HEALTH FINANCING WORKING PAPER No 1

Spending targets for health:

no magic number

Matthew Jowett
Maria Petro Brunal
Gabriela Flores
Jonathan Cylus
WHO/HIS/HGF/HFWorkingPaper/16.1

© World Health Organization 2016

All rights reserved. Publications of the World Health Organization are available on the WHO
website (www.who.int) or can be purchased from WHO Press, World Health Organization,
20 Avenue Appia, 1211 Geneva 27, Switzerland (tel.: +41 22 791 3264; fax: +41 22 791
4857; e-mail: [email protected]).

Requests for permission to reproduce or translate WHO publications –whether for sale or for
non-commercial distribution– should be addressed to WHO Press through the WHO website
(www.who.int/about/licensing/copyright_form/en/index.html).

The designations employed and the presentation of the material in this publication do not
imply the expression of any opinion whatsoever on the part of the World Health Organization
concerning the legal status of any country, territory, city or area or of its authorities, or
concerning the delimitation of its frontiers or boundaries. Dotted and dashed lines on maps
represent approximate border lines for which there may not yet be full agreement.

The mention of specific companies or of certain manufacturers’ products does not imply that
they are endorsed or recommended by the World Health Organization in preference to
others of a similar nature that are not mentioned. Errors and omissions excepted, the names
of proprietary products are distinguished by initial capital letters.

All reasonable precautions have been taken by the World Health Organization to verify the
information contained in this publication. However, the published material is being
distributed without warranty of any kind, either expressed or implied. The responsibility for
the interpretation and use of the material lies with the reader. In no event shall the World
Health Organization be liable for damages arising from its use.

Three of the authors are staff members of the World Health Organization. The authors
alone are responsible for the views expressed in this publication and they do not necessarily
represent the decisions or policies of the World Health Organization.
 Authors and citation
Authors:

Matthew Jowett, Senior Health Financing Specialist, Dept. Health Systems Governance and
Financing, WHO Geneva
Maria Teresa Petro Brunal, Intern, WHO Geneva (currently Health Economist, The Global Fund)
Gabriela Flores, Health Economist, Dept. Health Systems Governance and Financing, WHO Geneva
Jonathan Cylus, Technical Officer, European Observatory on Health Systems and Policies, WHO
Regional Office for Europe

Citation:

Please cite this paper as follows:

Jowett M, Brunal MP, Flores G, Cylus J. Spending targets for health: no magic number. Geneva:
World Health Organization; 2016 (WHO/HIS/HGF/HFWorkingPaper/16.1; Health Financing Working
Paper No. 1); http://apps.who.int/iris/bitstream/10665/250048/1/WHO-HIS-HGF-
HFWorkingPaper-16.1-eng.pdf

Acknowledgements
The idea for this paper emerged from discussions at a meeting on Fiscal space, public finance
management and health financing policy convened by WHO and held in Montreux, Switzerland
between 9-11 December 2014. A first draft of the paper was presented at an internal seminar in WHO
as part of the internship of Maria Petro Brunal, supervised by Matthew Jowett. A more developed
version of the paper was presented at the follow-up meeting Fiscal Space, Public Financial
Management, and Health Financing convened by WHO in Montreux, Switzerland between 26-28 April
2016.

Thanks are due to the following people for useful comments and suggestions on the paper: Rowena
Jacobs (Centre for Health Economics, University of York), Peter Smith (Imperial College Business
School), Rodrigo Moreno-Serra (University of Sheffield), Owen Smith (World Bank), Ajay Tandon
(World Bank), as well numerous colleagues within WHO and participants at the above meetings.

Financial support from the UK Department for International Development (DFID) under the Program
for Improving Countries’ Health Financing Systems to Accelerate Progress towards Universal Health
Coverage, and the Ministry of Health and Welfare of the Republic of Korea under the Tripartite
Program on Strengthening Health Financing Systems for Universal Health Coverage.

For further information please visit:

http://www.who.int/health_financing

2
 SUMMARY

Key observations and policy messages:

● This paper provides evidence which supports the message that all countries can make
progress towards UHC, including those with very low levels of public spending on health
(<40 $ per capita).
● Levels of public spending are central to UHC progress; in terms of service coverage as levels
of public spending increase we observe both systematic improvements in UHC performance
within countries, as well as convergence across countries.
● Whilst financial protection also improves with public spending (especially >$200 per capita),
progress remains highly variable across countries even at >$520 per capita (one of our
spending quintile thresholds). Improving financial protection can be complex, requiring both
the systematic development of policy, and the institutions that govern and manage health
systems.

Motivation and objectives:

● A number of health expenditure targets exist and are widely referred to. These include targets
based on absolute spending amounts, and those based on spending relative to a denominator
such as GDP or total government spending; those based on detailed bottom-up costing and
those without a clear evidence base; and some which clearly refer to public health spending, and
others which imply total health spending.
● Targets send the message to countries that at lower spending levels little or no progress can be
made in terms of service coverage and financial protection, which is clearly not the case given
the considerable variability across countries with similar levels of public expenditure on health.

Methods, indicators and data:

● We use data envelopment analysis (DEA) to assess and compare performance on agreed
indicators for both service coverage and financial protection, relative to a country’s level of
public spending on health in per capita terms.
● We measure performance using five service coverage indicators (DPT3, ART, TB, family
planning, skilled attendance at birth), and one measure of financial protection using public
expenditure on health as % total health expenditure as a proxy measure, given the lack of
widespread data on our preferred indicators.
● The latest validated and published data (2012 or most recent) are analysed for 83 low and
middle-income countries.

Key results:
● We observe high levels of variation across countries in terms of UHC performance at very low
levels of public spending i.e. <PPP$ 40 per capita; some countries achieve a performance less
than half of others with a similar levels of spending.
● UHC performance improves as countries increase public spending on health; convergence in
performance between countries is also observed as spending increases. This convergence is
driven primarily by improvements in service coverage, and occurs rapidly, once countries spend
more than PPP$ 40 per capita.
● In terms of financial protection, significant improvement is observed across our sample of
countries only once public spending is greater than PPP$ 200 per capita; convergence across
countries is not observed, however. Even at higher levels of public spending there remains
significant variation in how well countries translate greater public spending on health into financial
protection for their citizens.
3
 1. More money for health, more health for the money

a) Background
The World Health Report 2010 [1] put forward two central messages; first that countries
need to ensure adequate spending on health to make progress on UHC and, secondly, that
improving spending efficiency is central to the UHC agenda. This perspective has been
reinforced by the adoption in 2015 of both the Sustainable Development Goals (SDGs), and
the Addis Ababa Action Agenda on Financing for Development, which also recognise the
need to explore the nature of the resources available for health systems, and the use to
which they are put, rather than focusing solely on estimates of the level of resources
required to make progress toward UHC. How public resources are used has a direct impact
on both levels of service and financial coverage, as well as how equitably both are
distributed [2].

This paper considers these issues in the context of low and middle-income countries. A
number of estimates of how much countries should spend on health exist, are widely
referred to in policy discussions, and in some cases can play a useful role in advocating for
greater investment in the health sector. However, there is no single or simple answer to this
question [3], and many benchmarks or spending estimates offer little in terms of useful
guidance to country policy makers. Worse still, these estimates may divert policy focus away
from improving the way existing money is being spent. In the analysis which follows we aim
to provide insights for country policy makers by systematically analysing how performance
varies across countries in terms of the two main dimensions of UHC (service coverage and
financial protection), relative to levels of health spending.

b) Health spending targets

Health spending targets widely used in policy discussions concerning low and middle-
income countries are summarised below. A more complete list is provided in Annex 1.

Relative targets: The Abuja Declaration of 2001 recommended that governments allocate
15% of their budgets to the health sector, although the basis for this figure is not clear, with
no explicit connection to achieving a certain level of health system performance. Whilst
focused on the African Region, this target is widely referred to. In 2012 only 14% of
governments in low and lower-middle income countries met the Abuja target; indeed, only
29% of upper-middle income and high-income countries reached this level 1; as a result the
target is rarely considered useful or relevant to country policy makers.

An indicator which is increasingly used, and which builds on the Abuja Declaration target, is
the amount a country spends in terms of public spending on health as a %GDP. This
indicator captures both the priority given to health in budget allocations, as well as the fiscal
context i.e. how large government is relative to the economy, measured in terms of “total
public spending as %GDP”. The World Health Report 2010 noted that “...it is ‘difficult to get
close to universal health coverage at less than 4–5% of GDP’ 2”.

1
Author calculations based on WHO Global Health Expenditure Database (GHED).
2
See Table 5.2 on p98.
4
Subsequent analysis [4] makes a similar assertion saying that “Ensuring financial protection
at an adequate level generally requires GHE 3/GDP of at least 5 per cent. For example, such
a ratio is generally required for limiting the proportion of out-of-pocket payments to 20 per
cent of THE, which in turn is generally needed for achieving low rates of catastrophic and
impoverishing health expenditure.” The explicit link between this spending indicator and
financial protection, a fundamental objective of UHC, is more useful, and refers to previous
analysis [5] which looks at the correlation between a health systems reliance on direct out-
of-pocket payments and levels of catastrophic and impoverishing spending.

Absolute targets: the World Health Report 2010 also presents estimates of required health
spending prepared by the High-Level Taskforce on Innovative International Financing for
Health Systems [6]. The report concluded that low-income countries would need to spend on
average US$60 per capita by 2015 in order to deliver a set of essential health interventions 4,
with the caveat that for some countries the figure would be less than US$40 per capita, and
in others more than US$80 per capita. Subsequently, these estimates were independently
updated to 2012 US dollar terms (from 2005) resulting in an average figure of $86 per capita
[4], which was clearer in explicitly referring to the required level of government or public
health expenditure.

Estimates are not always explicit in referring to public rather than total spending on health,
which is problematic given that public and private revenue sources impact very differently on
how well countries perform in terms of UHC [7]. In 2012, whilst all governments in high and
upper-middle income countries spent at least $86 per capita on health, only 33 or 72% of
lower-middle income countries, and just two low-income countries (Kyrgyzstan 5 and
Rwanda) 6, reached this level.

c) Variation in UHC performance

One downside of these estimates of health spending requirements is that they hide wide
variation in performance across countries. In some cases, policy makers may consider that
unless they reach, or are close to reaching these levels of health spending, they will be
seriously limited in the progress they can make towards UHC. Clearly this is not necessarily
the case; for example public spending on health as %GDP in Thailand today is significantly
less than 4%, and stood at around 2.2% when the Universal Coverage Scheme was
introduced in 2004. Whilst variation across countries spending at similar levels will be due to
a wide range of factors, including many beyond the health system, the way in which health
systems are organised is likely to at least partially explain performance variation [8].

3
In the report the authors use GHE to refer to government health expenditure, equivalent to public spending on
health, and referred elsewhere in this paper as GGHE (general government health expenditure) in line with NHA
terminology. THE refers to total health expenditure.
4
Defined as those services required to increase coverage on MDGs 1, 4, 5, 6, and 8e to 50%. Costs related to
health systems activities or inputs shared across programmes were also estimated.
5
Note that since 2013 following a reclassification of GNI by the World Bank, Kyrgyzstan is now categorised as a
lower-middle income country.
6
Based on the World Bank’s income classification of countries for 2012.
5
Figure 1 further illustrates this point by showing two trends, one for the median level of
GGHE as % of GDP (see left panel) and the other for the median level of OOP as % of THE
(right panel) across the period 1995 to 2012 7. In both panels selected targets are included
as horizontal lines; 5% GGHE/GDP, and 20% OOP/THE. A focus on the median allow us to
show the performance for half of the countries in each year, whereas the average or mean
score can be skewed by one or two countries 8. In the left panel, the median GGHE/GDP
increased in almost all WHO regions (high-income countries are excluded) by almost one
percentage point but still remains significantly below 5%; in other words half of the countries
increased public spending on health relative to the size of their economies, except in the
Eastern Mediterranean and South-East Asia regions.

Figure 1: Trends in public spending on health as %GDP, and direct household contributions
9
relative to total health spending (1995-2012)

60%
6%

50%
5%
Median GGHE/GDP

Median OOP/THE

40%
4%

30%
3%

20%
2%

10%
1%
0%

1995-2012
123456789
1011
113
214
1167 1995-2012
5112123456789
10112
1114
31167 1995-2012
5112123456789
10112
1114
31167 1995-2012
5112123456789
10112
1114
31167 1995-2012
5112123456789
10112
1114
31167 1995-2012
5112123456789
10112
1114
3116
5112
7 1995-2012
123456789
10112
113
14
15
167 1995-2012
112123456789
10112
113
14
15
167 1995-2012
112123456789
1011
113
214
15
16
17 1995-2012
12123456789
1011
12
114
315
16
17 1995-2012
12123456789
1011
12
114
315
16
17 1995-2012
12123456789
1011
12
114
315
16
17
12

AFRO AMRO EMRO EURO SEARO WPRO AFRO AMRO EMRO EURO SEARO WPRO
Notes: 143 countries with information available at least for 18 years Notes: 143 countries with information available at least for 18 years
between 1995-2012 and excluding small states between 1995-2012 and excluding small states

Median government health expenditure Median out-of-pocket spending relative to

relative to GDP total health spending

1
Nb : Both charts excludes high-income countries

Most importantly, the right-hand side figure shows that the median OOP/THE ratio has
decreased in all regions since 1995 except in the WHO European, even in regions where
GGHE/GDP has not increased. Three points are worth making based on these findings:

• first, in most regions 50% of countries reduced their reliance on direct payments by households
despite being far from spending 5% GGHE/GDP
• secondly, there appears to be scope to reduce reliance on private out-of-pocket payments in the
absence of significantly more money for health, as illustrated by the EMRO and SEARO regions
• thirdly, macro analysis of health expenditures has serious limitations, not least understanding
simultaneous changes in levels of utilization of health services, something we address in this
paper

7
Henceforth referred to as GGHE/GDP and OOP/THE.
8
A focus on the mean could be influenced by a country outperforming others over a number of years, for
example in the case of prolonged economic downturn; whilst interesting in terms of how this affects a country’s
commitment to health, it is not the focus of this paper.
9
Only countries with information available for at least 18 of the 19 years between 1995 and 2012 are included.
Countries excluded are Afghanistan (only 12 years of data); Dem. Rep. Congo (11 years); Iraq (11 years); Liberia
(11 years); Somalia (7 years); South Soudan (5 years); Timor Leste (14 years). Small countries are also
excluded, i.e. 13 countries in Africa; 15 countries in the Americas; 1 country in the Eastern Mediterranean region;
3 countries in South East Asia and 10 countries in the Western Pacific; WHO regional classification is adopted.
6
d) The critical role of public revenue sources for progress towards UHC
Evidence shows that for countries to make progress towards UHC their health system needs
to rely predominantly on public revenue sources [7]. By public, we mean those revenue
sources which are prepaid, mandatory and pooled; this includes for example both
government budgetary allocations as well as mandatory contributions to health insurance
schemes, typically in the form of payroll taxes. Recent evidence confirms the importance of
fully and systematically executing public resources. Budget allocations to health reflect
political commitment, but effectively spending those funds the strength of the health system.
In many countries, governments do not fully execute budget allocations for a variety of
reasons, including deficiencies in public financial management [2].

Voluntary or private revenue sources tend to contribute little in terms of helping countries
move their health systems towards UHC, in particular cash payments at the point of service
use, the focus of much political attention in recent years [1, 9]. Voluntary health insurance
schemes, whether commercial for-profit or non-profit community-based schemes, do play a
role in risk-sharing but tend to reach only a small percentage of a country’s population [10];
furthermore, given the nature of these schemes they struggle to maintain financial stability
when faced with a population with high levels of unmet needs, and typically exclude either
those who need care the most, or relatively expensive health services.

2. Indicators, data and methodology

Given the critical role of public revenues for progress on UHC, our analysis focuses
exclusively on the relationship between a country’s level of public spending on health and its
progress in terms of both service coverage and financial protection. We draw on indicators
agreed in the joint UHC monitoring framework [11].

a) Public spending on health

Table 1 shows descriptive statistics for low, middle and high-income countries, for both
absolute and relative levels of public spending on health. GGHE/GDP was 3.66% on
average between 1995-2012 with a standard deviation of 2.2 percentage points; variance
across countries was three times greater than the variance observed within 10 countries over
this period. In contrast public spending on health per capita was an average of $691 with a
standard deviation of $947; here again variation across countries is twice as high across
countries as within countries. The high variation in per capita public spending across
countries is to be expected given that our sample includes countries at very different levels
of economic development.

10
“Across countries” refers to comparison between different countries. “Within countries” refers to change over
time in a single country.
7
 Table 1: Variance decomposition of absolute and relative indicators of public spending on
11
health (1995-2012)

Low, middle &

Excluding high-income & countries with a
high-income
Public spending on health population of less than 1.5 million population
countries
(GGHE)
Std. Median Median Q1 Q1
Mean Std. dev Mean
dev 1995 2012 1995 2012
overall 3.66 2.20 2.86 1.86 2 2.9 1.6 2.2
as a % of GDP between 2.11 1.75
within 0.69 0.71
overall 691.06 947.48 183.48 235.75 47.5 156.73 1.82 51
per capita in US$ between 888.53 207.42
within 335.53 116.24

Once we exclude both high-income and small countries 12, variance across countries reduces
but remains higher across than within countries for both indicators, with a mean barely 3%
for GGHE/GDP, and $183 in per capita terms. Given the greater variation between countries
than within individual countries over time, and the fact that the number of missing values
increases with time series data, we focus on a single year in our analysis. We also focus on
absolute rather than relative levels of spending; whilst countries at different levels of
economic development may have similar levels of GGHE/GDP, their absolute levels of
spending differs considerably 13. Likewise, median GGHE/GDP increased by only one
percentage point, from 2% in 1995 to almost 3% in 2012, a period of 19 years.

In contrast, per capita expenditures have almost quadrupled over this period, with the
median value increasing from $47.5 per capita to $157 per capita, and from $1.82 to $51 per
capita amongst the lowest 25% spenders (quartile 1 or Q1 in Table 1), with a 0.6 percentage
point change in the GGHE/GDP ratio for the latter. This again suggests there is potential for
increased spending in real terms, and hence progress on UHC, even if relative levels of
public spending change little. We use estimates of public spending on health per capita from
2012 expressed in international dollars (Purchasing Power Parity, or PPP) 14, the latest
confirmed and validated data available in WHO’s Global Health Expenditure Database
(GHED).

11
Notes: Author's own calculations based on an initial sample of 143 countries with information available at least
for 18 years between 1995-2012 and excluding small states. The final sample used is 94 countries after
excluding those classified as high income in 2012.
12
See justification in the section on Country Sample.
13
We do, however, repeat some of the analysis using GGHE/GDP. See for example Figures 9 and 10 in Annex
4, which rework Figures 3 and 4 in the paper.
14
National health accounts (NHA) indicators are based on expenditure information collected within an
internationally recognized framework. The per capita general government expenditure on health (GGHE)
indicator is expressed in PPP international dollars to facilitate international comparisons. It includes not just the
resources channelled through government budgets but also the expenditure on health by parastatals, extra-
budgetary entities and notably payroll contributions to compulsory or mandatory health insurance schemes. It
refers to resources collected and pooled by public agencies, and can also be referred to as “public spending on
health”. PPP series generated by the 2011 International comparison project (ICP) estimated by the World Bank
are used; where these are not available PPPs are estimated by the WHO. Population figures are taken from UN
population Division, OECD and EUROSTAT database.
8
b) Indicators of service coverage
A series of tracer indicators to measure country performance in terms of health service
coverage were agreed in the first monitoring report on tracking UHC [11]. Of the eight core
tracer indicators we exclude two, improved water and improved sanitation, given that public
spending on health generally does not pay for these interventions. We also exclude
antenatal care coverage due to major gaps in official data at the country level. The five
indicators used, each of which is measured in terms of the percentage of the target group
covered, are:

● Diphtheria, tetanus and pertussis immunization coverage among 1-year-olds (referred to as

DPT3)
● Family planning coverage with modern methods (referred to as Family Planning)
● Tuberculosis treatment coverage (referred to as TB)
● Antiretroviral therapy coverage (referred to as ART)
● Live births attended by skilled health personnel - doctors, nurses or midwives - (referred to as
SAB)

This set of service coverage indicators focus mostly on priority basic services under the
Millennium Development Goals, and hence are most relevant to low and middle-income
countries. The latest coverage data for each indicator, 2012 or closest year 15, were taken
from WHO’s Global Health Observatory (GHO) data repository 16; when values were missing,
data from World Bank’s World Development Indicator database were used.

c) Indicators of financial protection

Recent estimates of catastrophic and impoverishing levels of health expenditures are only
available for 37 countries. We hence use public expenditure on health as a percentage of
total health expenditure (GGHE/THE) as a proxy measure of financial protection, which is in
line with the message that moving towards a predominant reliance on public spending is
critical for UHC progress [7]. An alternative proxy indicator is out-of-pocket expenditures
(OOP) as a percentage of total health expenditure (THE), or rather 1-OOP as %THE; this
measure excludes, from our estimate of private spending, expenditures on voluntary health
insurance as well as external funding not flowing through government channels, i.e. these
would be included in our estimate of public spending. A sensitivity analysis was conducted
using this indicator (see Annex 2).

A country’s GGHE/THE is, however, only an indirect measure of financial protection as the
consequences of paying for health care on household living standards are not captured.
However, as Figure 2 shows, there is a negative correlation between GGHE/THE and the
incidence of catastrophic expenditures in the 37 countries for which nationally representative
data is available between 2002 and 2012; in other words financial protection improves as
reliance on public spending increases, suggesting that the indicator is a useful and valid
proxy measure.

15
In four countries the latest data used was for 2009.
16
For detailed information on how coverage estimates were calculated for each indicator see Reference 10.
9
 Figure 2: Direct and indirect measures of financial protection

0% 2% 4% 6% 8% 10% 12% 14% 16%

non-subsistence expenditure on health
Proportion of population spending
40% or more of their TJK

TZA

TZA

KHM NIC
GEO
BOL
GEO KEN
KHM
GEO KGZ RUS
MDA RUSMNG
GEO ARG BOL
EGY NICBOL
BIH
GEO
GEO
GEO RWA EST
LAO KGZ IRN UKRKORLVA
SEN BGR
IRNIRN KOR
KORLVA EST
LAO LVA
KOR
LVA BGR LVA EST
EST
KOR
KORUKR TUN BGREST
UKR
ZMB
NER ZMB RWAUKR
UKR BGR GHA BGR ESTFRA
PHL
PHL UKR
UKR EST
EST
PHL PHL TUR TUR
TUR FRA
PAK JOR TUR
TZA BIH PAN MWI
SEN

0% 20% 40% 60% 80% 100%

GGHE as a share of THE

prior to 2005 from 2005 onwards

2
Notes: Cata =.05 + -.04 x GGHE/GDP, R =.16
Source:
World Health Organization/World Bank group (2015)- Survey of 37 countries
Global Health Expenditure Database (GHED)

d) Methods
We use data envelopment analysis (DEA) which is a linear programming method commonly
used to assess efficiency in health systems and many other sectors of the economy [12].
This paper build on methods first used to analyse UHC performance relative to health
spending by Moreno-Serra and Smith [13]. DEA determines how well a decision making unit
(DMU), in our case a country, converts a set of inputs (public expenditure on health per
capita) into a set of outputs (progress on service coverage and financial protection). There
are a number of variants to the DEA approach; we use an output-oriented DEA model with
variable returns to scale17.

DEA assesses each countries performance relative to their level of public health spending
per capita, and then compares this assessment across a sub-sample of countries with
similar levels of inputs. A number of ‘best performing’ countries are identified, and other
countries assessed in relation to these countries, with each given a score out of 100%. For
instance, a country that has the highest achievement on our performance measures given its
level of spending (i.e. all other countries compared with it have lower performance on these
indicators) receives a score of 100%. A peer country with a score of 90% is considered to be
achieving 90% of the outputs achieved in the best performing country.

To calculate individual country scores, each country is assigned its own unique set of output
weights (i.e. weights on service coverage and financial protection indicators) 18; the ratio of
the weighted sum of country outputs divided by inputs is calculated and compared across
countries. We assign the simple constraint that each performance indicator must be given a
weight greater than 0, so that progress on each of the indicators must be taken into account;
17
An input-based approach can also be used; this would analyse variation in inputs for a given level of outputs.
18
Because we use only one input in each model, the input weights are the same for all countries (i.e. equal to 1).
10
otherwise indicators on which countries perform badly could essentially be ignored by the
DEA model. We use the Efficiency Measurement System (EMS) software. For further
discussion on methods see Annex 3.

e) Country sample
In a similar way that we remove progress on water and sanitation from our analysis due to a
lack of clear causal relationship between inputs and outputs, we also remove 55 countries
classified as high-income in 2012 out of the original sample of 191 countries. The reason is
that the indicators we use reflect priority services in low and middle-income countries, on
which many significant coverage challenges remain. In all high-income countries there is
near to full coverage on these indicators, and hence a different set of indicators would be
more appropriate for such analysis.

Given that we conduct a cross-sectional analysis for a single year, we remove countries
classified as small states 19 by the World Bank, due to the fact that both spending and
performance indicators in smaller populations can vary significantly from year to year due to
their greater exposure to international trade, economic shocks and income volatility 20; this
could have a significant impact on the results of a DEA using single year cross-sectional
data. We hence remove a further 35 countries from our analysis 21. However, we also
conduct a sensitivity analysis using averages for a five year period (see Annex 2), which
shows broadly similar results. Missing country values for the inputs or output variables were
also eliminated (18 countries had incomplete data) 22 leaving a final sample of 83 low and
middle-income countries (see Annex 5); see Annex 6 for limitations of the analysis.

3. Results

a) Descriptive analysis
Figure 3 shows descriptive statistics for both service coverage and financial protection, in
which the 83 countries are categorised into quintiles based on their level of per capita public
spending on health: Q1 represents the lowest spending countries and Q5 the highest 23. The
upper panel shows performance for countries, again categorised by spending quintile and
using boxplots, on the five services combined (right side in black), and financial protection
(left side in green); the more compact the box the less dispersion there is across the middle

19
The World Bank defines small countries as those with a population of 1.5 million or less.
20
World Bank (2015). Small States.
21
Forty-one countries in our sample were small states; six were already removed as high-income countries.
22
Angola; D.R. Congo; Hungary; Iraq; Jordan; D.P. Korea; Lebanon; Libya; Papua New Guinea; Somalia; South
Africa; Syrian Arab Republic; Turkmenistan; Uganda; Uzbekistan; Yemen; Zambia and Zimbabwe.
23
Quintile ranges for levels of public spending on health per capita are: Q1 = <$37.5; Q2 = $38.20 to $59.60; Q3
= $60.90 to $198.10; Q4 = $207.60 and $500.30; Q5 = >$520.50.
11
50% of countries 24. Average service coverage increases as countries spend more; variation
across countries also systematically falls i.e. performance converges. In contrast, such clear
convergence is not observed for financial protection; whilst the distribution across countries
becomes less skewed 25, performance remains as dispersed amongst middle quintile
spenders as it does amongst the lowest spenders (quintile 1), and reduces only marginally in
the highest spending countries (quintile 5).

Figure 3: Descriptive statistics (inputs and outputs) by public spending on health quintile17
10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

Q1 Q2 Q3 Q4 Q5

Financial Protection (proxy)

Average all service coverage indicators
Q1 to Q5 denote quintiles of public spending on health, US$(PPP)

Distribution of each service coverage indicator

0% 10% 20% 30% 40% 50% 60% 70% 80% 90%100%

Q1 Q2 Q3 Q4 Q5
TB DPT3
Fam. Planning SAB
ART
Q1 to Q5 denote quintiles of public spending on health, US$(PPP)

24
Boxplots emphasize trends in median performance; the median point is the line which divides boxes into two
parts. Dispersion in performance is shown by the box, the upper limit of which denotes performance at the 75%
quartile, the lower limit the performance at the 25% quartile.
25
Indicated by the median line which moves towards the centre of boxes across quintiles.
12
The data suggests that when low-spending countries start to increase public spending on
health, this translates first into greater use of essential health services, than financial
protection.

The lower panel of Figure 3 show boxplots for each individual service, with the horizontal line
at 80% representing agreed coverage targets. There is relatively good performance across
all spending quintiles in terms of DPT3 and TB coverage, the latter surpassing the 80%
coverage target [11] in half of our sample countries. Again, in half of countries within each of
the spending quintiles, DPT3 coverage rates are above 80% and even above 70% for the
lowest spenders.

What is particularly noticeable in Figure 3 (lower panel) is that increasing levels of public
spending strongly correlate with better performance on SAB, with the median increasing
from barely 40% in the lowest quintile (Q1) to almost 100% in the top quintile. Convergence
in family planning performance also increases with spending, but less so than for SAB. ART
coverage is low across all levels of spending, with little in terms of convergence in
performance to report. These differences in the scope of progress at different levels of public
spending are important for the DEA analysis.

b) Main DEA model

Figure 4 presents the results of the DEA model, with a score for each country which is also
identified in terms of its income-category. The DEA model identifies a number of ‘best
performing’ countries that achieve greater performance relative to their level of spending
than other countries with which they are compared (peer countries); these countries receive
a score of 100%. It is important to note that a score of 100% does not imply these countries
have achieved or are even close to achieving UHC, but simply that they outperform
countries with similar levels of public spending on health.

Figure 4: UHC performance relative to public spending in 83 low and middle-income countries

13
Six “best performer” countries are identified in our analysis: Myanmar, Cambodia, Malawi,
Rwanda, Thailand and Cuba, discussed in more detail later in this paper. Perhaps the most
striking observation from Figure 4 is the very large variation in performance at low levels of
public spending represented by a high dispersion on the y-axis, with increasing convergence
(or lower variation across the y-axis) as public spending increases. We measure this
variation more systematically by looking at the standard deviation of DEA scores within
spending quintiles (see Table 2).

Table 2: Variation in relative performance at low levels of public spending on health

Level of public spending Number Std. dev. of
Quintile Mean DEA score Min DEA score
on health per capita countries DEA scores
1 GGHE less than $37.50 17 14.4% 77.0% 45.5%
GGHE between $38.20
2 17 9.9% 83.0% 59.2%
and $59.60
GGHE between $60.90
3 16 10.7% 86.8% 57.1%
and $198.10
GGHE between $207.60
4 17 6.5% 88.1% 79.1%
and $500.30
GGHE higher than
5 16 5.6% 87.8% 78.3%
$520.50

Those countries in the lowest spending quintile i.e. below $PPP 37.50 per capita public spending
on health have by far the greatest variation in DEA scores at 14.4%, with quintiles 2 and 3
showing much lower variation at around 10%; the lowest variation is in the highest spending
quintiles 4 and 5. UHC performance improves, and variation in performance across countries
falls, as absolute levels of public spending on health increase. We also conduct the analysis
using country income groupings which shows a similar trend (see Table 7 in Annex 4).

In order to understand what drives the observed convergence in UHC progress relative to
public spending on health, we refer back to Figure 3 which shows that this trend is driven
primarily by improvements in service coverage rather than financial protection. Even at
higher levels of public spending i.e. in Q3, Q4 and Q5, whilst financial protection improves
on average, there remains considerable variation across countries in terms of how they
perform.

c) Best-performing countries
The six countries identified as showing the best progress for their given level of per capita
public spending on health are Myanmar, Cambodia, Malawi, Rwanda, Thailand and Cuba.
The latter three are widely acknowledged in the literature as good performers, Cuba since
being identified as a country providing good health at low cost in the 1980s [14], and
Thailand for its expansion of service coverage and range of reforms including the Universal
Coverage Scheme in 2004 [15-17]; both these countries are upper-middle income and score
well across most of the progress indicators. Rwanda is a low-income country, known for its
establishment of almost universal population coverage through highly organised mutuelles
[18]; in the DEA analysis Rwanda does well on most indicators, although worse than Cuba
and Thailand on family planning and skilled attendance at birth coverage, two indicators
which appear to be particularly sensitive to absolute levels of public spending.

The other three countries are low-income, like Rwanda, but spend considerably less in
absolute terms. Cambodia’s introduction of health equity funds led to a reduction in out-of-
14
pocket spending by 26% [19] at the same time as a remarkable increase in the utilization of
child and maternal health services, in part due to reforms such as performance-based
incentive payments to midwives [20] and the use of vouchers to promote maternity care [21].
Whilst Cambodia performs particularly well across most of the service indicators, it still does
poorly on financial protection despite recent improvements. For Malawi and Myanmar
progress is again driven mostly by good service coverage (apart from ART), rather than
financial protection, the latter being a particular challenge in Myanmar.

d) Variation in UHC progress at low levels of public spending

The large variation in performance at low levels of public spending suggests that there is
significant room for improvement in the way that money is spent and services are delivered,
which in turn could impact significantly in terms of UHC. This point is of particular interest
given that the greatest levels of unmet health needs are generally in the poorest countries;
furthermore it supports the assertion that even at levels of public spending significantly
below the widely referred to spending targets summarised at the beginning of this paper,
UHC progress should be possible. We look more closely at countries with low levels of
public spending on health in Figure 5, which presents the main DEA model results for
countries spending below $160 per capita.

Figure 5: UHC performance in countries with public spending on health below $160 pc

Focusing on two of the best performing countries which spend below $86 per capita,
Cambodia and Malawi, we look more closely at the significant difference in performance
between these countries, and selected peer countries in the DEA.

Cambodia: Table 3 shows DEA scores for a selection of the 24 countries for which
Cambodia was a peer in the analysis.

15
 Table 3: Cambodia and selected DEA peers

DEA peer
DEA Score GGHE per capita GGHE % GDP
countries
Sierra Leone 77.8% 31.4 1.95%
Ethiopia 80.5% 36.8 2.97%
Kenya 87.8% 39.9 1.82%
Cambodia 100% 41.1 1.45%
Cameroon 77.2% 42.1 1.62%
Tanzania 87.0% 45.5 2.78%
Senegal 87.2% 48.2 2.18%

We focus even more specifically, using a spider chart (see Figure 6), to compare
performance between Cambodia and Cameroon which have almost identical levels of public
spending on health per capita, PPP$ 41.10 and PPP$ 42.10 respectively, but significantly
different performance. Both countries spend less than half of the US$86 per capita spending
estimate, but Cameroon achieves significantly less in terms of UHC progress than Cambodia
as shown its lower DEA score (77.2% compared with 100%). Both countries perform poorly
in terms of financial protection overall, perhaps not surprising given the very low level of
absolute spending in both, but Cambodia performs better on all of the service coverage
indicators in particular family planning and ART coverage.

Figure 6: Comparison of UHC progress in Cambodia and Cameroon

Cambodia (DEA=100%) versus Cameroon (DEA=77.2%)

Financial Protection
100%

80%

60%
ART Coverage TB Coverage
40%

20%

0%

SBA Coverage DPT3 Coverage

Cambodia ($41.1)
Cameroon ($42.1) Fam. Planning
Coverage

At this stage we are interested in looking at the main differences in performance, and asking
questions which merit further analysis in terms of the way in which the country’s health
systems are organised. Clearly, however a wide-range of non-health system factors could at
least in part explain such differences, such as levels of education and cultural attitudes
towards health-seeking behaviour, as well as population density which is much greater in
Cambodia may also explain some of the variation [22]. Interestingly, in addition to being
richer, Cameroon actually gives a higher priority to health than Cambodia in terms of
government allocations, something which illustrates broader fiscal capacity and the effect of

16
using PPP$, given that public spending per capita on health is almost identical in the two
countries (see Table 4).

Table 4: Selected indicators for Cambodia and Cameroon

Population Education:
Population
Priority: Income level: density (people Mean years of
Country GGHE/GDP size (millions
GGHE/GGE GNI per capita per sq. km of schooling (of
of hab.)
land area) adults)
Cambodia
1.45% 6.7% 880 14.8 84 5.8
(LIC)
Cameroon
1.62% 8.5% 1,230 21.7 46 5.9
(LMIC)

The best performing country for its level of spending, Cambodia has undergone a number of
health sector reforms in the past decade, as noted earlier. Cameroon has looked to cover its
population through its “Strategic Plan for the Promotion and the Development of Mutual
Health Insurance (2005–2015)”, but limited progress has been made with only one percent
of the population covered through these schemes in 2010. In 2013, Cameroon also
introduced programme budgeting to improve public expenditure management which has led
to low budget execution and poor strategic allocation of resources in the health sector.
Whilst enjoying one of the highest densities of nurses and doctors in Sub-Saharan Africa,
they are heavily concentrated in urban areas in Cameroon; the country is now starting to
pilot voucher-type schemes to improve maternal health, and results-based financing [23].
Clearly a more systematic comparison of how the two health systems are organised is
required to understand these differences in performance on core essential health services in
more detail.

Malawi: Table 5 shows DEA scores for a selection of the 15 countries for which Malawi,
another best-performer, was a peer in the analysis.

Table 5: Malawi and selected peers

DEA peer
DEA Score GGHE per capita GGHE % GDP
countries
Tajikistan 87.3% 41.1 1.88%
Tanzania 87.0% 45.5 2.78%
Senegal 87.2% 48.2 2.18%
Malawi 100% 50.9 5.16%
Sudan 59.2% 54.3 1.50%
India 83.5% 59.6 1.28%
Philippines 87.4% 82.3 1.35%

Figure 7 compares Malawi with Sudan, which has a very similar level of spending, PPP$
50.9 compared with PPP$ 54.3 but very different performance; Sudan achieves a DEA score
of 59.2%.

17
 Figure 7: Comparison of UHC progress in Malawi and Sudan

Malawi (DEA=100%) versus Sudan (DEA=59.2%)

Financial Protection
100%

80%

60%
ART Coverage TB Coverage
40%

20%

0%

SBA Coverage DPT3 Coverage

Malawi ($50.9)

Sudan ($54.3) Fam. Planning

Coverage

Malawi performs dramatically better on ART, SAB and family planning coverage, with a
similarly high level for DPT3 and TB treatment coverage. Despite a slightly lower level of
public spending in absolute terms, and with 70% of total health expenditures coming from
external sources which may have some influence on the quality of spending (an important
issue beyond the scope of this paper), Malawi performs significantly better in terms of
financial protection.

Table 6 shows selected indicators for both countries; Malawi has a much lower GNI per
capita than Sudan, and only achieves a similar level of public spending by giving a far higher
priority to health. With lower levels of education, Sudan is also a far larger and less densely
populated country with many remote communities.

Table 6: Selected indicators for Malawi and Sudan

Population Education:
Income Population
Priority: density (people Mean years of
Country GGHE/GDP level: GNI size (millions
GGHE/GGE per sq. km of schooling (of
per capita of hab.)
land area) adults)
Malawi
5.16% 22.1% 320 15.7 167 4.2
(LIC)
Sudan
1.50% 11.1% 1,650 37.7 20 3.1
(LMIC)

In terms of health system organisation, Malawi has invested significantly in health systems
strengthening activities in recent years, and increasingly focuses its public resources on a
clearly defined essential health package which can be accessed free at the point of service
[24, 25]. The provision of services and management of funds are decentralised with districts
having extensive responsibility. Interestingly, Malawi allocates significant public resources to
cross-cutting health systems activities e.g. management of medical equipment and supply
chain.

Sudan has a long-established federal health insurance system although its population reach
has been limited, and is only recently looking to establish a set of essential health services

18
for the entire population [26]. Again, this analysis aims to service as an initial investigation of
the data, with a more systematic comparison of how the two health systems are organised,
as well as other influential environmental factors, required to understand the differences in
performance in more detail.

4. Conclusions

This paper was motivated by a desire to bring greater perspective to the literature on health
expenditure targets, which are limited in terms of providing country policy makers with useful
analysis. We systematically explore the relationship between levels of public spending on
health, and performance on a number of agreed indicators of UHC progress; we use both
descriptive analysis and data envelopment analysis.

There is very high variation in how countries perform, at low levels of public spending (below
PPP$40 per capita). As spending increases, there is convergence in UHC performance
across countries; interestingly, increased public spending quickly translates into improved
service coverage (between PPP$40-60 per capita), but not financial coverage which only
improves significantly when public spending is greater than PPP$200 per capita.

Overall the analysis suggests significant scope for improvements in how public funds are
spent, particularly at low levels. Poor performance is observed across the sample on family
planning with modern methods and coverage of anti-retroviral treatment. Further analysis is
required to understand and explain why low-spending countries with similar levels of public
spending have significant variations in performance; clearly both non-health systems factors
such as population density, as well factors relating to how health systems are organised and
how money is spent, will be important.

One issue which merits further investigation is the role of external funding; several countries
in our sample have a high reliance on external funding which in most cases is focused on
the services we use as UHC tracer interventions. Whilst we capture the majority of external
funding in our measure of public spending (GGHE), we do not capture potential effects on
the quality of spending. Indeed the quality of public spending, whether external or domestic,
is relevant for all country and merits deeper investigation.

Finally, this analysis clearly demonstrates that absolute levels of public spending on health
are critical for progress on UHC, but that at any given level of spending, but particularly at
low levels, there is significant scope for UHC progress through greater efficiency in how
money is spent.

19
Annex 1: Summary of international health expenditure benchmarks and targets

HEALTH EXPENDITURE TARGETS

YEAR SOURCE Absolute terms As a share of As a share of

OOPs/THE
(per capita) GDP GGE

McIntyre et al. [4] $86 GGHE in LIC 5% GGHE

2014 WHO/PAHO [27] 6% GGHE

HTGSD [28] 5% GGHE

HLTF, WHO and $60 THE LIC by

partners [29] [30] 2020
2010
WHO [1] [5] 5%-6% GGHE 15%–20%

WHO/EMRO [31] GGHE of 8%

High Level Task $54 THE LIC by

Force [6] 2015

2009 WHO/WPRO SEARO

4%-5% THE 30%–40%
[32]

WHO/PAHO [33] 6% GGHE

Abuja Declaration [34] GGHE of 15%

Commission on $34 and $38 LIC

Macroeconomic and by 2007 and
2001
Health [35] 2015*

WHO, Evans et al. $80 THE per

[36] capita

WHO, Evans et al. $60 THE per

2000
[37] capita

$12 and $22 per

1993 World Bank [38]
capita LIC, LMIC*

1980s WHO [39] [40] 5% THE/GNP

Note: LIC: Low-income countries; LMIC: Lower-middle income countries.

* Related to the minimum package required and which should be achieved by joint government, donor
agencies and patients efforts.

20
Annex 2: Sensitivity analysis

Three sensitivity analyses were performed:

• first, we replace the proxy indicator for financial protection by an alternative based
on 1-(OOPs as %THE)
• second, we replace all indicator scores for 2012 with average scores for the 5-year
period 2008-2012.
• third, we remove all imputed indicator scores which further restricts our sample to 60
countries.

Alternative indicator of financial protection: we re-run the main model replace our proxy
for financial protection (GGHE/THE) with the indicator (1–OOP/THE) as per the discussion in
Section 3: Indicators of financial protection. The results remain similar; five of the six “best
performer” countries in the main model remain, with Cambodia excluded, albeit still with a
very high DEA score. Overall across the sample, DEA scores do not change more than 1.2%
from the main model. Only where a country has high expenditures on voluntary health
insurance and or external funding not flowing through government channels would we
expect significant change; Figure 8 shows countries where these flows represent more than
5% of THE (see grey bar), and hence the use of this alternative indicator results in a
significantly different indicator score; subsequent DEA scores under the main and alternative
model are plotted. The only country’s where the DEA score differs significantly are Haiti,
Mozambique and Liberia, and in each case improve as a result.

Figure 8: Impact on DEA scores of using alternative proxy measure of financial protection

100%
95%
90%
85%
80%
75%
70%
65%
60%
55%
50%
45%
40%
35%
30%
25%
20%
15%
10%
5%
0%
Dominican Rep.

Niger
Peru

Togo

Guinea
Mozambique

Malaysia

Colombia

Vietnam
Malawi

Rwanda

Burundi
Georgia
Cote d'Ivoire
Indonesia

Senegal
Argentina
Tajikistan
Guatemala

China

Moldova

Turkey

Venezuela
Afghanistan

Iran

Bolivia
Nicaragua
Tanzania
Haiti

Cambodia
Sierra Leone

Madagascar

Azerbaijan

Morocco

Burkina Faso

Panama
Nepal

Sri Lanka

Pakistan

Tunisia
Liberia

Brazil

Kenya
Philippines

Thailand
India
Lao PDR

Difference between two proxy indicators of financial protection DEA score main model DEA score sensitivity analysis

21
Longitudinal analysis: We ran the DEA model using time series averages across a period
of five years (2008-2012) i.e. averages for input and output values were calculated for the 83
countries in question. Although complete data was available across the five years for the
health spending indicator, this was not the case for all output values. We hence calculated
averages using whatever data was available in service coverage indicators. The DEA model
still returns six countries as best performers, namely Myanmar, Cambodia, Malawi, China,
Thailand and Cuba. The only change was that China replaces Rwanda, which still scores
highly (98%); the other five countries remaining the same.

Remove of imputed data: out of the sample of 83 countries 23 countries contain estimated
data in 2012. We conducted sensitivity analysis by removing these countries leaving a
sample of 60 countries, with Myanmar and Rwanda, two of the best performer in the main
model removed from the sample with Haiti and Vietnam now shown as best performers.

Across these three sensitivity analyses, three of the six countries identified in the main
model as “best performers” for their level of public spending remain the same, namely
Malawi, Thailand and Cuba.

22
Annex 3: Methodological approach to the analysis

Unlike parametric approaches used to measure efficiency, DEA does not require any
assumptions to be made regarding the functional form of the frontier, which is essentially
made up of those DMUs/countries with the best outputs/performance for a given of inputs. In
our case outputs are indicators of UHC progress and inputs are public spending on health
per capita. DEA is however more sensitive to outliers as it produces a deterministic frontier
which does not distinguish noise from efficiency scores, nor hypothesis testing or model fit.

In DEA, a DMU is assessed based on the ratio of a weighted sum of its outputs divided by a
weighted sum of its inputs. To do this, each DMU is assigned its own unique set of input and
output weights, which cast it in the best possible light [41]. While this flexibility in determining
weights is an appealing characteristic of DEA, one potential drawback is that if a DMU
performs extremely poorly on some outputs but very well on others, the DEA program may
assign a 0% weight to the poorly produced outputs (or alternatively, assign a 100% weight to
a very high performing output). A 0% weight would imply that said output has no intrinsic
value to the DMU (i.e. it is dropped from the model), which in our analysis is not acceptable
given our concern with country performance on both service coverage and financial
protection. Furthermore, a number of countries score poorly on service coverage but well on
financial protection, which may be due to non-use of health services and unmet health
needs.

Various authors have suggested ways of imposing restrictions on weights in DEA [42]. In this
study we assign the simple constraint that each UHC performance indicator be given a
weight greater than 0. This assures that all indicators are evaluated for all DMUs, while still
permitting a considerable degree of flexibility in the weights. However we also run a DEA
model with no weight restrictions and find similarities in the sense that the most efficient
countries remain the same.

23
Annex 4: Additional analysis

Table 7: Descriptive statistics by country income group of score related to UHC performance
relative to levels of public spending on health

UHC performance relative Number Mean DEA Std. dev. of Min DEA Max DEA
to levels of public spending countries score DEA scores score score
on health

Low income countries 28 82.22% 13.71% 45.48% 100%

Lower-middle income 28 83.53% 9.92% 57.13% 98.87%

countries

Upper-middle income 27 87.50% 6.80% 73.83% 100%

countries

Figure 9: Coverage progress relative to public spending on health per capita

24
Figure 10: Public spending on health as % GDP by quintile relative to coverage performance

10% 20% 30% 40% 50% 60% 70% 80% 90% 100%

<1.6%
Q1 1.6%-2.18%
Q2 2.22%-3.16%
Q3 3.25%-4.21%
Q4 >4.4%
Q5

Financial Protection (proxy)

Average all service coverage indicators
Q1 to Q5 denote quintiles of public spending on health as a share of GDP n=16 or 17

Figure 11: UHC performance relative to public spending on health as %GDP

100%
95%
90%
UHC Performance relative to Public spending

85%
80%
75%
70%
65%
60%
Low Income
55%
Low-middle Income
50%
45% Upper-middle
Income
40%
0.0% 0.5% 1.0% 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 4.5% 5.0% 5.5% 6.0% 6.5% 7.0% 7.5% 8.0% 8.5%
Public spending on health as a share of GDP

25
Annex 5: List of sample countries with official indicator values

Public spending Fam.

Financial TB DPT3 SBA ART
DEA on health per planning
Countries protection coverage coverage coverage coverage
Score capita in Int.$ coverage
(proxy) 27 28 29 31 32
(PPP) 26 30

LOW-INCOME COUNTRIES
1 Afghanistan 66.2% 33.10 20.8% 88.0% 71.0% 39.9% 38.6% 4.7%

2 Bangladesh 85.4% 27.20 31.9% 92.0% 97.0% 70.1% 31.7% 11.3%

3 Benin 87.4% 48.95 57.4% 90.0% 76.0% 19.4% 84.1% 33.6%

4 Burkina Faso 88.8% 55.76 58.5% 80.0% 90.0% 38.7% 65.9% 37.4%

5 Burundi 95.1% 37.53 60.1% 89.0% 96.0% 37.1% 60.3% 39.7%

6 Cambodia 100% 41.12 19.7% 94.0% 95.0% 55.4% 74.0% 67.3%

Central
7 African 67.8% 18.22 51.1% 68.0% 47.0% 25.0% 53.8% 13.9%
Republic
8 Chad 52.0% 27.82 38.1% 69.0% 45.0% 8.5% 22.7% 20.5%

9 Eritrea 89.7% 16.61 46.5% 87.0% 94.0% 28.1% 34.1% 51.5%

10 Ethiopia 80.4% 36.80 60.6% 91.0% 69.0% 51.9% 10.0% 40.0%

11 Guinea 68.5% 21.88 39.2% 82.0% 63.0% 12.5% 45.3% 22.1%

12 Haiti 74.9% 10.81 9.2% 81.0% 68.0% 45.4% 37.3% 39.4%

13 Kenya 87.8% 39.90 40.9% 86.0% 83.0% 60.0% 43.8% 41.0%

Kyrgyz
14 93.9% 125.54 60.2% 84.0% 96.0% 63.1% 99.1% 13.4%
Republic
15 Liberia 83.0% 25.85 34.5% 79.0% 93.0% 34.8% 61.1% 21.2%

16 Madagascar 77.7% 23.44 50.7% 82.0% 70.0% 53.5% 43.9% 1.0%

17 Malawi 100% 50.93 56.1% 82.0% 96.0% 62.6% 71.4% 46.3%

18 Mali 78.9% 38.22 38.8% 93.0% 74.0% 25.8% 56.1% 29.6%

19 Mozambique 84.3% 28.41 49.4% 87.0% 76.0% 32.5% 54.3% 31.8%

20 Myanmar 100% 8.32 23.9% 89.0% 84.0% 69.2% 70.6% 35.2%

21 Nepal 82.7% 46.78 39.5% 91.0% 90.0% 60.3% 36.0% 22.9%

22 Niger 70.9% 18.04 33.1% 77.0% 74.0% 27.7% 29.3% 29.8%

23 Rwanda 100% 92.71 58.8% 84.0% 98.0% 61.0% 69.0% 66.3%

24 Sierra Leone 77.8% 31.38 17.9% 90.0% 91.0% 30.4% 62.5% 15.9%

25 South Sudan 45.5% 17.37 33.3% 52.0% 59.0% 5.1% 19.4% 4.5%

26 Tajikistan 87.3% 44.33 29.4% 83.0% 94.0% 52.7% 87.4% 10.1%

27 Tanzania 87.0% 45.54 39.1% 90.0% 92.0% 49.7% 48.9% 36.6%

28 Togo 81.5% 55.52 50.9% 86.0% 84.0% 29.0% 59.4% 30.5%

26 Global Health Expenditure Database (GHED), 2012. http://apps.who.int/nha/database/Select/Indicators/en.

27 Global Health Expenditure Database (GHED), 2012. http://apps.who.int/nha/database/Select/Indicators/en.
28 Global Health Observatory (GHO) data repository, 2012 or nearest year.

http://apps.who.int/gho/data/node.main.1335?lang=en
29 GHO data repository, 2012 or nearest year. http://apps.who.int/gho/data/node.main.1335?lang=en
30 World Bank & GHO, 2012 or nearest year. http://databank.worldbank.org/data/reports.aspx?source=health-

nutrition-and-population-statistics & http://apps.who.int/gho/data/view.main.94330

31 World Bank, 2012 or nearest year. http://databank.worldbank.org/data/reports.aspx?source=health-nutrition-

and-population-statistics
32 World Bank & GHO, 2012 or nearest year. http://databank.worldbank.org/data/reports.aspx?source=health-

nutrition-and-population-statistics & http://apps.who.int/gho/data/view.main.23300?lang=en

26
 Public spending Fam.
Financial TB DPT3 SBA ART
DEA on health per planning
Countries protection coverage coverage coverage coverage
Score capita in Int.$ coverage
(proxy) 27 28 29 31 32
(PPP) 26 30

LOWER-MIDDLE INCOME COUNTRIES

29 Armenia 83.6% 139.14 41.9% 81.0% 95.0% 38.7% 99.5% 15.8%

30 Bolivia 83.9% 238.23 71.8% 84.0% 93.0% 48.0% 71.1% 19.5%

31 Cameroon 77.2% 42.13 32.4% 79.0% 85.0% 31.7% 63.6% 21.8%

Congo,
32 82.6% 180.45 79.2% 70.0% 69.0% 33.0% 93.6% 28.4%
Rep.
Cote
33 74.8% 54.97 30.4% 79.0% 82.0% 30.5% 59.4% 30.4%
d'Ivoire
Egypt, Arab
34 86.5% 207.62 39.0% 88.0% 93.0% 80.4% 78.9% 15.7%
Rep.
35 El Salvador 98.5% 318.81 62.8% 93.0% 92.0% 77.4% 95.5% 47.5%

36 Georgia 86.1% 116.00 18.0% 85.0% 92.0% 51.9% 99.8% 32.9%

37 Ghana 85.6% 133.41 68.3% 84.0% 92.0% 33.5% 68.4% 33.8%

38 Guatemala 81.2% 159.40 34.9% 88.0% 96.0% 62.5% 51.5% 31.0%

39 Honduras 92.5% 188.78 44.4% 89.0% 88.0% 75.8% 82.9% 39.1%

40 India 83.5% 59.56 30.5% 88.0% 72.0% 71.1% 52.3% 35.9%

41 Indonesia 85.8% 108.26 39.6% 86.0% 83.0% 79.3% 83.1% 6.2%

42 Lao PDR 90.2% 27.21 31.8% 90.0% 79.0% 62.8% 41.5% 47.8%

43 Mauritania 73.5% 57.57 47.0% 68.0% 80.0% 25.7% 65.1% 21.5%

44 Moldova 85.0% 226.92 45.6% 76.0% 92.0% 61.2% 99.2% 16.8%

45 Mongolia 90.9% 333.25 62.8% 88.0% 99.0% 71.1% 98.8% 14.5%

46 Morocco 87.6% 152.87 35.5% 89.0% 99.0% 73.7% 73.6% 21.1%

47 Nicaragua 98.9% 192.73 54.3% 87.0% 98.0% 86.8% 88.0% 34.8%

48 Nigeria 57.1% 60.91 33.2% 86.0% 26.0% 26.2% 48.7% 19.8%

49 Pakistan 73.4% 45.18 36.9% 91.0% 72.0% 45.1% 49.0% 6.5%

50 Paraguay 82.6% 294.11 42.6% 70.0% 87.0% 80.8% 81.9% 26.8%

51 Philippines 87.4% 82.28 30.4% 88.0% 86.0% 51.4% 72.2% 49.2%

52 Senegal 87.2% 48.21 50.6% 84.0% 91.0% 33.1% 65.1% 35.4%

53 Sri Lanka 93.7% 105.40 39.1% 86.0% 99.0% 69.4% 99.9% 18.2%

54 Sudan 59.2% 54.27 22.5% 75.0% 92.0% 27.2% 23.1% 6.7%

55 Ukraine 81.7% 351.60 55.4% 71.0% 76.0% 64.3% 99.9% 26.5%

56 Vietnam 98.7% 124.23 42.6% 91.0% 97.0% 79.9% 92.9% 33.3%

UPPER-MIDDLE INCOME COUNTRIES

57 Albania 89.1% 249.58 49.3% 92.0% 99.0% 21.4% 99.3% 52.1%

58 Algeria 88.0% 573.91 73.5% 90.0% 95.0% 72.8% 96.9% 18.4%

59 Argentina 88.2% 1073.65 69.3% 56.0% 91.0% 78.2% 97.1% 62.8%

60 Azerbaijan 73.8% 198.14 22.6% 83.0% 89.0% 28.3% 99.4% 13.6%

61 Belarus 89.3% 669.17 77.2% 85.0% 98.0% 73.6% 100.0% 20.6%

Bosnia and
62 81.3% 665.62 71.2% 84.0% 92.0% 22.4% 99.9% 44.2%
Herzegov.
63 Brazil 87.9% 659.13 47.5% 72.0% 97.0% 86.7% 98.1% 45.9%

64 Bulgaria 83.7% 659.19 56.3% 87.0% 95.0% 57.0% 99.5% 31.3%

65 China 99.3% 323.34 56.0% 95.0% 99.0% 94.5% 99.8% 28.7%

66 Colombia 88.6% 612.37 75.8% 72.0% 91.0% 82.6% 99.3% 29.4%

67 Costa Rica 96.5% 1008.18 74.7% 86.0% 91.0% 89.0% 99.1% 56.5%

68 Cuba 100% 1642.50 94.2% 85.0% 96.0% 87.7% 99.9% 62.0%

27
 Public spending Fam.
Financial TB DPT3 SBA ART
DEA on health per planning
Countries protection coverage coverage coverage coverage
Score capita in Int.$ coverage
(proxy) 27 28 29 31 32
(PPP) 26 30

Dominican
69 94.5% 309.05 50.9% 82.0% 85.0% 84.3% 98.6% 47.0%
Republic
70 Ecuador 87.8% 300.76 44.8% 75.0% 99.0% 73.9% 90.5% 31.5%
Iran, Islam
71 79.1% 491.66 40.4% 87.0% 99.0% 70.3% 96.4% 6.4%
Rep.
72 Kazakhstan 85.0% 543.80 55.8% 86.0% 99.0% 72.6% 100.0% 17.4%
Macedonia,
73 78.3% 520.48 65.3% 86.0% 95.0% 22.1% 98.3% 29.9%
FYR
74 Malaysia 80.1% 493.93 55.2% 78.0% 97.0% 55.6% 98.6% 20.1%

75 Mexico 90.4% 550.17 51.8% 80.0% 99.0% 80.5% 96.0% 51.1%

76 Panama 85.4% 891.65 68.6% 80.0% 85.0% 69.1% 88.5% 46.6%

77 Peru 85.3% 373.18 58.7% 67.0% 95.0% 63.0% 86.7% 43.0%

78 Romania 92.7% 788.39 80.3% 85.0% 89.0% 66.4% 98.5% 54.4%

79 Serbia 80.7% 759.53 61.2% 84.0% 91.0% 37.5% 99.7% 38.6%

80 Thailand 100% 500.29 79.5% 81.0% 99.0% 90.7% 99.5% 56.5%

81 Tunisia 86.5% 441.33 59.0% 89.0% 97.0% 69.9% 98.6% 15.9%

82 Turkey 88.5% 745.38 76.8% 88.0% 97.0% 58.0% 95.0% 36.5%

83 Venezuela 87.5% 281.87 32.4% 82.0% 81.0% 77.0% 95.7% 42.3%

28
Annex 6: Limitations of the analysis

There are a number of limitations to this analysis. As noted in section 2c) of the main text,
our measure of financial protection does not capture the effects of a household’s private
spending on health on its overall economic situation and is hence a very limited measure of
financial protection. As data on the preferred and more precise indicators of catastrophic and
impoverishing health expenditures become more widely available, they can be used to
rework this analysis.

Second, due to a lack of verified and published data for individual countries on a number of
service coverage indicators, we use only a limited set of basic interventions in this paper.
There is clearly no 1:1 relationship between inputs and outputs i.e. we use total public
spending on health per capita, which covers a much broader set of interventions (outputs)
than those captured in our analysis. We are hence likely to be underestimating the full effect
of public health spending. Section 2e) provides further discussion. The analysis can be
reworked when a broader set of service indicators are more widely available.

Thirdly, there may be non-public spending which influences performance on the outputs,
which would mean we overestimate the effects of public spending. Given the nature of our
outputs, however, we expect non-public spending to be limited.

Finally, DEA is ideally used to compare decision-making units (DMU) such as a factory, or a
hospital, which operate under very similar conditions. In our analysis, the DMU is a country,
and whilst we remove high-income countries from the sample there is clearly significant
difference on a wide range of contextual and other factors across the 83 countries. For this
reason, we focus more on variation across countries than on the frontier countries
themselves.

29
References:

1. World Health Organization, Health systems financing: the path to universal

coverage, in World Health Report. 2010: Geneva.
2. World Health Organization, Public financing for health in Africa: from Abuja to the
SDGs. 2016, WHO: Geneva.
3. Savedoff, W.D., How much should countries spend on health? FER/EIP discussion
paper. 2003, Geneva: World Health Organization. 14 p.
4. Mcintyre, D.M., F., Shared Responsibilities for Health: A Coherent Global
Framework for Health Financing, in Final Report of the Centre on Global Health
Security Working Group on Health Financing. 2014, Chatham House: London, UK.
5. Xu, K.S., P. Jowett, M. Indikadahena, C. Kutzin, J. Evans, D.B., Exploring the
thresholds of health expenditure for protection against financial risk. 2010, World
Health Organization: Geneva.
6. World Health Organization (High-level Taskforce on Innovative International
Financing for Health Systems), More money for health and more health for the
money. 2009.
7. Jowett M and Kutzin J, Raising revenues for health in support of UHC: strategic
issues for policy makers, in Health Financing Policy Brief No. 1. 2015.
8. World Health Organization, Health systems: improving performance, in World
Health Report. 2000, World Health Organization: World Health Organization,
Geneva.
9. Lagarde, M. and N. Palmer, The impact of user fees on health service utilization in
low- and middle-income countries: how strong is the evidence? Bull World Health
Organ, 2008. 86(11): p. 839-848.
10. Kutzin, J., Anything goes on the path to universal health coverage? No. Bulletin of
the World Health Organization, 2012(90): p. 867-868.
11. World Health Organization and World Bank, Tracking universal health coverage:
first global monitoring report. 2015: World Health Organization, Geneva.
12. Hollingsworth, B., The measurement of efficiency and productivity of health care
delivery. Health Econ, 2008. 17(10): p. 1107-28.
13. Moreno-Serra, R. and P.C. Smith, An exploratory application of data envelopment
analysis to the efficiency of health service coverage and access. Results for
Development Institute Working Paper, 2012.
14. Halstead, B., Good Health at Low Cost. 1985, New York: Rockefeller Foundation.
15. Yiengprugsawan, V., et al., The first 10 years of the Universal Coverage Scheme in
Thailand: review of its impact on health inequalities and lessons learnt for middle-
income countries. Australas epidemiol, 2010. 17(3): p. 24-26.
16. Prakongsai, P., S. Limwattananon, and V. Tangcharoensathien, The equity impact of
the universal coverage policy: lessons from Thailand. Adv Health Econ Health Serv
Res, 2009. 21: p. 57-81.
17. Limwattananon, S., et al., Universal coverage with supply-side reform: The impact
on medical expenditure risk and utilization in Thailand. Journal of Public
Economics, 2015. 121: p. 79-94.
18. Lu, C., et al., Towards universal health coverage: an evaluation of Rwanda
Mutuelles in its first eight years. PLoS One, 2012. 7(6): p. e39282.
19. Flores, G., et al., Financial protection of patients through compensation of
providers: The impact of Health Equity Funds in Cambodia. Journal of Health
Economics, 2013. 32(6): p. 1180-1193.
20. Van de Poel, E., et al., Impact of Performance-Based Financing in a Low-Resource
Setting: A Decade of Experience in Cambodia. Health Econ, 2015.
21. Van de Poel, E., et al., Can vouchers deliver? An evaluation of subsidies for maternal
health care in Cambodia. Bulletin of the World Health Organization, 2014. 92(5): p.
331-339.

30
22. Hanlon, M., et al., Exploring the relationship between population density and
maternal health coverage. BMC Health Services Research, 2012. 12: p. 416-416.
23. World Bank, Cameroon Economic Update. Towards Greater Equity: A Special
Focus on Health. 2013.
24. Ministry of Health Malawi, Malawi Service Provision Assessment 2013-2014.
November 2014.
25. Ministry of Health Malawi, Malawi Health Sector Strategic Plan 2011-2016: Moving
towards equity and quality. 2011.
26. Ministry of Welfare and Social Security Sudan, Report of Proceedings of the
International Conference: Road to Universal Health Coverage for Sudan. 2014.
27. World Health Organization (Regional Committee for the Americas), Strategy for
Universal Access to health and Universal Health Coverage. 2014.
28. Sustainable Development Network (Thematic Group on Health for All of the
Sustainable Development Solutions), Health in the framework of sustainable
development: Technical Report for the Post-2015 Development Agenda. 2014.
29. World Health Organization (Taskforce on Innovative International Financing for
Health Systems), Constraints to Scaling Up the Health Millennium Development
Goals: Costing and Financial Gap Analysis. Geneva, 2010.
30. World Health Organization (Taskforce on Innovative International Financing for
Health Systems), Constraints to Scaling Up and Costs. Working Group 1 Report.
Geneva, 2009.
31. World Health Organization (Regional Committee for the Eastern Mediterranean),
Strategic directions to improve health care financing in the Eastern Mediterranean
Region: moving towards universal coverage 2011–2015. 2010.
32. World Health Organization (Regional Offices for South-East Asia & the Western
Pacific), Health financing strategy for the Asia Pacific Region (2010-2015). 2009,
New Delhi & Manila.
33. Organization of American States, Towards the Fifth Summit of the Americas:
Regional Challenges. Summits of the Americas Secretariat. Washington D.C., 2009.
34. Abuja Declaration on HIV/AIDS, Tuberculosis and Other Related Infectious
Diseases. April 2001. OAU/SPS/ABUJA/3.
35. Sachs, J., Macroeconomics and Health: Investing in Health for Economic
Development. 2001, Commission on Macroeconomics and Health, World Health
Organisation: Geneva. p. 202.
36. Evans, D.B., et al., Comparative efficiency of national health systems: cross national
econometric analysis. BMJ, 2001. 323(7308): p. 307-10.
37. Evans, D.B., Ajay, T., Murray, C. J., & Lauer, J. A., The comparative efficiency of
national health systems in producing health: an analysis of 191 countries.
Discussion Paper Series, World Health Organization., 2000.
38. World Bank, Investing in Health. 1993, New York: World Bank.
39. World Health Organization., Development of indicators for monitoring progress
towards Health for All by the Year 2000. Health for all series. 1981, Geneva: World
Health Organization. 91 p.
40. World Health Organization, Planning of the finances of Health for All : economic
strategies to support the Strategy for Health for All, report by the Director-General.
1985, Geneva: World Health Organization. 25 p.
41. Thanassoulis, E., Introduction to the theory and application of data envelopment
analysis : a foundation text with integrated software. 2001, Norwell, Mass.: Kluwer
Academic Publishers. xvi, 281 p.
42. Jacobs, R., P. Smith, and A. Street, Measuring efficiency in health care: analytic
techniques and health policy. 2006, Cambridge: Cambridge University Press.

31
32
Health System Health Cost Effectiveness,
Governance, Policy and Financing Expenditure and
Aid Effectiveness (HGS) (HEF) Priority Setting (CEP)

WE ALSO HOST THE

SECRETARIATS FOR:

For additional information, please contact:

Department of Health Systems Governance and Financing

Health Systems and Innovation
World Health Organization

Email: [email protected] Website: http://www.who.int/health_financing