Midwives and Maternal Mortality: How Effective Has Indonesia

1
Midwives and Maternal MortalityHow Effective
Has Indonesias Village Midwife Program Been?

• Shailender Swaminathan (Brown)
• Tomoya Matsumoto (GRIPS)
• Jeffrey B. Nugent (USC)
• March 2010

2
I. Motivation

• Millennium Development Goal (MDG5) is to reduce
  the maternal mortality ratio (MMR) by three
  quarters between 1990 and 2015.
• Decline in mortality among adults may be more
  important for development than among children
• Given budget constraints and difficulty of
  reducing MMR, quest for cost-effective measures
  for doing so is important
• What program(s) to select?

3
Midwives Programs Commonly Identified as Likely
Candidate

• Historical Precedents
• Sweden in Late 19th Century MMR reduced
  substantially as Midwives spread throughout
  country
• Sri Lanka 1946-60
• Malaysia 1960-1985 with 75 reduction in MMR
• Methodological Shortcomings No controls for
  other factors changing concurrently despite the
  fact that many factors were changing
  simultaneously
• Consensus Recommendations of International
  Experts and Organizations like WHO
• The Lancet Maternal Survival Series Steering
  Group 2006 Maternal Survival 2 Strategies for
  Reducing Maternal Mortality Getting on with What
  Works
• Is This priority justified?

4
Obstacles to Careful Study and Shortcomings in
Existing Studies

• Absence of reliable reports on MMRs since
  majority of deliveries are at home
• Those programs which have been studied have had
  inadequate controls
• Absence of useful randomized trials

5
Dearth of Randomized Trials on Determinants of MMR

• Those randomized trials that do exist on MMRs
• Examine Specific Drugs Effectiveness, other
  health conditions
• Most relevant study a Pakistani study on the
  effectiveness of giving short-term training to
  traditional midwives (as opposed to training new
  midwives for a longer period as in the Indonesian
  case).
• Result by itself at least, this program did not
  significantly reduce maternal mortality in
  Pakistan (Jokhio et al, 2005).

6
II. Indonesia Village Midwife Program

• Motive In mid 1980s Indonesias MMR gt400 per
  100,000 live births
• VMP Established in 1989 to train midwives
• 1 year of training (generally on top of several
  years of nursing)
• Primarily deployed only beginning in 1993
• By 1998 54,000 midwives deployed to underserved
  rural villages
• Midwives per capita increased more that 10 fold,
  from 0.2 per 1000 to 2.6 per 1000 between 1990
  and 1998.

7
Features of Indonesias Program
8
Known for low cost methods Burning the Umbilical
Cord
9
Although Midwives allocated to underserved areas,
Shortcomings Noted

• Hatt et al (2007) Did not reduce urban-rural
  gap in access to emergency care, such as to
  Caesarian sections and even widened it
• Makowiecka et al (2008) VMW services became less
  equitable between less remote and more remote
  areas
• Early studies comparing 1989 with 1994 had shown
  little MMRate reduction

10
Data

• Demographic and Health Surveys in Indonesia of
  1994, 1997 and 2000 Data on Maternal Deaths
• MMRate Maternal deaths at t /100K Women aged
  15-49
• Sisterhood Method (WHO 2004) Females asked to
  identify female siblings born of same mother,
  living or dead. If dead, asked to identify cause
  of death, if during pregnancy or within 2 mos.
  Post-partum.
• This done at national, rural, urban and province
  levels

11
Explanatory Variables

• Fraction of districts with village midwife (VMP)
• at national, rural, urban, province
  levels since 1975 using data from IFLS 1993,
  1997, 2000.
• Other Controls Age of woman, income,
  education of mother, time, alternative service
  provision

12
Regression Model

• MMRater-ut a0 a1time a2VMP a3Xr-u t ut (1)
• Midwife r-ut ß0 ß1time ß2VMP ß3 Xr-u t v t
  (2)
• T 1975, 1976,.2000
• a 2 and ß2 represent difference in effects of
  VMP between rural and urban areas where VMP 0
  before 1993 (diff in diff)
• If other factors constant between pre and post
  program periods, these effects could be causal,

13
RRRRURAL RR
14
(No Transcript)
15
(No Transcript)
16
Results of Table 1

• -12.1 for the VMW Rural Dummy interaction term
  implies that the difference in MMRates between
  rural and urban areas was 12.1 lower post-program
  than pre-program. Pre-program it was 21
  percentage points higher in rural areas
  representing almost 60 of the differential.
• Next (Table 2) a more detailed look at the
  difference in means before and after and
  difference in trends

17
(No Transcript)
18
Results of Table 2

1. Almost 11 percentage point reduction in means of
   w/o adjusting for midwife availability (col. 2)
2. But after controlling for midwife availability,
   only a 3 fall (col. 4)
3. Hence over 70 of the fall in rates can be
   explained by VMW availability

19
Province level Analysis

• Explanatory variables linear time trend,
  fraction of districts with VMW, province level
  fixed effect
• Table 2 shows that VMW availability explains 6
  of intra province changes in MMRate. Use this to
  predict the MMRates by province Figure 3

20
(No Transcript)
21
(No Transcript)
22
Compare Figures 3 and 2

• For the provinces that are most substantially
  rural the drops are very sharp whereas the most
  urban least affected by the VMP, there is
  virtually no decline.

23
(No Transcript)
24
Robustness

• Other access differentials between rural and
  urban areas
• Changing the date of startup of the program

25
Conclusions

• Indonesia has made significant progress to MDG
  Goal for 2015
• But by no means enough to achieve it
• Despite small numbers of data points available ,
  most of the decline achieved seems to be due to
  VMP
• Possible areas for improvement over time
• Possible areas for future research

26
Benefits Costs of VMP

• Much could be done to improve it
• But, even if the payoff in MMRate is deemed
  insufficient to justify it, because of other
  benefits that have been documented, such as on
  birth weights, child health, increasing BMI of
  women, increasing child spacing and reducing
  fertility rates, overall benefits might well
  outweigh the costs.

27
Reasons Offered for the Disappointing Results of
the VMP

• Low quality of training
• Lack of financing and access to hospitals when
  needed
• Lack of sustainability urban born women not
  happy about having to live in villages w/o many
  amenities
• Insufficient incentives for high quality service
  delivery
• Shortages of materials and equipment at local
  level
• Numerous transport and communications problems
  locally
• Conclusion Despite its promise, Not obvious that
  the program has had significant effects on MMR

28
Other Changes Affecting Results

• 1999 Decentralization Program Failures local
  governments revenue constrained, hence often
  cannot pay the midwives and buy the supplies
• 1997-8 financial crisis loss of income,
  inflation, supply shortages weakening provisions
  at local level
• Gradual increase in malaria which would weaken
  pregnant women and make them more vulnerable to
  infection, death during pregnancy and childbirth

29
Shortcomings and Extensions

• MMR data not reliable Could use other
  indicators excessive bleeding at birth, prompt
  referral of complicated cases.
• Use of still other sources of midwife
  availability. How sensitive would results be to
  such alternatives.
• Experience as of 2000 quite short, future rounds
  of IFLS should provide more reliable estimates
• How much substitution among providers has there
  been?
• Has the quality of care by other providers been
  affected by presence of VMW?