Equity in health and health care:lessons from an Asian comparative

studyEddy van Doorslaer

Erasmus School of Economics & Erasmus Medical Centre

Rotterdam

Merck Foundation Lecture London School of Economics, 16 March 2007

Introduction - background

Inequality and inequity in the distribution of– Health– Health care– Health care payments

Some analogies with measurement of income inequality, but not perfect

Approaches to study equity in health care finance and delivery in high-income (EU/OECD) countries, developed in ECuity Project

Usefulness for low-to-middle income countries, examined in EQUITAP Project (Asian-Pacific)

2

Draws on several papers

1. Van Doorslaer, E, O O’Donnell, R Rannan-Eliya et al (20 authors) Effect of payments for health care on poverty estimates in 11 countries in Asia: an analysis of household survey data, The Lancet, 2006, 368, Dec

2. O’Donnell, O, E van Doorslaer, R Rannan-Eliya et al (18 authors) The incidence of public spending on health care: comparative evidence from Asia, World Bank Economic Review, (Published online, Jan 2007)

3. Van Doorslaer, E, O O’Donnell, R Rannan-Eliya et al (23 authors) Catastrophic expenditures on health care in Asia, Health Economics, (Published online, Feb 2007)

I. Distribution of health care payments

In OECD context: mostly publicly financed health care systems with universal coverage have disconnected receipt of care from payment for care

Payments according to ability to pay Tools of progressivity and redistribution can be used to

examine health payments by income Descriptive x-country comparative results informative

on potential consequences of alternative health care financing choices

E.g out-of-pocket payments most regressive source

The health care financing triangle - OECD(Wagstaff, Van Doorslaer et al, JHE, 1999)

US

UK

Switzerland

Sweden

Spain Portugal

Netherlands

Italy

Ireland

Germany

France

Finland

Denmark

Belgium

0%

20%

40%

60%

80%

100%

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

% social insurance

% g

en t

ax

0% private finance

Overall progressivity hc finance — OECD(Wagstaff, Van Doorslaer et al, JHE, 1999)

-0.15 -0.10 -0.05 0.00 0.05 0.10 Kakwani index

Denmark (1987)Finland (1990)France (1989)

Germany (1989)Ireland (1987)

Italy (1991)Netherlands (1992)

Portugal (1990)Spain (1990)

Sweden (1990)Switzerland (1992)

United Kingdom (1992)United States (1987)

Distribution of health care finance

In low-income settings: high shares of out-of-pocket spending, small pre-paid shares

And receipt of care linked to payment for care Payments largely according to use Tools of progressivity and redistribution can be

very misleading Rich pay more, even in proportion to their income Interest in (lack of) prepayment and protection

against income loss

Health care financing triangle – Asia(O’Donnell, Van Doorslaer et al, 2006)

Thailand

Taiwan

Sri Lanka

Punjab Philippines

Nepal

Kyrgyz Rep.

Korea Rep.

Japan Indonesia

Hong Kong

China

Bangladesh

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

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

% Out-of-pocket

% G

en T

ax R

evR

Inequality and progressivity – total health care payments, Asia (O’Donnell et al, 2006)

-0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6

Japan Taiwan

Korea Rep. Kyrgyz Rep.

China Punjab Nepal

Sri Lanka Philippines Hong Kong Indonesia

Thailand Bangladesh

Concentration index Kakwani index

Measures of income protection of interest High health expenditure out-of-pocket (OOP) can affect

household welfare through– Effect on health care use: see BIA– Effect on living standards: income protection

Wagstaff and van Doorslaer (2003) proposed to look at measures of ‘impact’ of OOPs on– Incidence and intensity of high or catastrophic

expenditure– Incidence and intensity of poverty changes

Impoverishment

Most health spending in low-income settings not discretionary

Standard poverty headcount and gap measures do not account for health needs

Variability and unpredictability of OOPs makes adjustment of poverty line for health needs difficult

How many more individuals are below the poverty line – and by how far – if OOP spending is deducted from total spending?

Change in headcount and gap makes ‘hidden poor’ visible

Pen’s Parade and OOPs in China, 2000

Pen' s parade - China 2000

0

5

10

15

20

25

30

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

HHs ranked by pre-payment expendi ture per capi ta

HHs

expe

ndit

ure

per

capi

ta a

s m

ulti

ple

of$1

.08

PL

PL1（$1.08 ） pre-payment exp. post-payment exp. PL2（$2. 15 ）

Poverty headcount (at $1) increase by OOP share

Vietnam

Thailand Sri Lanka

Philippines

Nepal

Malaysia Kyrgyz

Indonesia

India

China

Bangladesh

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0%

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

OOP share of total health expenditure

Cha

nge in

head

co

unt ($

1 p

ov

lin

e)

Poverty headcount (at $1) increase by population at risk

Vietnam

Thailand Sri Lanka

Philippines

Nepal

Malaysia Kyrgyz

Indonesia

India

China

Bangladesh

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0%

0% 10% 20% 30% 40% 50% 60%

% with pre-payment consumption between $1 and $2

Change in

headcount ($

1 p

overt

y li

ne)

Some EQUITAP Project findings – Asia(Van Doorslaer et al, Lancet, 2006)

– Estimated increase in poverty headcount from 19.3% to 22% of total population (so plus 2.7% of population)

– or an extra 78.25 million Asian people poor– who are currently not counted as poor because their

OOP health spending lifts them above poverty line– Largest relative increases in Bangladesh, India, China

and Vietnam– Magnitude clearly linked to OOP finance share and to

population share at risk– But again, even at similar OOP share levels, some

countries appear more successful at protecting living standards than others. So not just a question of affordability

II. Distribution of health care utilisation in rich countries (Van Doorslaer et al, CMAJ, 2006)

In OECD context: horizontal equity (HE) concern requires equal treatment for equal need, irrespective of income

Measurement of inequality in need-standardised distribution of use by income

Norm = average relationship between use and need Many systems close to achieving HE, but varies by type

of careE.g.: – GP care often pro-poor– specialist care everywhere pro-rich– more pro-rich when private options available

Access to health care in Asia?

In low-income countries: self-reported health measures often show less pro-poor gradient in LDCs

Unlike more objective indicators of need Need measurement problematic Therefore: assume need is equal or at least

not pro-rich and look at distribution of actual utilization, i.e.

unadjusted for need

Poor kids more likely to die before age five 21

But poor kids and mothers less likely to receive health services

22

Who makes use of public health services? Benefit-incidence analysis (BIA)

Important to realize: Even if benefits not distributed according to needs In-kind transfers still achieve a welfare

redistribution even if incidence itself is pro-rich if incidence of public expenditure less

concentrated among the rich than income itself

Use formal tests of stochastic dominance to compare concentration curves of public health

subsidy with diagonal and Lorenz curves

Example: concentration of public health subsidy in Malaysia and India

BIA of public health subsidy – Asia (O’Donnell, Van Doorslaer et al, WBER, 2007)

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Concentration index

Kakwani index

Pro-rich

Pro-poor

Inequality-increasing

Inequality-reducing

Cross-country dominance results – Asia(O’Donnell, Van Doorslaer et al, WBER, 2007)

Malaysia Thailand Sri Lanka Vietnam Bangladesh Indonesia India Gansu Heilongjiang Nepal

Hong Kong SAR D* D* D* D D* D* D* D* D* D*

Malaysia ns ns D D D* D* D* D* D*

Thailand ns D D D* D* D D* D*

Sri Lanka ns ns D D D D* D*

Vietnam D D* D D* D D*

Bangladesh ns ns ns ns ns

Indonesia ns D ns D

India D ns D

Gansu (China) ns ns

Heilongjiang (China) ns

D= Stochastic Dominance (more pro-poor) ; * = strict D= Stochastic Dominance (more pro-poor) ; * = strict dominance ; ns = not signif 5%dominance ; ns = not signif 5%

Cross-country dominance results – Asia(O’Donnell, Van Doorslaer et al, WBER, 2007)

Very pro-poor: Hong Kong Mildly pro-poor: Malaysia, Thailand, Sri Lanka Mildly pro-rich: Vietnam Very pro-rich: Bangladesh, Indonesia, India, Gansu,

Heilongjiang and Nepal

Two questions

Does public policy fail in low-income countries where poor do not even get their fair (pop) share of public subsidy? Subsidy still narrows rich-poor gap, and poor may benefit more from same subsidy.

Why is public health subsidy more pro-poor in Malaysia, Thailand, Sri Lanka (and Vietnam)?All three emphasize universality, minimize user charges, seek to exempt the poor and spend more.And private sector offering attractive alternative seems to lead to better targeting and redistribution.

Distributions of public and private inpatient care(O’Donnell et al, 2006)

Thailand

Sri Lanka

Malaysia

IndonesiaIndia

Hong KongBangladesh

0.0

0.2

0.4

0.6

0.8

-0.6 -0.4 -0.2 0 0.2 0.4

Concentration index of public care

Con

centr

ati

on index o

f pri

vate

car

e

III. Health inequality

Ultimately of greatest concern Which inequality is inequitable? Application of rank-based measures such as

concentration indices and curves less analogous for health measures, and dependent on measurement scale of health

Welfare basis for income-related health inequality rests on strong assumptions (Bleichrodt & V Doorslaer, 2006)

Two issues here:– Reporting heterogeneity– Decomposition analysis

Heterogeneous health reporting?

Socioeconomic gradients in self-reported health often not found in LDC survey data (e.g. LSMS illness questions)

Suspicion of systematic heterogeneity, i.e. at same true but unobserved health, poor report better health

Implicit health scale cutpoints may vary (eg by SES) Proposed fix: identify any reporting biases (eg by SES)

from the evaluation of health vignettes Then use this information to ‘correct’ health scale

cutpoints And use corrected cutpoints when analyzing individual’s

own health (eg with hierarchical ordered probit model)

Example from Asia(taken from Bago d’Uva et al, 2005)

WHO Multi-Country Survey Data for Indonesia, India (Andrah Pradesh) and 3 Chinese provinces

Ratings for 6 health domains For own health and vignettesReporting effects by income: are poor more likely to report

same condition as very good? Yes in India and China (some domains), not in Indonesia (Fig)

Does heterogeneity correction ‘resurrect’ the SE gradients? Yes, for some domains, and some countries. Not for others.

Reporting heterogeneity evidence: rich-poor disparities in China example

Ratio of top to bottom quintile of prob of reporting very good health (on the vignettes): rich have higher expectations

Gansu, Henan & Shan-dong (China)

0.92

0.94

0.96

0.98

1.00

1.02

mobility cognition pain self usual affect

Effect of heterogeneity correction on rich-poor disparities in China example

Gansu, Henan & Shang-dong (China)

1.00

1.05

1.10

1.15

1.20

mobility cognition pain self usual affect

Ratio of top to bottom quintile of prob of being in very good health (own)

Beyond measurement: decomposition of health inequality

In general, for any linear additive explanatory model

such as :

where y is health, X is a vector of determinants, and is

a disturbance term, one can write (Wagstaff et al,

2003):

And in (time) differences:

i k ki iky x

( / ) /k k kkC x C GC

ttktktk ktk ktkttk GCCCCC /111

Example: Child malnutrition in Vietnam 93-98 Decomposing change in CI of haz (Wagstaff et al, 2003)

-0.12

-0.1

-0.08

-0.06

-0.04

-0.02

0

0.02

1993 1998 changeResidual

Commune

Mum educ.

Head educ.

Sanitation

Water

Consumption

Sex

Age

Lessons from EQUITAP: methods

Egalitarian based equity measurement approaches require some reconsideration in low-income, low-coverage contexts

Payments: focus on income protection rather than redistribution

Health care use: examine redistribution through (public) subsidies in absence of need adjustment

Health: inequality decomposition studies using ‘objective’ indicators need to be complemented with studies using self-reported indicators, after purging any reporting biases

In general, descriptive x-country work on measurement and explanation needs to be complemented with well-controlled evaluation work on distributional impact.

Lessons from EQUITAP: findings

Some countries in Asia achieve far better distributional outcomes than others, even at similar income levels

Unfortunately, it is often the larger countries that perform worse (China, India, Indonesia, …)

What explains relative success of the good examples? (Sri Lanka, Thailand, Malaysia)Income and growth helps, but is not sufficientTargeted protection of worst off also helpsInterestingly, a combination of (near) universal public provision, linited user charges and good geographic dispersion of services, coupled with an attractive private alternative sometimes leads to best protection.