A

BhsuMGbRbtfabCa©

K

1

wiH

0

Health Policy 78 (2006) 149–156

Priority setting of public spending in developing countries:Do not try to do everything for everybody

Rob Baltussen ∗

Institute for Medical Technology Assessment, Erasmus Medical Center Rotterdam, PO Box 1738, 3000DR Rotterdam, The Netherlands

bstract

ackground: Public spending on health care in many developing countries falls short to provide a comprehensive set of essentialealth services, which indicates the need to target and prioritize resources. However, governments often attempt to provide freeervices to the whole population, and often spend resources on low-impact services. This results in an inequitable and inefficientse of resources.ethods: This paper presents a rational approach to targeting and prioritization of public spending, with an application tohana. First, interventions were tested against the economic justification for public funding, to define to whom spending shoulde targeted. Second, resulting interventions were prioritized on the basis of medical and non-medical criteria.esults: The step-wise approach led to a rank ordering of interventions with a specification whether public spending shoulde targeted at the whole population or the poor only. Disease control priorities are prevention of mother-to-child HIV/AIDSransmission and oral rehydration therapy to treat diarrhea in childhood, and public funding of these interventions is warrantedor the whole population. Case-management of pneumonia in childhood is also a priority but public funding should be targetedt the poor only. Low priorities for public funding are certain interventions to control blood pressure, tobacco and alcohol abuse,

e it for the whole population or the poor only.onclusion: Governments should not try to provide everything for everybody. This may help health systems to move towardsmore equitable and efficient use of resources.2005 Elsevier Ireland Ltd. All rights reserved.

eloping

mfa

eywords: Priority setting; Public finance; Economic analysis; Dev

. Introduction

Different technical approaches have been put for-

ard in disease control in developing countries, rang-

ng from vertical programs such as the Global Fund forIV/AIDS, Malaria and Tuberculosis [1], to invest-

∗ Tel.: +31 10 4082821; fax: +31 10 4089081.E-mail address: r.baltussen@erasmusmc.nl.

ocv[

l

168-8510/$ – see front matter © 2005 Elsevier Ireland Ltd. All rights resedoi:10.1016/j.healthpol.2005.10.006

countries

ents in health systems [2,3]. Although solutions dif-er, what unites them – while there has been no explicitcknowledgement of it – is the need to target and pri-ritize public spending. This is especially relevant inountries where funding levels fall much short to pro-

ide a comprehensive set of essential health services4].

However, governments choose to finance a mucharger share of health services than would be narrowly

rved.

1 Policy

jpoohcuaidSboiectaog

wb[bib

tatatppio

da(twdGMtgfi((

50 R. Baltussen / Health

ustified by economic criteria and often attempt torovide free services for the whole population. Butffering free care to all typically leads to some formf rationing in which better-situated populations oftenave an advantage. Resources may be excessively con-entrated in urban facilities serving the middle andpper classes. The poor, especially those in rural areas,re left with low-quality public services [5–7]. Benefit-ncidence surveys detect persistent anti-poor bias in theistribution of public subsidy in health care in sub-aharan Africa [8]. In addition, resources are ofteneing spent on low impact services such as curative caref non-catastrophic illness. For example, while stud-es show that investing in primary health care is moreffective than investing in specialized health care, allo-ations to primary care in Ghana have remained behindhose allocated to tertiary care [9]. These inequitiesnd inefficiencies indicate the need for a careful usef scarce public resources in health, which could beuided by a more rational approach [10–12].

Developing countries have increasing experienceith explicit rationing of health care, as manifestedy studies in East Africa, Northern Africa and India

13–15]. Typically, these processes are based on a com-ination of cost-effectiveness and burden of diseasenformation, but recently many other criteria have alsoeen put forward, including medical criteria (such as

ivso

Fig. 1. Step-wise approach to targe

78 (2006) 149–156

reatment effectiveness) and non-medical criteria (suchs patient age) [16,17]. Yet, the relative importance ofhese criteria has not yet been determined in a way thatllows a rank ordering of interventions incorporatinghese different considerations. In addition, research onriority setting generally assumes some form of com-rehensive public funding of all health care and therebygnores the more fundamental discussion on the criteriaf state intervention per se.

In response to these shortcomings, this paper intro-uces a step-wise rationale approach to targetingnd prioritization of public spending on interventionsFig. 1). It combines insights from the economic jus-ification of public funding of interventions to defineho should be targeted (step 1), with empirical evi-ence on the prioritization of those interventions fromhana (step 2). It thereby builds on earlier work byusgrove [18]. It argues that, rather than achieving lit-

le in a vain attempt to provide everything for everyone,overnments could provide significant health bene-ts to a large number of people by a better targeting“do not always target everybody”) and prioritization“do not always provide everything”) of its public

nvestments. The result is a rank ordering of inter-entions with a specification whether public spendinghould be targeted at the whole population or the poornly.

t public spending in health.

Policy

od2Ui

2

dohe

ctwaapsfscnitettpfibc

fieippoepm

t

peietms

gaiiip

3

fqpalach

-

-

the same size improvement in health for a less severecondition [25].

- For an equal total effect size, societies may favorinterventions with a large health impact on a few

R. Baltussen / Health

The rational approach is applied to Ghana, a countryf extreme resource-scarcity. Per capita public expen-iture on health in Ghana equalled US$7 per year in004 [19], which is less than a fifth than the requiredS$36 to provide a package of essential health services

n a developing country [4].

. Targeting of public spending

The first step defines to whom public funding ofifferent interventions should be targeted, and is basedn the economic justifications of public spending in theealth sector [18] as derived from mainstream publicconomic theory [20].

Because of the presence of market failures in healthare, there are some important health-related activitieshat the public sector must fund for, potentially, thehole population if they are to be provided at all or

t socially optimal levels. These relate to public goodsnd externalities. Public goods are activities that therivate sector will not undertake, or will undertake atub-optimal levels, because users cannot be chargedor them. Many public health interventions, such aspraying for malaria control and health informationampaigns, are usually considered public goods. Exter-alities are effects to others than those directly receiv-ng the services. For example, an individual who isreated successfully for a sexually transmittable dis-ase derives benefits from that treatment, but so doeshe community at large as the risks of transmissiono others are reduced. An individual’s willingness toay for these services may not fully reflect the bene-ts they generate for society, and public support maye necessary to achieve the optimal level of STDontrol.

In addition, policy makers may also want to publiclynance health for reasons of horizontal equity, to securequal treatment for equal need [21]. For example, hor-zontal equity is achieved when all malaria patients,oor and non-poor, receive equal treatment. However,oor patients may not be able to afford the treatment ifut-of-pocket payments are required. Thus horizontalquity considerations strongly indicate that fees for the

oor should be subsidized, so to help equalize treat-ent.In summary, public subsidies may potentially target

he whole population (including the poor and the non-ag

78 (2006) 149–156 151

oor1) in case interventions constitute public goods orntail large externalities. If this is not case, public fund-ng of interventions may still be warranted to guaranteequal access to care and should in that case be targetedo the poor. Some other interventions, for example treat-

ent of non-communicable disease of the non-poor,hould not be publicly subsidized.

Note that all criteria for public spending are wellrounded in the theory of public economics, but requiresubjective assessment of its importance, and hence of

ts need for public funding. For example, passive smok-ng can be considered as an externality of smoking, butts health effects may be regarded as too minor to justifyublic subsidy of smoking cessation interventions.

. Prioritization of public spending

If interventions qualify for public spending – be itor the whole population or the poor only – the nextuestion is how they should be prioritized. It is oftenroposed to do this on the basis of cost-effectiveness,s this would generate the largest health gains at popu-ation level for the available budget [22]. However, it islso acknowledged that cost-effectiveness is only oneriterion for priority setting, and many other criteriaave been proposed and debated [15–18].

Societies may want to give preferential treatment todisadvantaged populations because they are in somemoral sense more deserving of health resources thanothers [23]. In general, it is argued that the poor have agreater need for support than less poor sections of thecommunity, due to their lower income and typicallylower ‘stock’ of health [24].Societies may favor interventions that target severelyill patients on the basis of their greater need for healthcare, and the diminishing marginal utility of health:an improvement in health from a severe health con-dition is then valued more highly by individuals than

1 In this paper, we refer to horizontal equity as guaranteeing equalccess for poor and non-poor people, although it may also refer touarantee equal access for people in, e.g. remote versus central areas.

1 Policy

-

-

tt

4G

wt‘rtTss

jtftlcusqbpeativdc

dTpt

ttrhowcttiiitp‘d‘Aas

spTliidpoverty, target severe disease, or target the younghave a higher probability of being chosen thanothers.

52 R. Baltussen / Health

individuals compared to those but with a small healthimpact on many individuals since the former has agreater capacity to reduce health inequalities [26].Societies may also have age-preferences becauseof ethical considerations (“disadvantage the oldbecause they have had their fair innings”) or eco-nomic considerations (“advantage the adults becausethey are more productive”) [27].Societies may favor interventions with a relativelylow overall budget impact because of affordabilityconsiderations.

It is suggested that interventions should be priori-ized by taking into account the relevant criteria andheir relative importance [15–17].

. Application of the step-wise approach tohana

As a starting point for the application of the step-ise approach to Ghana, we considered the interven-

ions as put forward in the World Health Report 2002Reducing risks, promoting healthy life’ [28] whicheflect strategies to reduce risks and address an impor-ant part of the burden of disease in Ghana (Table 1).o answer the question which of these interventionshould be publicly financed and for whom, the twoteps are discussed in turn.

In the first step, interventions were tested against theustification criteria for public spending. Table 1 listshe interventions and indicates the presence of marketailures that prevent the provision of these interven-ions at socially optimal levels. If interventions entailarge externalities or are public goods, the table indi-ates that public subsidies may target the whole pop-lation. For example, population-based interventionsuch as fortification and law enforcement programsualify for public spending for the whole populationecause they constitute public goods, and would not berovided without public funding. Other interventions,.g. in HIV/AIDS and water and sanitation programs,lso qualify to be subsidized for the whole popula-ion because they entail externalities in the sense they

nterrupt disease transmission. A range of other inter-entions, mainly individual-based and clinical serviceso not entail market failures but may nevertheless beandidate for public spending if society wishes to subsi-

Fm

78 (2006) 149–156

ize these services for the poor to ensure equal access.he result is a typology of interventions eligible forublic funding targeting the whole population or onlyhe poor.

In the second step, we prioritized the interven-ions on the basis of medical and non-medical criteriahrough a discrete choice experiment (DCE). In a DCE,espondents choose their preferred option from sets ofypothetical interventions, each consisting of a bundlef criteria that describe the intervention in question,ith each criterion varying over a range of levels. The

riteria are constant in each scenario, but the levelshat describe each criterion may vary across interven-ions. Analysis of the options chosen by respondentsn each set reveals the extent to which each criterions important to the decision at hand [29]. The DCEncluded a number of criteria that were selected onhe basis of a literature review and discussions witholicy makers in Ghana, and were referred to above:cost-effectiveness’, ‘poverty reduction’, ‘severity ofisease’, ‘age of target group’, ‘budget impact’ andindividual health effect’. All criteria had two levels.n example of a pair of interventions is given in Fig. 2,

nd 12 pairs of interventions were included in theurvey.

The DCE survey was administered during a work-hop in Ghana, including 30 health policy makers oreople otherwise involved in decision making in health.he results were analyzed using a random effects

ogistic regression model. Four criteria had significantmpact on the choice of intervention (as indicated byts regression coefficients) with its sign in the expectedirection. Interventions that are cost-effective, reduce

ig. 2. Example of pair of interventions in discrete choice experi-ent.

R. Baltussen / Health Policy 78 (2006) 149–156 153

Table 1Targeting and priority setting of public spending

Interventions Targeting of public spending (step 1)a Priority settingof publicspending (step2), compositeleague table

Market failures To whom

Presence Type

HIV/AIDS: prevention of mother to child transmission Yes Externalities Whole population 1Childhood undernutrition: case management of pneumonia No – Poor 1Childhood undernutrition: oral rehydration therapy for diarrhoea Yes Externalities Whole population 1Childhood undernutrition: vitamin A fortification of staple food Yes Public good Whole population 4Childhood undernutrition: vitamin A supplementation No – Poor 4Childhood undernutrition: zinc fortification of staple food Yes Public good Whole population 4Childhood undernutrition: zinc supplementation No – Poor 4Disinfection at point of use Yes Externalities Whole population 4Improved water supply and sanitation, low technologies Yes Public good/

ExternalitiesWhole population 4

HIV/AIDS: mass media campaigns Yes Public good/Externalities

Whole population 10

HIV/AIDS: voluntary counseling and testing Yes Externalities Whole population 10Iron deficiency: iron supplementation No – Poor 12Childhood undernutrition: improved complementary feeding No – Poor 13Improved water supply and sanitation, high technologies Yes Public good/

ExternalitiesWhole population 13

Smoking: brief advice to stop smoking No – Poor 15Alcohol abuse: complete advertising ban Yes Public good Whole population 15Tobacco: excise tax Yes Public good Whole population 17Tobacco: comprehensive advertise banning Yes Public good Whole population 18Safe practice: decreased reuse of injection equipment w/o

sterilizationYes Public good/

ExternalitiesWhole population 18

Safe practice: decreased unnecessary use of injections Yes Public good/Externalities

Whole population 18

HIV/AIDS: antiretroviral drugs Yes Externalities Whole population 18Alcohol: brief advice to stop drinking No – Poor 18Alcohol: reduced sale hours Yes Public good Whole population 18Cholesterol: population-wide health education through mass media Yes Public good Whole population 18Cholesterol: individual-based treatment and education. No – Poor 18Blood pressure: population wide salt reductions. Yes Public good Whole population 18Smoking: nicotine replacement therapy No – Poor 27Tobacco: clean indoor air law enforcement Yes Public good Whole population 27Tobacco: nicotine replacement therapy No – Poor 27Alcohol: drink drive legislation and enforcement Yes Public good Whole population 27Alcohol: excise taxes Yes Public good Whole population 27Blood pressure: individual-based hypertension treatment and

educationNo – Poor 27

for puba

pstot

ac

a The presence of catastrophic costs is also a justification criterionntiretroviral drugs in HIV/AIDS control.

On the basis of these results, we computed a com-osite index that represents the relative priority of the

et of interventions as a function of their characteris-ics. We mapped the characteristics of the interventionsn the levels of the various criteria, and consideredhe regression coefficients of the particular levels of

woiI

lic spending, but is not included in the table. It could apply to, e.g.

ll criteria from the DCE survey as weights. We thenomputed a composite index score as the sum of the

eights of all criteria levels of an intervention. A rankrdering of all intervention on the basis of this compos-te index results in a composite league table (Table 1).t shows that the interventions with the highest priority

1 Policy

atcvat[

leHtttaacap

5

tatmmGsfiwntcaw

airceihet

vn

hftoent(

iasboHfitepnw

atspwdtpepscf[

mapi

54 R. Baltussen / Health

re prevention of HIV/AIDS transmission from mothero child, and those that treat pneumonia and diarrhea inhildhood. Low priority interventions are certain inter-entions to control blood pressure, tobacco and alcoholbuse. More detail on the DCE survey and the deriva-ion of the composite league table is provided elsewhere30].

Combining the results of the two analytical stepseads to an interesting typology of interventions. Dis-ase control priorities are prevention of mother-to-childIV/AIDS transmission and oral rehydration therapy

o treat diarrhea in childhood, and public funding ofhese interventions is warranted for the whole popula-ion. Case-management of pneumonia in childhood islso a priority but public funding should be targetedt the poor only. Low priorities for public funding areertain interventions to control blood pressure, tobaccond alcohol abuse, be it for the whole population or theoor only.

. Discussion

Public spending in developing countries falls shorto provide free services for the whole populationnd the public sector should limit itself to its coreasks. A rational approach as presented in this paper

ay help governments to achieve best value for theironey and from international initiatives such as thelobal Fund for HIV/AIDS, Malaria and Tuberculo-

is. The approach rank orders interventions and speci-es whether public spending should be targeted at thehole population or only the poor. This underpins theotion that the public sector should not try to do every-hing for everybody. This may inform decisions on thehoice of interventions when more resources becomevailable, but also to the reallocation of resourcesithin the existing budget [31].The approach outlined in this paper is hierarchic:

policy maker should first ask whether public fund-ng is justified and to whom and should subsequentlyank order the intervention to set priorities. The hierar-hy is essential: only because an intervention is cost-ffective, it does not necessarily mean that public fund-

ng should target the whole population. An exampleere is case management of pneumonia which is cost-ffective but where public funding should be targeted athe poor only. In the absence of a justification to inter-

ifior

78 (2006) 149–156

ene, such services should not be subsidized for theon-poor.

The approach uses two different notions of equity:orizontal equity is defined as a justification criterionor public spending, whereas vertical equity is used inhe weighing procedure to prioritize interventions. Inther words, poverty may be a reason to intervene tonsure that the poor receive equal treatment for equaleed (horizontal equity), and policy makers may wanto go further and give preferential treatment to the poorvertical equity).

Another important rationale for public interventionn health financing, but not included in the two-steppproach, relates to the cost of interventions. Someervices carry catastrophic costs that cannot be paidy someone who is non-poor, without it making himr her poor, e.g. the cost of antiretroviral treatment inIV/AIDS. This is often regarded as a reason for publicnance but is really a reason for insurance, which can

ake the form of private or public insurance [18]. How-ver, insurance schemes usually cover a minority of theopulation in many poor countries, and if insurance isot feasible, public financing of such health services isarranted.The results of this study are not directly generaliz-

ble to other settings, e.g. other countries: the criteriahat have been identified as being relevant to priorityetting in Ghana may not be relevant to other countries,olicy makers in other countries may attach differenteights to those criteria, and interventions may haveifferent characteristics in different countries, e.g. inerms of cost-effectiveness, or their ability to reduceoverty. However, the methodological approach is gen-ralizable, and has, e.g. now also been used to guideriority setting in Nepal. The study led to the inclu-ion of similar criteria as in the present study, but theriteria ‘age of target group’ and ‘health effects’ wereound to be more important than ‘cost-effectiveness’32].

This study may not lead to direct policy recom-endations in Ghana. First, the research is explorative

nd the interventions considered only establish a smallart of those potentially applicable to disease controln Ghana. Second, the research should be embedded

n the policy and planning process, e.g. relate to anancial budgeting analysis, to assess the feasibilityf implementing the priority interventions. Third, weecognize that a rational approach to priority setting

Policy

moprlbwb

uitmr

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

R. Baltussen / Health

ay only have limited impact on the eventual choicef interventions, and that past spending patterns orressure of political interests groups influence the wayesources are allocated in practice [10–12]. Neverthe-ess, we believe this information is generally welcomedy policy makers and provide practical suggestions onays in which the public spending in health care coulde better employed.

In conclusion, this paper has proposed a potentiallyseful rationale approach to public spending by argu-ng that governments should not try to provide every-hing for everybody. This may help health systems to

ove towards a more equitable and efficient use ofesources.

eferences

[1] Brugha R, Donoghue M, Starling M, Ndubani P, Ssengooba F,Fernandes B, et al. The global fund: managing great expecta-tions. Lancet 2004;364:95–100.

[2] Travis P, Bennett S, Haines A, Pang T, Bhutta Z,Hyder AA, et al. Overcoming health-systems constraints toachieve the millennium development goals. Lancet 2004;364:900–6.

[3] WHO. The World Health Report 2003. Health Systems: princi-pled integrated care (Chapter 7). http://www.who.int/whr/2003/chapter7/en/ [accessed 28 April 2005].

[4] WHO. Commission on Macroeconomics and Health. Macroe-conomics and health: investing in health for economicdevelopment. Report of the Commission on Macroeco-nomics and Health. Geneva: World Health Organization;2001.

[5] Tsui AO, Wasserheit JN, Haaga JG, editors. Reproductive healthin developing countries: expanding dimensions, building solu-tions. Commission on Behavioral and Social Sciences and Edu-cation; 1997.

[6] Rhodes G. 1978: A good year, but a vintage past its best. HealthPolicy 2004;67:241–4.

[7] Streefland P. Public health care under pressure in sub-SaharanAfrica. Health Policy 2005;71:375–82.

[8] Overseas Development Institute Briefing Paper. Can we attainthe millennium development goals in education and healththrough public expenditure and aid? 2003.

[9] MOH Republic of Ghana Ministry of Health. Health sector five-year programme of work: 1997–2001. Accra, Ghana: Ministryof Health; 1998.

10] Martin DK, Singer PA. Priority setting and health technol-ogy assessment: beyond evidence based medicine and cost-

effectiveness analysis. In: Ham C, Coulter A, editors. The globalchallenge of health care rationing. Buckingham: Open Univer-sity Press; 2000.

11] Segal L, Chen Y. Priority setting models for health. Centre forhealth program evaluation. Working paper 119; 2001.

[

78 (2006) 149–156 155

12] Ham C. Priority setting in health care: learning from interna-tional experience. Health Policy 1997;42:49–66.

13] Bobadilla JL, Cowley P, Musgrove P, Saxenian H. Design,content and financing of an essential national package ofhealth services. Bulletin of the World Health Organization1994;72:653–62.

14] Zergaw A, Mariam DH, Ali M. Cost-effectiveness analysis ofhealth care interventions in Meskaan Mareko Wereda Ethiopia.Ethiopian Journal of Health Development 2002;16:267–76.

15] Kapiriri L, Norheim OF, Heggenhougen K. Using burden ofdisease information for health planning in developing coun-tries: the experience from Uganda. Social Science Medicine2003;56:2433–41.

16] Kapiriri L, Norheim OF. Criteria for priority setting in healthcare in Uganda: exploration of stakeholders’ values. Bulletin ofthe World Health Organization 2004;82:172–9.

17] Kapiriri L, Arnesen T, Norheim OF. Is cost-effectiveness anal-ysis preferred to severity of disease as the main guidingprinciple in priority setting in resource poor settings? Thecase of Uganda. Cost Effective Resource Allocation 2004;2:1.

18] Musgrove P. Public spending on health care: how are differentcriteria related? Health Policy 1999;47:207–23.

19] World Health Organisation Statistical Information Sys-tem. http://www3.who.int/whosis/country/indicators.cfm?country=gha&language=en#economic [accessed 28 April2005].

20] Stiglitz JE. Economics of the public sector. 3rd ed. New York,London: W. W. Norton and Company; 2000.

21] Wagstaff A, van Doorslaer E. Measuring and testing for inequityin the delivery of health care. Journal of Human Resources2000;35:716–33.

22] World Development Report. Investing in Health. World Bank;1993.

23] Daniel Wikler, editor. Fairness and goodness in health. Geneva:World Health Organization; 2004.

24] McIntyre D, Gilson L. Redressing dis-advantage: promotingvertical equity within South Africa. Health Care Analysis2000;8:235–58.

25] Stolk EA, van Donselaar G, Brouwer WB, BusschbachJJ. Reconciliation of economic concerns and health pol-icy: illustration of an equity adjustment procedure usingproportional shortfall. Pharmacoeconomics 2004;22:1097–107.

26] James C, Carrin G, Savedoff W, Hanvoravongchai P. Clarify-ing efficiency-equity tradeoffs through explicit criteria, witha focus on developing countries. Health Care Analysis, inpress.

27] Tsuchiya A. Age-related preferences and age weighting healthbenefits. Social Science Medicine 1999;48:267–76.

28] World Health Report. Reducing risks, promoting healthy life.

Geneva: World Health Organization; 2002.

29] Ryan M, Gerard K. Using discrete choice experiments tovalue health care programmes: current practice and futureresearch reflections. Applied Health Economic and Health Pol-icy 2003;2:1.

1 Policy

[

[

[32] Baltussen R, Ten Asbroek G, Shrestha N, Bhattarai P,

56 R. Baltussen / Health

30] Baltussen R, Stolk E, Chisholm D, Aikins M. Towards a multi-

criteria approach for priority setting: an application to Ghana,2005. Health Economics, in press.

31] Murray CJ, Evans DB, Acharya A, Baltussen RM. Developmentof WHO guidelines on generalized cost-effectiveness analysis.Health Economics 2000;9(3):235–51.

78 (2006) 149–156

Niessen LW. Weighing criteria for policy making: shouldmanagement of lung health be a priority program in Nepal?Institute for Medical Technology Assessment DiscussionPaper.