Going beyond horizontal equity: An analysis of health expenditure allocation across geographic areas in Mozambique

Highlights

• We analyze horizontal and vertical equity in the distribution of public health expenditure.

• We quantify the contributions of service use and resource allocation to inequity.

• We assess equity in government and donor expenditure for 2008–2011 in Mozambique.

• Over time, government and donor expenditure aligned and became more equitable.

Abstract

In contexts where health services are mostly publicly provided and access is still limited, health financing systems require some mechanism for distributing financial resources across geographic areas according to population need. Equity in public health expenditure has been evaluated either by comparing allocations across spending units to equitable shares established using resource allocation formulae, or by using benefit incidence analysis to look at the distribution of expenditure across individual service users. In the latter case, the distribution across individuals has typically not been linked to the mechanisms that determine the allocation across geographic areas, and to the utilization of specific services by individuals.

In this paper, we apply benefit incidence analysis in an innovative way to assess horizontal and vertical equity in the geographic allocation of recurrent expenditure for outpatient health care across districts in Mozambique. We compare the actual distribution of expenditure with horizontal and vertical equity benchmarks, set according to measures of economic status and need for health care. We quantify the observed inequities and the relative contributions of service use and resource allocation. We analyse government and donor expenditure separately and combined, for the years 2008–2011 to compare changes over time and funding source. We use data from a number of national routine sources.

Results show improvements in both horizontal and vertical equity, along with the gradual alignment of government and donor resources over time, which resulted in almost horizontally and vertically equitable resource allocation in 2011. However, inequities in the distribution of expenditure across beneficiaries persisted and were driven by inequities in service use. The discrepancy between economic and need indicators highlighted initial differences in government and donor expenditure targets, raising questions about the purpose of public health expenditure and confirming the importance of clearly defining equity objectives to inform and evaluate resource allocation policies.

Introduction

In contexts where health services are mostly publicly funded, the allocation of resources across levels of care and geographic areas is a key determinant of equitable service provision and access (Kutzin, 2013, WHO, 2010). Governments are ultimately responsible for establishing appropriate mechanisms to guarantee access to health care services and financial protection to citizens. In low and middle-income countries (LMICs) in particular, where access is still limited and resource scarcity is pronounced, distributing public financial resources across geographic areas according to population need, is the mechanism through which essential services are made available (Diderichsen, 2004, Green, 2007).

Equity in the allocation of public financial resources has been assessed using either resource allocation formulae (RAF) or benefit incidence analysis (BIA), two different but complementary analytical approaches. In the literature on LMICs, the separate use of these two approaches has so far provided only limited evidence to policy makers on how they can better allocate their resources to meet the needs of their populations (Anselmi et al., 2014).

RAFs have been developed to provide a normative benchmark for the allocation of financial resources across local health authorities, identified with geographic areas. An equitable share for each area is calculated based on population, adjusted for some indicator of need for healthcare to account for differences in the expected intensity and cost of treatment to be provided (McIntyre et al., 2007, Rice and Smith, 2002). RAFs were originally developed to promote horizontal equity (equal treatment of individuals or groups in the same circumstances) by allocating resources across areas to allow equal care for individuals with the same need (Rice and Smith, 2002, Wagstaff et al., 1991).

In high income settings service utilisation, adjusted for the influences of health care availability and individual characteristics, is used as an indicator of need (Gravelle et al., 2003, Smith, 2008). By contrast, in LMICs, where more important differences in access to services exist across geographic areas, morbidity, mortality, or socio-economic deprivation, are used as proxies for healthcare need. In LMICs, RAFs can contribute to promote vertical equity (differential treatment of individuals or groups in different circumstances) by allocating expenditure proportionally, or even progressively, to need, to accelerate health improvements among the neediest (Diderichsen, 2004, McIntyre et al., 2007, Mooney, 2000, Mooney and Jan, 1997). Although helpful in promoting changes in resource allocation patterns, the RAF approach lacks consideration for how resources allocated to local health administrations ultimately reach the intended beneficiaries (Sheldon and Smith, 2000).

Unlike RAFs, BIA evaluates the distribution of public expenditure across individual beneficiaries by attributing to each individual a monetary benefit, or subsidy, according to the frequency and average cost of the public service used. Individuals are aggregated in subgroups, typically defined according to socio-economic status, to analyse the relative distribution of subsidy across the whole population (Demery, 2000, O'Donnell et al., 2008).

BIA was originally conceived to analyse the potential that public expenditure in social sectors has to redistribute economic resources (Demery, 2000, Van de Walle and Nead, 1995). When applied to the health sector, BIA has therefore been used to explore whether public subsidies reach the poor, but without explicit reference to need for health care (Castro-Leal et al., 2000, O'Donnell et al., 2008, Wagstaff, 2012). The existing BIA studies have implicitly assumed a vertical equity perspective and considered a pro-poor distribution of expenditure desirable, either because from a public finance perspective it redistributes economic resources (Demery, 2000, Van de Walle and Nead, 1995), or because inequalities in health tend to disadvantage the poor (Mahal et al., 2000, O'Donnell et al., 2007, Wagstaff, 2002). However, from a strict health sector perspective, inequalities in health expenditure across individuals with different economic status could be deemed horizontally inequitable. Unless explicitly associated with differences in need for health care, such inequalities would reflect a differential treatment based on individuals' ability to pay, potentially with the same health status.

Although the two are often empirically correlated, poverty refers to lack of economic means, while need for health care (henceforth “need”) refers to lack of health. Additionally, there is no evidence (so far) of a direct effect of economic wealth on health status, other than through access to higher education and improved living condition, which depend on household economic capacity, as much as on expenditure decisions (O'Donnell et al., 2014). Measuring horizontal and vertical equity separately is therefore important to identify the most effective policies to address inequities. A few recent studies have compared the distribution of benefit with the distribution of need, measured by self-assessed health, across socio-economic strata (Akazili et al., 2012, Ataguba and McIntyre, 2012, Chuma et al., 2012, Mtei et al., 2012). However, none of the existing BIA studies has analysed both horizontal and vertical equity. In particular no study has measured vertical equity by ranking individuals according to their health need, nor have they quantified vertical inequity to allow inter-temporal and cross-setting comparisons.

BIA results are driven by the interplay of two separate factors: the expenditure on different services and the individual use of those services (Castro-Leal et al., 2000). Some studies have complemented BIA with an analysis of healthcare utilisation to infer the redistributive implications of funding different levels of care (O'Donnell et al., 2007), and others have accounted for differences in expenditure across provinces (Mahal et al., 2000). However, none of the BIA studies has disaggregated the relative contributions of service use and resource allocation to the observed inequities, nor have they interpreted the observed distribution of benefit across individuals in relation to the allocation of resources across geographic areas. Finally, in spite of international aid constituting on average 16% (up to over 80%) of health expenditure in low income countries (WHO, 2012), no BIA study has separately analysed government and donor allocations to further inform resource allocation decisions.

In this paper we use BIA to analyse equity in the allocation of recurrent expenditure for primary and secondary outpatient care across geographic areas in Mozambique. We depart from the standard BIA practices in two ways. We introduce the distribution of need as an equity benchmark and we integrate analysis of healthcare use into BIA. We make several contributions to the existing literature. First, we evaluate separately horizontal and vertical equity by ranking individuals according to their economic status and their need for healthcare. Second, we account for differences in expenditure across districts and quantify the relative contributions of resource allocation and service use to the existing inequities. Third, we analyse government and donor expenditure, separately and in combination, for a four year period.

The paper is structured as follows. Sections 2 Study setting, 3 Data set the context and describe the data. Section 4 details the methods used. Sections 5 Results, 6 Discussion present and discuss results and section 7 concludes.