Similar to Child Poverty Research Day: Reducing Non-Economic Poverty - Anja Sautmann, 'Subsidies, Information, and the Timing of Children's Health Care in Mali'
Similar to Child Poverty Research Day: Reducing Non-Economic Poverty - Anja Sautmann, 'Subsidies, Information, and the Timing of Children's Health Care in Mali' (20)
Child Poverty Research Day: Reducing Non-Economic Poverty - Anja Sautmann, 'Subsidies, Information, and the Timing of Children's Health Care in Mali'
1. Subsidies, Information, and the Timing of
Children’s Health Care in Mali
Anja Sautmann, Samuel Brown
(Brown University)
and Mark Dean
(Columbia University)
Child Poverty Research Day
November 18, 2016
3. Introduction
I Lack of adequate care for acute illness contributes to continually
high child mortality rates
I Broad policy swing in primary care for children: from “user fees”
(Bamako Initiative) to free access (e.g. Burkina Faso 2016)
I New focus on acute care and urban areas
I Debate over subsidies for acute care:
I Absent other distortions, subsidies can cause overuse and waste:
price is a measure of value
I But subsidies can overcome underuse if there are access barriers or
inefficiencies: lack of access to credit, lack of information, benefits
not taken into account by parents (child welfare, longrun health,
infection risk)
I Complementary policy: health education and information to
encourage efficient use
Question:
Can subsidies, supplemented with information policies, curb underuse of
care without creating overuse (in urban populations)?
4. Introduction
I Two common health
policies:
I Biweekly healthworker visits
I Subsidies for “5 killers of
children” “Action for Health”
NGO Mali Health
I Idea of “integrated care” and
information as a tool to
optimize healthcare use
I Following guidelines of
Integrated Management of
Childhood Illness (IMCI) by
WHO/Unicef
5. Introduction
What constitutes overuse and underuse?
I Value of care depends on (often unobserved) health status
I In this paper:
I Model and estimate timing of care within an illness spell
I As benchmark for overuse/underuse use guidelines of
Integrated Management of Childhood Illness (IMCI) by WHO/Unicef
Supply-side effects of large-scale demand change
I In this paper: randomized control trial, supply-side fixed
7. Model: Timing of Care
I Assume a child in an ongoing spell of (given) symptoms
I If illness absorptive, or full information about its course: either go to
doctor immediately, or never.
Intuition for delaying a visit:
I Initially, child may recover on her own; can save a visit
I If symptoms do not abate: probability of not recovering is increasing
over time
) Longer illness is more likely “serious”
8. Model: Timing of Care
Show: optimal “care seeking strategy” specifies after how many days to
go see a doctor
I Depends on
I seriousness of the illness/symptoms
I cost vs. benefit of a visit.
I Parents may disagree on the optimal choice
) Parents seek care too early or too late
I Subsidies: reduce the cost threshold and lead parents to seek care
earlier
I Information: can teach parents about the optimal action according
to policy
9. Predictions
1. Free care leads to earlier care: can reduce underuse, but may
increase overuse.
2. Better information
2.1 can reduce underuse and overuse
2.2 but may increase underuse if parents do not agree with the
information
3. Free care and better information may be complements: potential to
reduce underuse without creating overuse.
) Motivates policies that combine the two (e.g. IMCI)
11. Data: Action for Health RCT
Fall 2012: baseline survey
I Location: Sikoro, peri-urban area of Bamako, Mali
I 650 compounds, 1544 children; below local poverty line Attrition
I Two public health clinics provide basic primary care – “CSCom”
I Large households (>6 members), USD 63 weekly income, 50%
literate, undernourished children (-0.61 W4H z-score)
Typical of the fast-growing population of urban poor in Sub-Saharan
Africa
Urban setting means better care, but also a risk for overuse!
13. Data: Action for Health RCT
January 2013: independently randomize
1. “Free” care: free services and medications for children under 5 at
local CSCom
only for diarrhea/malnutrition, malaria, vaccinable disease,
respiratory disease
2. Healthworker visits: monitor health, teach symptoms, and guide
use of formal care – e.g. accompany to clinic – based on IMCI
standards when care is required
Fall 2013: 10-week follow-up survey
I Formal consultations: 514 CSCom, 67
other (private); average cost of USD 5-10
I Symptom records
15. Data: Spells and Need for Care
Spell: contiguous period of symptoms, ending with doctor visit or
recovery.
Policymaker preference:
I Unicef/W.H.O.’s Integrated Management of Childhood Illness
(IMCI)
Classify symptom days in the spell as “early” for care or “care
required”
I Example:
I Diarrhea < 5 days: home remedies
I Diarrhea with blood in the stool: immediate care (dysentery)
17. Outcomes: Unconditional Utilization
I Subsidies
I decrease CSCOM visit costs by 70% (2964 to 893 CFA on average)
I increase formal demand by 317% per child (from 0.18 to 0.57 visits)
I Healthworkers
I have little average demand effects.
18. Outcomes: Over- and Underuse
early
care
required early
care
required early
care
required early
care
required
374 407 327 463 368 430 353 438
% with a consultation 3% 11% 2% 10% 6%** 31%*** 8%*** 27%***
Significance levels: *** 1%, ** 5%, * 10%, t-test on mean difference from control.
# spells that did/did not enter
"care required"
Control Healthworker Free care HW & FC
I Control and HW only groups:
I Rampant underuse, no overuse
I Subsidies
I remaining underuse of at least 69% of “care required” spells
I Healthworkers have no clear effects
I Proportion of consultations that are overuse constant at about
16%.
19. Day by day probability of care seeking
0%
2%
4%
6%
8%
10%
12%
14%
1 2 3 4 5 6 7 >7
Probability of Formal Care
Spell Day
Control
Early
Care required
0%
2%
4%
6%
8%
10%
12%
14%
1 2 3 4 5 6 7 >7
Probability of Formal Care
Spell Day
Healthworkers only
Early
Care required
0%
2%
4%
6%
8%
10%
12%
14%
1 2 3 4 5 6 7 >7
Probability of Formal Care
Spell Day
Subsidy only
Early
Care required
0%
2%
4%
6%
8%
10%
12%
14%
1 2 3 4 5 6 7 >7
Probability of Formal Care
Spell Day
Subsidy and Healthworkers
Early
Care required
21. Implications for Overuse and Underuse
1. Overuse:
I Near zero probability of care-seeking on “early” days
I No additional reduction by healthworkers
I Subsidies have only small effect on demand on early days
2. Underuse:
I Subsidies increase care-seeking, but to at most 14% daily probability
I Healthworkers decrease use by 37% (10% significance)
Implications:
I Parents can discern “early” days
I Substantial underuse even with subsidies
I Healthworkers increase underuse, as predicted when parents have
high cost threshold
Cost/benefit is the binding constraint; information not the main barrier
22. Predicting Care Seeking For Other Disease Environments
I Care-seeking probabilities based on symptoms: allow out-of-sample
predictions
I Here: use hemorrhagic fever spell descriptions to code set of typical
symptom spells
I Predict proportion without care for each spell day
Model 1, group HWFC
Model 1: each day classified as early/care-required according to C-IMCI.
Model 2, group HWFC
Model 4: Indicators for disease com
diseases (i.e. generalized fever, ma
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5 6 7 8 9 10
Ebola: pred. proportion without formal care
(Model 1: early/care-required)
Model 1, group C Model 1, group FC
Model 1, group HW Model 1, group HWFC
Ebola: pred. proportio
(Model 3: C-IMC
0.5
0.6
0.7
0.8
0.9
1
1 2 3 4 5
Ebola: pred. proportio
(Model 4: day-t
Model 4, group C
Model 4, group HW
Ebola: pred. proportion without formal care
(Model 2: symptom-specific early/care-req.)
Result: in an (undetected/unexpected) Ebola outbreak, subsidies would
lead to 20-30% higher use of formal care by day 5.
23. Other Results
Health Outcome Effects of subsidies:
I Average illness spell length reduced by 0.8 days – recall, only 30%
receive a visit!
I Mothers self-report significantly less worried about their children;
20% of days instead of 29% of days
25. Summary of Results
I Open the black box of healthcare demand, estimate timing of care
conditional on illness incidence
I Results encouraging for opponents of user fees:
I Families recognize need for care
I Overuse and moral hazard not a primary concern
I Unintended consequences of (only) providing information
26. Policy Relevance
I Immediate policy impact:
I Changes to the programs of our cooperating partner Mali Health
I Focus on subsidies, re-focus health workers onto prevention
I Many open questions:
I How were care-seeking guidelines formulated?
I Should we trust parents’ observations, but not their decisions? How?
I Can we get more data, and how to use it?
I Broader lessons for child poverty and healthcare access
I Urban healthcare is different
I Access 6= use; parents as gatekeepers of children’s use of resources
I Non-monetary costs of care are important: mutiple dimensions of
scarcity