All Time Service Available Call Girls Marine Drive 📳 9820252231 For 18+ VIP C...
Bruno08 10 Lindelow The Impact Of Health Insurance In Rural China
1. The Impact of Health Insurance
in Rural China: Evidence from
the New Cooperative Medical
Scheme
POVILL Conference, Vientiane, October 2008
Magnus Lindelow, World Bank
Joint work with:
Adam Wagstaff, World Bank
Gao Jun, CHSI, Ministry of Health, China (now WHO, Manilla)
Xu Ling , CHSI, Ministry of Health, China
Qian Juncheng , CHSI, Ministry of Health, China
2. Background
Many countries are experimenting with reforms to
expand health insurance to the informal sector
Subsidized participation in national schemes (e.g. Colombia,
Philippines and Vietnam)
Separate subsidized contributory scheme (e.g. Mexico)
Separate fully subsidized scheme (Thailand)
China introduced a new health insurance scheme in rural
areas in 2003
Initially rolled out in around 300 counties—rapidly expanding
towards national coverage
Aim of research is to evaluate early impact of scheme on
use of health services and health expenditures
3. China’s New Cooperative Medical
Scheme (NCMS)
‘Old’ CMS collapsed during 1980s—at same time health care costs
increased dramatically
Resulted in high incidence of catastrophic spending and increased
barriers to utilization
Aims of NCMS are to reduce impoverishment resulting from illness
and encourage utilization of needed care
Key features of NCMS
County-based (200-300K population)
Voluntary participation (~80% in most counties)
Variation in design and implementation across counties
In principle focus on IP; in practice OP also covered in most counties
Few measures to control costs
Financing (at time of evaluation—fall of 2005)
Flat-rate household contribution (min. 10 RMB / USD 1.3)
Government subsidy (min. 40 RMB / USD 5.5)
Small share of overall p.c. rural health spending (~250 RMB)
4. Evaluation question and approach
Evaluation question
What has been impact of NCMS on health expenditures and use of
health services?
Also component based on facility level data—not covered here
Pre-intervention data from 2003 National Health Survey (MOH)
Post-intervention data from new household survey (Sept. 2005)
10 counties with NCMS, 5 without NCMS (purposively selected)—8,476
households
Also broader cross-section sample
Control group: households in “non-exposed” counties
Alternative approach: non-participating households in NCMS counties
Evaluation of impact though combination of differences-in-
differences (DD) and matching
DD to deal with unobserved (time-invariant) heterogeneity
Matching to control for observed heterogeneity
Estimate impact for sample & individual income deciles
Quantitative analysis complemented with qualitative work
5. Methods
Propensity score matching (PSM) to match treated &
untreated HHs and estimate average treatment on the
treated (ATT)
Probit analysis (1=NCMS member, 0=HH in non-NCMS county)
Nearest neighbor matching (5), with caliper (requiring
sufficient “closeness”)
Significant differences between NCMS and non-NCMS
counties
Poorer and more rural counties
Matching achieves good balancing on observables,
except when treated HHs are matched with HHs in non-
NCMS counties
Some bias on observables remain
Better matching if use non-members in NCMS counties
But problem of unobservables
6. Use of services
OP visit last 2 weeks 52%
OP visit village clinic 56%
OP visit THC 21%
OP visit county hospital 81%
IP episode last 12 months 42%
IP episode: THC 24%
IP episode: county hosp. 53%
# IP episodes (12 months) 119%
Expenditures
HH OOPs (12 months) 61%
OOP per OP visit 236%
OOP per IP episode 103%
OOP delivery -61%
ATT (%
change)
OP visits have increased in
village clinics (the poorest
Q) and county hospitals
(other Q)
IP visits have increased (no
significant differences
between bottom and higher
Qs)
Increase in OOPs,
particularly OP visit;
decline in OOPs for
deliveries. Impact on OOPs
less pronounced among
bottom Q
Key findings from HH data
7. Conclusions
Increase in utilization consistent with findings from other studies
Most studies from other countries find reduction in OOP—China
seems to be different
Implications?
Welfare gains from improved access must be weighed against welfare
losses from demand- and supply-side moral hazard
Reasons for concern about (supply-side) moral hazard in China
(overuse of drugs and procedures)
Key limitations
Short life of program at time of evaluation (<2 years)
No evidence on impact of health outcomes
Limited insight into how impact varies with design & implementation
Potential bias due to imperfect balancing and possible time-variant
unobservables
Uncertain generalizability due to non-random sample and non-random
program placement
8. Thank you!
For more information on WB work on China’s health system, see:
www.worldbank.org/chinaruralhealth
For copies of paper, email
mlindelow@worldbank.org, with cc to sthitsy@worldbank.org