Who is the vulnerable child

1
Who is the vulnerable child?

• Using data from DHS and MICS to identify
  vulnerable children in the era of HIV/AIDS

2
A problem of definition at the global level

• Problem
• We know that HIV and AIDS makes children
  vulnerable
• We agree to work to ameliorate the effects of
  that vulnerability
• We struggle to define the vulnerable child in the
  context of HIV/AIDS
• Ramifications
• We cannot reliably identify the population at
  risk
• We are unable to measure effects of interventions
• We are unable to discern trends in prevalence of
  vulnerability

3
A problem of definition at the global level

• Whose definition?
• Local definitions of vulnerability are critical
  for effective program implementation, yet
  difficult to harmonize measurement at the global
  level
• Effects of a global AIDS epidemic on child
  vulnerability demands a globally-coordinated
  response
• Multinational organizations need cross-nationally
  applicable definitions and indicators to monitor
  and evaluate service coverage and progress
  towards international goals and targets
• Global donors need evidence that their
  investments are having the desired effect for the
  appropriate population in a cost-effective manner

4
So, what have we been doing until now?

• Orphaning and living arrangements
• the most frequently-used markers of child
  vulnerability
• however, empirical evidence supporting their use
  has been equivocal.
• Chronic illness or HIV serostatus of adult
  household members
• Intuitively appealing, but can they help identify
  children at risk of poor outcomes?

5
Purpose of the research

• To explore the utility of frequently-used
  indicators of child vulnerability by
  investigating two questions
• Are orphans worse off than non-orphans?
• Are children who co-reside with chronically ill
  or HIV-positive adults worse off than those who
  dont?

6
Organization of the rest of the presentation

• Data methods
• Definitions of child outcomes (dependent
  variables)
• Results
• Are orphans worse off than non-orphans?
• Are children who co-reside with chronically ill
  or HIV adults worse off than those who dont?
• Summary
• Conclusions/next steps

7
Data Sources and selection criteria

• Data sources
• MICS (Multiple Indicator Cluster Surveys) ?
  UNICEF
• DHS (Demographic and Health Surveys)
• AIS (AIDS Indicator Surveys (AIS) ?
  USAID/Macro
• Selection of countries to focus on those most
  affected
• orphan prevalence gt 8 percent OR adult HIV
  prevalence gt 1 percent
• and
• conducted a DHS/AIS in 1995 or later, or
  conducted a MICS-3 (2005-06)
• and
• data were available by June 2008

8
Data Sample sizes and regional coverage

• Sample sizes
• Range 4,368-40,511 households
• Average 9,638 households
• Median 8,380 households
• Regional coverage (38 countries)
• West and Central Africa 16 countries
• Eastern and Southern Africa 15 countries
• Latin American and Caribbean 5 countries (4 are
  Caribbean)
• Other 2 countries (Ukraine and Thailand)

9
Bivariate analysis analyzes 2 variables at a time
10
Multivariate analysis uses multiple explanatory
variables in effort to explain outcome
Child is undernourished

• Age
• Urban/rural
• Wealth
• Sex
• Recent illness
• Mothers education
• Fathers education
• Orphanhood status/living arrangements
• Number of siblings
• Mothers empowerment
• Recent illness (diarrhea)

11
Data Availability according to type of analysis

• Bivariate analysis (simple correlations, with
  chi-square test plt0.05) data are used from 36
  surveys from about 2001-2007 whichever surveys
  included data on the outcome and key explanatory
  variables were included in the bivariate
  analyses.
• Multivariate analysis (accounts for multiple
  factors that could also explain the outcome
  plt0.10) requires a larger set of variables to be
  common across all countries thus, fewer surveys
  were included. Final set of surveys include
• Cameroon 2004 Côte dIvoire 2005 (AIS)
• Haiti 2005 Malawi 2004
• Mali 2006 Rwanda 2005
• Tanzania 2003 (AIS) Uganda 2004 (AIS)
• Zimbabwe 2005-2006

12
Definition of child outcomes

• Wasting (weight-for-age).
• Among children aged 0-4 years old, whether they
  are considered underweight, i.e., lt2 standard
  deviations below the median of the new WHO Child
  Growth Standards. (0/1)
• School attendance.
• Among children aged 10-14 years old, whether
  they have attended school in the past year. (0/1)
  
• Early sexual debut.
• Among girls and among boys age 15-17, whether
  their first sexual intercourse occurred before
  age 15. (0/1)

13
Are orphans worse-off than non-orphans?

• Analysis 1

14
MethodsAre orphans worse-off than non-orphans?

• Key explanatory factor whether child is an
  orphan
• Bivariate analysis
• 0 no parent has died
• 1 mother, father, or both parents have died
• Multivariate analysis multi-category variable
  that covers all combinations of orphan status
  and living arrangements of child reference
  category is no parent has died, living with both
  parents
• Other explanatory factors of interest
• Multivariate models also contain household,
  community-level, health status, demographic, and
  education-related variables

15
Nutritional status Percent wasted (lt-2SD from
WHO ref median)Percent of children age 0-4 who
are wasted, by orphanhood status, MICS DHS
2001-2007
16
School attendance Percent attended school in
past yearPercent of children age 10-14 who
attended school in the past year, by orphanhood
status, MICS DHS 2001-2007
17
Boys Percent with early sexual debutPercent of
boys age 15-17 who first had sex before age 15,
by orphanhood status, MICS DHS 2003-2006
18
Girls Percent with early sexual debut Percent
of girls age 15-17 who first had sex before age
15, by orphanhood status, MICS DHS 2003-2006
19
Multivariate resultsAre orphans worse off than
non-orphans?

• Wasting There is very weak if any evidence that
  guardianship or orphaning are important
  determinants of wasting in children age 0-4
  years. Wealth was consistently significantly
  associated with wasting in all six countries
  analyzed.
• Schooling Lack of guardianship by a blood
  relative is the factor that is most frequently
  correlated with lower school attendance. In
  nearly all countries, wealth was strongly and
  significantly associated with school attendance,
  as was education of the head of the household or
  of the eldest female household member (in most
  countries, effect of educ of oldest woman is
  stronger than that of head of hh).
• Early sexual debut Few to none of the orphaning
  and living arrangement variables were associated
  with early sexual debut among boys or girls the
  models do not explain variation in early sexual
  debut well. Among girls, orphaning and living
  arrangements were not statistically important for
  any country except Uganda.

20
Odds of wasting wealth quintile Among children
age 0-4, DHS 2004-2006
21
Loss of a second parent is more likely than the
loss of a first parent Programmatic implications
22
Key pointsAre orphans worse-off than
non-orphans?

• Orphans are not significantly worse off than
  non-orphans on all three outcomes of child
  well-being
• In most countries, orphans do not have
  significantly worse outcomes than non-orphans.
  In these countries, targeting resources to
  children based on orphan status may not the best
  approach to using scarce resources.
• Small differences may nevertheless be meaningful
• Even if lacking in statistical significance, the
  consistency with which orphans are more likely to
  experience poorer outcomes on these selected
  indicators of vulnerability especially as
  demonstrated in the bivariate suggests that
  while orphanhood alone fails to definitively
  identify a child as vulnerable, it likely
  nevertheless contributes to a childs
  vulnerability profile.
• National context matters
• Countries with high levels of overall wasting
  did not have large differentials between orphans
  and non-orphans. At high levels of overall
  school attendance, both orphans and non-orphans
  attended school in nearly the same percentages.
  Differences between orphans and non-orphans were
  more distinct where levels of the respective
  outcomes were lower.

23
Are children living with a chronically ill or
HIV adult comparatively worse off?

• Analysis 2

24
Methods Definition of key explanatory variable

• Health status of adult household members
• No chronically ill or HIV household member (ref.
  category)
• HIV adult in household
• Chronically ill adult in household
• Both HIV and chronically ill adult in household
• Variable based on two pieces of information from
  DHS surveys
• Chronic illness among adult (age 18-59 years)
  members of the household Has (NAME) been very
  sick for at least three months during the past 12
  months? By very sick, I mean that (NAME) was too
  sick to work or do normal activities around the
  house for at least 3 of the past 12 months.
• HIV-positive serostatus of adult members of the
  household Established through anonymously linked
  biomarker data after obtaining informed consent,
  dried blood spots are obtained in the field,
  taken to lab, and anonymously tested for HIV
  antibodies. The results are then linked through
  use of bar codes to the rest of the demographic
  and other health data. (DHS)

25
Note Most chronically ill adults are not HIV
(Barrere 2007)
26
Odds of wasting illness or HIV in the household
Among children age 0-4, DHS 2004-2006
27
Odds of school attendance chronically ill or
HIV-positive adult in the household Among
children age 10-14, DHS 2004-2006
28
Odds of early sexual debut (boys) Among boys age
15-17, DHS 2004-2006
29
Odds of early sexual debut (girls) Among girls
age 15-17, DHS 2004-2006
30
Key pointsAre children living with a
chronically ill or HIV adult comparatively worse
off?

• In 2 out of 6 countries, having an HIV adult in
  the household increases the odds of childhood
  wasting.
• In 7 out of 9 countries, adult chronic illness or
  HIV serostatus decreases the odds of school
  attendance.
• In 3 of 8 countries, adult chronic illness
  increases the odds of an early sexual debut for
  girls for boys, it is only in 1 of 8 countries
  where this holds true.
• In sum having a chronically ill or HIV adult in
  the household is fairly consistently associated
  only with school attendance.

31
Summary thus far

• Orphaning, living arrangements, and adult health
  are not useful on their own in identifying
  vulnerable children, even in countries with high
  HIV prevalence.
• Household chronic illness or HIV-positive
  serostatus
• Lack of familial guardianship (though not
  necessarily orphaning)
• ? These measures were consistently associated
  with poor outcomes only for school attendance
• Other indicators of vulnerability were more
  frequently associated with poor outcomes. For
  both research questions
• Household wealth status was associated in the
  great preponderance of cases for wasting and
  school attendance.
• Poor education of adult household members was
  significantly associated with low school
  attendance among children.
• Models of early sexual debut not very useful

32
Are there good alternative measures of
vulnerability?

• Analysis 3

33
Methods Type of statistical analysis used

• Are there good alternative measures of
  vulnerability?
• Analysis bivariate only counting exercise, no
  statistical testing
• Outcome measures underweight, school attendance,
  early sexual debut
• Key explanatory factor s large set of
  theoretically-plausible indicators of
  vulnerability see following slides
• Surveys used Same 9 surveys as were used in
  multivariate analysis
• Cameroon 2004 Côte dIvoire 2005 (AIS)
• Haiti 2005 Malawi 2004
• Mali 2006 Rwanda 2005
• Tanzania 2003 (AIS) Uganda 2004 (AIS)
• Zimbabwe 2005-2006

34
Counting exercise
35
Conclusions

• Standard OVC-related indicators of vulnerability
  by themselves are not very satisfying for policy
  programmatic purposes
• Wealth and parents education are better
  correlates of child health and well-being
  outcomes
• Models of early sexual debut not very useful
• The analysis results support a multivalent
  approach to defining vulnerability which
• Incorporates age-specific vulnerabilities
• Tries to capture multiple aspects of
  vulnerability
• Is limited to data that are commonly collected in
  surveys like MICS DHS
• Even if we conclude that orphanhood is not the
  best or only marker of vulnerability, there is
  still value in continually assessing the status
  of orphans.

36
Thank you!
37
Data Regional coverage
38
MethodsAre orphans worse-off than non-orphans?

• Analysis bivariate multivariate (logistic
  regression)
• Outcome measures underweight, school attendance,
  early sexual debut
• Key explanatory factor whether child is an
  orphan
• Bivariate analysis
• 0 no parent has died
• 1 mother, father, or both parents have died
• Multivariate analysis multi-category variable
  that covers all combinations of orphan status
  and living arrangements of child reference
  category is no parent has died, living with both
  parents
• Other explanatory factors of interest
• Multivariate models also contain household,
  community-level, health status, demographic, and
  education-related variables

39
Vulnerability measures Household composition,
child endowments and household endowments

• Child endowments
• Child does not own a blanket
• Child does not own shoes
• Child does not own an extra set of clothes
• Child does not own all three items (blanket,
  shoes, extra set of clothes)
• Child does not sleep under a mosquito net
• Household endowments
• Household head has lt primary education
• Best educated adult woman has lt primary education
• Household in lowest 40 of national wealth
  distribution
• Household in lower half of community wealth
  distribution
• Household does not own a radio
• Household does not own a television
• Less than half of the adults in the hh have at
  least a primary education
• At least one age-eligible child in the household
  is not in school
• Rural residence

40
Vulnerability measures Household composition,
community measures chronic illness in the
household

• Household composition
• Household dependency ratio gt 1
• Female-headed household
• Crowded household
• Community measures
• Community is poorer than the national average
• Community is less accepting of HIV people than
  the national average
• Community has a high level of orphaning
• Community has a high level of OVCs
• Chronic illness or HIV in the household
• Child lives in a household with an HIV person
• Community HIV seroprevalence is higher than
  national average
• Child lives in a household with a chronically ill
  adult
• Child lives in a household in which a person died
  in the past year after experiencing at least 3
  months of illness
• Childs parent has been chronically ill in the
  past year

41
Vulnerability measures Family characteristics,
orphaning and living arrangements

• Family characteristics
• Childs mother was 18 years old or younger at
  childs birth
• Childs mother has parity 4
• Mother has less than a primary education
• Childs birth is not registered
• Orphan status
• Child is a maternal orphan
• Child is a paternal orphan
• Child is a double orphan
• Child is an orphan (maternal, paternal or double)
• Living arrangements
• Child does not live with both parents
• Child does not live with mother
• Child does not live with father

42
Results Alternative measures of vulnerability

• So, what are the best overall measures associated
  with vulnerability?
• household wealth and education of adults in the
  household
• almost invariably positively related with good
  outcomes for all three indicators of
  vulnerability
• lower probability of wasting
• higher school attendance
• reduced early sexual debut
• Measures associated with underweight The health
  of household members
• Measures associated with school attendance
• Orphan status
• Living arrangements
• Child ownership of blankets, clothes and shoes
• Measures associated with early sexual debut
• Child ownership of blankets, clothes and shoes
• Living arrangements