Community Based Health Insurance Scheme (Mutuelles) in Rwanda: an evaluative note using household surveys

1
Community Based Health Insurance Scheme
(Mutuelles) in Rwanda an evaluative note using
household surveys

• Abebe Shimeles
• Development Research Department
• African Development Bank
• October 2009

2
1. Introduction

• According to WHO (2005), every year 100 million
  people are driven into poverty due to
  catastrophic expenditure on health related needs.
  
• Certainly the problem is more pervasive in Africa
  where there are little risk-mitigating mechanisms
  against health-related negative shocks.
• Out of pocket household health expenditure is
  generally high and non-monotonic across the
  income divide poor income countries spend as
  much as middle income economies as a share of
  household income (see Figure 1) with considerable
  variation on health outcomes.

3
Figure 1 Share of household out of pocket
expenditure on health in 47 African countries
4
2. CBIs in Rwanda

• In recent years Community-based health insurance
  schemes (CBHIs) emerged in Africa in response to
  failures by the state and the market to provide
  health insurance (e.g. Ghana, Senegal Rwanda)
• CBHIs in Rwanda however are perhaps the largest
  (close to 85 coverage in 2008) and linked to
  national health policy.

5
2. CBIs in Rwanda (contd)

• Some of the features of CBIs in Rwanda include
• Premiums are flat (earlier it used to be
  different across CBIs)
• Members have access to basic health care services
  and medication at a discount rate.

6
2. CBIs in Rwanda (contd)

• Why is CBHIs in Rwanda interesting?
• . Scale up took place in the midst of
  controversy. The pros and cons are as follows
• Rwanda being a poor country and basic health
  services are unaffordable to the majority
  (despite government subsidy), CBHIs are the only
  alternative to increase demand for modern health
  care and reduce illness related consumption risk

7
2. CBIs in Rwanda (contd)

• Others argue that flat rate is inherently
  discriminatory. The insurance premium is high for
  the extreme poor (about 6 of total income) and
  in fact could reduce health service utilization
  due to other layers of expenses. Since
  subscription to the program is not voluntary,
  there is no guarantee that the poor are protected
  from health related income shocks.

8
3. Objectives of the paper

• Do CBIs increase demand for modern health care
  services?
• Are insured households protected from
  catastrophic out of pocket health-related
  expenditure?
• Do the poor fare well compared to the non-poor
  since they contribute proportionately more to the
  system than the non-poor?

9
4. Data

• Nationally representative household survey
  conducted in 2005/06 covering 6900 households.
• The data is standard Living Standard Measurement
  Survey complete with information on household
  demographics, consumption, income, labor market
  conditions, education and health, etc.
• According to the survey 34 of households were
  members of CPIs (39 rural and 22 urban areas).

10
4. Data (contd..)

• 21 of households reported as having fallen sick
  in the previous two weeks of the survey.
• Of these only 30 sought medical care.
• Overall, 20 of households reported positive
  health related expenditure

11
5. Variable definition

• Dependent variables
• Dummy if a household sought treatment from health
  providers after reporting sick
• Dummy if a household experienced catastrophic
  expenditure which is defined as top decile of the
  share of health expenditure to total expenditure.

12
5. Variable definition (contd..)

• Covariates
• Dummy if a household was enrolled in community
  based health insurance scheme (key variable of
  interest)
• Age, size of household, sex of head, level of
  education, real consumption expenditure in adult
  equivalent, district dummies, urban dummy,
  disability status, etc..

13
6. Estimation issues

• Membership in CBHIs is very likely not random so
  that there is a real possibility of households
  self-selecting into the system which introduces
  biases into its effect on the dependent
  variables.
• One example is sick people self-selecting into
  the insurance system
• Or the flat premium provides built-in incentives
  to well-off households
• Well-run districts get far more members than
  weaker districts, etc..

14
6. Estimation issues(continued)

• Generally membership to CBIs was driven by the
  following factors
• Household consumption quintile (richer households
  tend to enroll)
• Demographic factors are important Male headed
  households, large families and older family heads
  tend to enroll into CBIs.
• Some districts have significantly higher
  enrollment than others
• But, there is also substantial pressure from
  local administrators that may be correlated with
  the above variables (the higher the stake, the
  higher the rate of compliance-richer and educated
  households tend to comply more than poor ones,
  etc. )

15
7. Empirical method

• Broadly speaking, the empirical literature uses
  two approaches to deal with the above research
  questions econometric models (regression
  approach) and the matching estimator commonly
  used in the evaluation literature though
  conceptually the two are related

16
7. Empirical approach (contd..)

• The general specification of the econometric
  model follows the latent variable approach with
  endogenous dummy regressor (Heckman, 1974 and
  others) (see equation below)

17
7.1 regression approach (a bi-variate discrete
choice model)
18
7.1 regression approach (contd..)

• It is safe to assume that membership to the CBIs
  is endogenous in the econometric model for a
  number of reasons (s12 0)

19
7.2. Matching estimator

• This is a popular method used extensively in the
  evaluation literature. Some focus on before and
  after a program and often most focus on with
  and without a program
• The idea is to create a treated versus
  control group that would be matched on the
  basis of specified household and community
  characteristics.
• Works well when the bias introduced by unobserved
  factors are minimum.

20
8. Discussion of results

• Finding instruments that affect health
  utilization and health related risk only through
  membership to insurance is not easy.
• We identified two potential instruments. One is
  cluster level enrollment rate (to isolate some of
  the confounding factors in individual decision)
• and the other is a dummy whether or not a
  household owns title deeds for land ownership
  (proxy for well-run districts)

21
Table 1 marginal effects of membership to
Mutuelles on selected variables simple probit
(with Blundell-Smith, 1986 test for
Weak-exogenity)
22
Table 2 Average treatment effect of being
insured on selected variables using Matching
Estimator
23
6. Discussion of results (contd..)

• Conditional on other household and community
  level covariates (age, sex of head of the
  household, educational attainment, dummies for
  district, dummies for serious illness,
  occupation, etc) we find that membership to CBHIs
  have significant and positive impact on
• Utilization of modern health care and protection
  of households from catastrophic health related
  expenditure, which is indeed reassuring.

24
6. Discussion of results (contd..)

• Generally the poor do not seem to have come out
  badly, though the non-poor seem to have better
  access.
• Catastrophic expenditure is not any different
  between insured and uninsured among households
  that reported sick.
• Generally the results with the Matching Estimator
  are very comparable and consistent (see Table 2)

25
6. Discussion of results (contd..)

• Despite the weak result on the effect of CBIs on
  health related expenditure, it is possible to see
  that generally insured households have less
  health related expenditure risk than uninsured
  households (see Figure 2 and Figure 3 below)

26
Figure 2 Health expenditure profile for the
uninsured
27
Figure 3 Health expenditure profile for the
insured
28
Summary and conclusions

• CBIs in Rwanda play an important role in
  increasing demand for modern health care
  controlling for other factors. In general,
  membership increases health service utilization
  by about 17 more among the non-poor than the
  poor.
• When illness strikes, the CBIs seem to protect
  member households from catastrophic expenditure.

29
Summary and conclusions

• The potential of CBIS seem to be very high among
  the non-poor than the poor in both cases that may
  reinforce the inequity inherent in the system.