How much does community-based targeting of the ultra-poor in the health sector cost? Novel evidence from Burkina Faso

Universal Health Coverage (UHC) aspires to provide affordable and quality health care to all [1]. However, many low and middle income countries (LMICs) continue to rely on user fees which severely limits access to care and use of health services, especially for the poorest segments in a society [2]. In order to protect the most vulnerable people from financial hardship, some countries have opted for exempting the poorest from user fees [3] or for subsidizing health insurance premiums [4]. The identification and selection of the poor for user-fee exemption or insurance premium subsidization has been proven to be very challenging [5].

In the absence of universal criteria to define poverty and identify the poorest, countries have traditionally adopted various targeting methods. In the literature, the identification of beneficiaries is presented as the process of selecting the persons that benefit from a certain intervention. Targeting, on the other hand, is commonly used as a broader term which includes the identification as well as the actual act of allocating scarce resources to the poorest with the aim of achieving certain policy objectives to preserve or improve health equity [6].

In the health sector, most targeting experiences derive from those with the greatest health risks or the highest ability to benefit from the programs [7]. Those are generally the poorest [8] but can also be other groups of people such as the elderly, orphans or people with disabilities [9].

Generally, high-income countries use a form of means-testing and identify ultra-poor based on a certain income threshold. On the contrary, targeting strategies without a direct observation of income, such as proxy means testing (PMT) and community targeting, are predominantly used in LMICs [6]. While PMT uses a formal algorithm to identify households’ wealth, community based targeting (CBT) collects information from local leaders [10].

All targeting mechanism are prone to errors and generate costs [11]. Various studies have shown that there is no best solution to identify beneficiaries for targeting health benefits. Relative accuracy and cost effectiveness are said to be best achieved by a well-designed and implemented mechanism [9]. However, the scientific community continues to debate the relative costs and benefits of universal vs. targeted strategies in allocating resources to the poorest and alleviating poverty [12‐14]. The question is whether the costs associated with the CBT program actually compensate for the savings accrued when offering subsidized or free health services only to a selected number of individuals.

These arguments are mostly postulated in the light of theoretical assumptions on the costs and benefits associated with targeting procedures, but are not substantiated by any extensive empirical evidence [15]. The literature suggests that CBT can potentially contribute to a reduction in administrative costs compared to other targeting methods [8, 16, 17], since the bulk of resources, such as time and expertise, are provided by community members who are usually not remunerated. Prior studies [18, 19] addressed the costs of pro-poor targeting programs briefly by focusing on the financial expenses and showed that engaging the poor to identify the poorest is not costly. However, both studies disregarded, amongst other components, the time volunteers (e.g. community members) invested into the targeting exercise. Prior studies which addressed the effectiveness of pro-poor targeting showed inconsistent results [20‐23]. These evaluations have focused primarily on estimating how effective programs are in relation to how well all members of a certain target group are included, and members of the non-target group excluded from participating in a targeted program [20‐22]. The available evidence, however, is almost completely silent as to whether targeting policies for the poor offer good value for money, as measured in terms the relationship between their costs and their benefits.

Towards this ends, our economic evaluation began to fill an important gap in knowledge by assessing both the financial and the economic costs associated with designing and implementing a pro-poor targeting project, with specific focus on the identification and targeting component integrated within the Performance Based Financing (PBF) intervention recently rolled out in Burkina Faso [24]. We plan to include these first cost estimates into more comprehensive cost-effectiveness analysis of pro-poor targeting policies in the future.

Table 2 presents the financial and economic costs of the CBT intervention across activity clusters by phase. The estimated financial costs accounted for USD 587,510 and represent about 48% of the total economic value. The total economic cost of the CBT intervention was USD 1,213,447 The selection process represents the most relevant economic cost component. With an estimated economic value of USD 392,060 the selection of the ultra-poor carried out by community members represents one third of the total value of the intervention, followed by the data collection step with an estimated value of USD 328,958.

With a total of 102.609 identified ultra-poor persons, the financial costs amounted to USD 5,73 and the total economic costs to USD 11,83 per identified ultra-poor person.

This study is first of its kind to accurately trace both the financial and economic costs of CBT. By assessing the overall economic value of a CBT intervention, our study makes a unique contribution to the very limited literature available on the cost of targeting in general, both within and beyond the health sector. The lack of comparable empirical literature seems to be due to the fact that the budget statements of targeting interventions in LMICs do not generally indicate costs by process or activity [11]. Lack of comparable studies limits our ability to discuss our results extensively in light of prior evidence, especially when considering the economic value of the intervention. Nevertheless, when relevant, we do appraise our findings in relation to the two published studies that assessed the financial costs of CBT [18, 19] and to the one unpublished report [11] which also accounted for the value of the time contributed by uncompensated community members.

A unique trait of our study is its reliance on ABC, a conceptual approach identified as the preferred one to cost complex health interventions, also in LMICs [35]. The application of ABC allowed us to estimate both the financial and economic costs of the CBT intervention per activity in a highly systematic and accurate manner. This high level of accuracy was ensured by the fact that instead of relying on financial statements as our primary source of information, we started our work by tracing all activities that made up the intervention and all resources consumed by these activities, looking for corresponding financial values only afterwards. Furthermore, combining ABC with Resource Consumption Accounting [36] allowed us to trace the single cost categories within each activity, generating a very detailed map for policy makers of what activities and what items within activities drove the intervention costs. It needs to be noted, however, that the application of this methodological approach is usually not inexpensive [31]. In our specific case, we could rely on ABC costing only because our access to all implementation documents was facilitated by our team’s close relation with the implementation teams and its development partners.

We estimated a financial cost of USD 587,510 and an economic cost of USD 1,213,447 for the entire CBT intervention, including its design and implementation phase. With a total of 102,609 identified ultra-poor, this corresponds to a financial cost of USD 5.73 and an economic cost of USD 11.83 per ultra-poor person identified. As demonstrated in Additional files 2, 3 and 4, our results were neither highly sensitive to variation of the applied wage nor to changes in the percentage of overhead costs or the budget provided by SERSAP. The consistency of our results from primary analysis with those of the sensitivity analysis suggests a good robustness of our findings.

Our study estimated financial costs per ultra-poor person identified to be within the same range of what was estimated in the few prior comparable studies. In particular, Ridde et al. [18], traced the financial expenses of a CBT action research project in one district of Burkina Faso and calculated financial cost of USD 10.16 per identified person. The financial costs Aryeetey et al. [19] calculated for a Participatory Wealth Ranking exercise in Ghana were calculated to be USD 3.83 per identified person in an urban setting, USD 10.63 for a rural setting and 2.71 for a semi-urban setting (calculation: survey costs without exemption premium divided by the total number of people exempted). Compared to the study of Watkins [11], who calculated financial cost of approximately USD 25 per recipient, our estimate is much lower. Differences are likely to be attributable to different implementation strategies as well to differences in methodological pathways to cost estimation.

Additionally, the study draws attention to the large discrepancy between financial and economic cost.Similarly to Watkins [11], we found the difference to be driven by the valuation of human resources, specifically the fact that we valued the time contributed towards the initiative by community volunteers. The heavy reliance on communities as key players in the implementation of the targeting intervention is well aligned with the push towards decentralization and community involvement, which has prevailed across sub-Saharan Africa following the Bamako Initiative [37]. Still this high reliance on communities as actual implementing agents raises important questions in terms of acceptability and feasibility. On the one hand, targeting of health services based on community-based approaches is likely to result in a more accurate and more acceptable identification strategy than one based on experts’ (i.e. healthcare providers) judgment [9, 16, 28, 38] and is likely to be more attractive in the eyes of implementing agencies given its relatively low financial costs [11]. On the other hand, our study indicates that the time committed by volunteers has a clear opportunity cost, measured in terms of their foregone economic value. This observation questions the adequacy of executing CBT by adding an additional burden to communities already operating in resource-constrained settings and often struggling to meet their most basic needs [16, 38]. The challenge is that of relying on the expertise of communities, while at the same time offering adequate compensation for their engagement [39]. This is meant to sustain their motivation in contributing to such activities and avoid withdrawal during the implementation phase, as observed in some communities during the intervention described in our study [29].

Looking at the distribution of costs across activities provides additional evidence that financial and economic valuations do not necessarily align, as we see that the actual identification of the ultra-poor (the core activity delegated to communities) accounted for less than one sixth of all financial costs, but for one third of all economic costs. Moreover, the high cost associated with preparing for and producing identification cards may be surprising for those unfamiliar with the intervention, since the two activities combined accounted for nearly 50% of all financial costs. While we have been reassured by key stakeholders (personal communication) that the cost of card production was much lower in Vietnam compared to what it would have been in Burkina Faso, our appraisal of this specific set of findings inevitably questions the need for sophisticated individual digitalized cards vs. simpler household traditional paper-based cards. It is beyond the scope of our work to assess the comparative advantage of a system based on digitalized vs. traditional paper-based identification cards, albeit we do imagine this to be a relevant area for further research, estimating cost-effectiveness of one system compared to the other. Likewise, the government and its partners, in Burkina as elsewhere, could test the possibility of lowering CBT costs by integrating activities related to the database establishment into the initial identification effort.

Furthermore, when assessing both the financial and the economic value of targeting interventions, one must consider the extent to which these are driven by national governments or by development partners [40]. In the specific case analyzed in our study, the largest portion of the financial costs (60%), excluding the card production that was paid directly by the WB, were accrued by a local for-profit organization, SERSAP, which in turn was contracted by the governmental agency PADS. In practice, however, the resources deployed for the identification and selection procedure originated in a grant made by the HRITF to the government of Burkina Faso, channeled via the PADS. This explicit donor involvement sheds yet a different light on the appraisal of the financial costs associated with targeting programs. Spending close to USD 6 per ultra-poor person identified may be affordable for international agencies specifically committed to fostering greater equity in access, but is likely to be unaffordable for most LMICs, considering overall low levels of public spending on health [41]. Specifically, without aiming at conducting a full budget impact analysis – a task beyond our initial study objectives – it is worthwhile to appraise financial costs in relation to the overall health budget of Burkina Faso. The USD 5.73 per ultra-poor person identified are equivalent to 21% of Burkina Faso’s per capita government expenditure on health (USD 27 (in PPP International USD, year: 2015)) [42]. At the same time, however, one must consider that targeting approaches are implemented beyond the health sector alone and hence there may be potential in sharing costs for pro-poor identification and selection procedures across sectors [43, 44].

Universal Health Coverage can only be achieved by ensuring that the poorest are not left behind. Thus, it is necessary that once implemented, measures to identify and address health inequity are systematically evaluated. With our study, we set the first example of how to systematically assess both the financial and economic costs of a community-based targeting program in a LMIC setting. We view the contribution of this costing study as meaningful, particularly because pro-poor targeting programs continue to expand, within and beyond the health sector. Future research should not only replicate this approach in other settings to generate additional and comparable evidence to better inform policy, but should also reach beyond the mere estimation of the costs to assess the costs of targeting in relation to the benefits accrued, within the framework of a comprehensive economic evaluation, such as a cost-effectiveness analysis.