputs of RMP by Activity
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made

Sensitivity Analysis
Shared costs across Naledi and Moses Kotane per unit of output were determined on a proportional basis according to the distribution of assigned or "pre-allocated' costs across communities. This approach was based on a proportional distribution of South . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint African headquarters office, sub-district, and unknown costs. In order to ensure that our approach was appropriate and representative, we conducted a sensitivity analysis based on the equal distribution of headquarters, shared sub-district, and unknown costs across the two communities. Based on this analysis, costs per unit of output or outcome were not found to vary significantly. For example, cost per wellness event changed from $2,644.78 (Naledi) and $2,185.77 (Moses Kotane) to $2,648.02 and $2,188.45, respectively.

Key Findings
Our key findings suggest that (1)

Cost Distribution across Sites
When reviewed by site, in both Year 1 and Year 2 of the RMP intervention costs were focused on the University of Washington --with particular reference to personnel costs associated with program management --indicating the importance of international involvement and support for the RMP intervention as a requirement for successful service delivery (with specific reference to the initial stages of the intervention) as well as protocol development and planning for the intervention, as

Component and Activity-Related Costs
Analysis of activity-related costs suggested that the RMP intervention was highly dependent on the use of personnel. The development of labor-intensive rather than capital intensive interventions for low-income settings is particularly important in treating and preventing HIV/AIDS and other health conditions (Marseille & Kevany 2011), with particular reference to regions where a high availability of low-skilled personnel is available (as opposed to potentially scarce and expensive capital equipment). The associated importance of program management activities was also reflected in the personnel share of intervention costs. In addition, the low costs related to utilities and supplies also suggested that the RMP intervention would be easily transferrable to other low-income settings, a key feature of adaptable HIV/AIDS interventions (Kevany et al 2013)..

Explaining Differences in Efficiency across Sub-Districts
Program efficiency was found to be driven by productivity of outputs rather than by differences in sub-district-level costs. For example, the Moses Kotane sub-district recorded approximately double the level of outputs (compared to Naledi) across most indicators. A range of contextual, environmental and logistical considerations should be taken into account in this context. First, the Naledi sub-district was a geographically larger area, with populations spread over larger distances, which may affect uptake of services. By contrast, the Moses Kotane sub-district was more geographically concentrated, with less significant distances between population centers and RMP activities.
Second, the Moses Kotane sub-district was more accessible from the South Africa project headquarters office, meaning that travel expenses from Pretoria to Naledi (e.g. per diems and accommodation) were relevant to Naledi only. Third, the Moses Kotane sub-district was dominated by the mining industry, which impacted both population and migration. In this regard, several RMP activities took place at mining locations which resulted in higher numbers of participants and greater uptake in HCT per wellness day. Fourth, study staff at the field level observed that social, political and community level support for the RMP intervention diverged across communities, with greater community-level support for the intervention in the Moses Kotane subdistrict. This was reflected in the number of communities engaged by the intervention (10 in the Naledi sub-district as compared to 22 in the Moses Kotane sub-district).
Finally, the intervention was initially rolled out at the Moses Kotane sub-district and . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint then moved to the Naledi sub-district two weeks later, which may have marginally contributed to differences in outputs across sub-districts.

Policy Implications and Context
These findings are in keeping with key results from related literature. In comparable settings, combination HIV/AIDS interventions have been shown to focus resources on personnel and medical supplies, as well as transport and basic equipment, to facilitate However, given the lack of alternative resources in rural areas, combined with potentially high levels of detected and undetected HIV/AIDS prevalence, the case for continuing to provide services despite reduced cost-efficiency may be made on a health outcome, accessibility, and equity basis.

Costing Perspective
The findings presented in this report represent a highly inclusive costing process, which encompassed all possible costs related to each component of the RMP intervention. Such holistic cost analyses, though still exclusive of patient costs, should be borne in mind when interpreting these results. Our approach is best represented by the health system perspective, which encompasses direct and indirect costs as well as costs borne by local headquarters offices and subcontracting support organizations. Other studies which present findings on the cost of, for example, HIV/AIDS counseling and testing from the provider perspective (with a focus on personnel time and medical and non-medical supplies) may therefore not be directly comparable in this regard (Ekwueme et al 2003).

Limitations
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint There were a number of limitations to our analysis. Our review focused on a limited time period (Years 1 and 2 of the intervention). The possibility that Year 1 costs were not representative of usual operational costs due to start-up issues at this stage should therefore be considered. Our results also did not consider health outcomes beyond the number of HIV-positive patients identified through the wellness day process. The combination of cost and efficiency results with such results in future studies may help to further inform resource allocation and efficiency decisions.

Locally-Informed Resource Allocation Decisions
This analysis of the costs and outcomes of the RMP combination HIV/AIDS prevention program may help to inform associated resource allocation decisions and strategic planning at the in-country level. Our approach provided necessary data around activity-based costs, so that, in the country ownership context, the South African government could make necessary decisions related to efficiencies and effectiveness particularly in the context of changing proportions of funding investments between international and local management offices over time.
Awareness of the key costs and efficiency findings of RMP may therefore have helped in-country actors such as the South African Department of Health to better anticipate associated resource requirements. was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted February 27, 2019. ; https://doi.org/10.1101/562926 doi: bioRxiv preprint