HOUSEHOLD in the Demographic and Health Surveys:
Experiences from research on health inequity
Tanja AJ Houweling Dept. Epidemiology & Public Health, UCL
Commission on Social Determinants of Health
Tanja AJ Houweling
Outline presentation
1. Health inequalities research– use of concept of household – focus on inequalities in childhood mortality in low/middle
income countries, DHS data
2. Problems with use of ‘household’ in DHS
3. Implications for policy making & research
1. HEALTH INEQUALITIES RESEARCH
Tanja AJ Houweling
Under-5 mortality rate
Source: www.who.int/
Tanja AJ Houweling
Under-5 mortality by wealth group
0
50
100
150
200
250
300
Und
er-5
mor
talit
y ra
te
Peru Indonesia Mali
RichestFourthMiddleSecondPoorest
Tanja AJ Houweling
The average is not the only messagedistribution
Reducing health inqualities within countries: increasingly important policy objective
Tanja AJ Houweling
Health inequalities research
Comparing health outcomes between socially stratified groups on basis of characteristics measured at level of:– individual (education, occupation)– household (economic status: wealth, income)
Child health low/middle income countries:– maternal education, household wealth
Tanja AJ Houweling
Demographic and Health Surveys
Set up as primary source health & population data
1984 - present75 low and middle income countries in Africa, Asia, Latin America, North Africa/Near East
nationally representative hh surveyswomen aged 15-49
child mortality (retrospective birth histories)
direct determinants: e.g. immunizationeducation & household assets
Tanja AJ Houweling
Household ownership assets:durable consumer goodshousing characteristicswater and sanitation facilitieselectricity
Principle Components Analysis
Household wealthIncome/expenditure data unavailable
Relative position in national wealth hierarchy
Some problems. Further research needed
Tanja AJ Houweling
Under-5 mortality rate by wealth group
0
50
100
150
200
250
300
350
poorest quintile
richest quintile
average MozambiqueUzbekistan
Cameroon
Namibia
Bolivia
Mali
Source: Houweling 2007
Tanja AJ Houweling
0
10
20
30
40
50
60
70
80
90
100
Bangladesh
Nepal
Chad
Niger
Pakistan
Morocco
Nigeria
India
Guatemala
Uganda Mali
Burkina Faso
Ghana
Mozambique
Kenya
Côte d’Ivoire
C.A.R.
Haiti
Senegal
Zambia
Tanzania
Bolivia
Madagascar
Indonesia
Togo
Comores
Malawi
Peru
Philippines
Cameroun
Benin
Nicaragua
Paraguay
Namibia
Zimbabwe
Turkey
Viet Nam
Colombia
Brazil
Dominican Rep.
Uzbekistan
Kyrgyz Rep.
Kazakstan
Niger
Peru
Richest quintile
Poorest quintile
India
Skilled delivery attendance by wealth group
Source: Houweling 2007
Tanja AJ Houweling
Irregular patternsUnder-5 mortality by wealth group
0
50
100
150
200
250
Und
er-5
mor
talit
y ra
te
Chad
RichestFourthMiddleSecondPoorest
2. SOME PROBLEMS
Tanja AJ Houweling
Why measure economic status at household level in research on child health?
Theoretical: “household production of health”: households as a locus of provision of care & production of health. Level at which resources are pooled for daily needs & decisions that influence child health are made.
Practical: no info. on income/expenditure levels in DHS. Assets: assumed to be characteristics at level of household
BUT: does the DHS definition correspond with ‘household production of health’ model?
Tanja AJ Houweling
DHS definition household
“A person or group of people that usually lives andeats together”.
– If people eat in one household and sleep in another: “consider the person to be a member of the household where he sleeps” (source 1)
Source 1: DHS Interviewer’s manual
Tanja AJ Houweling
i. Definition & lived reality
Details on definitions & implications in specific contexts: not readily available in DHSAssessing extent of problem difficult– Polygyous compounds: one or
separate households?– Complex living arrangements /
resource pooling arrangements– Differences in operationalisation
of ‘household’ between countries?
Could this partly explain cross-country differences in household structure?
Tanja AJ Houweling
i. Definition & lived reality
Co-residential group: not necessarily the same as group that poolsresources for daily living -> problems measurement economic status
Eg. sisters with respective husbands and children living under one roof, but not pooling resources for food (Jakarta)
If asset of both sisters would be included in the household asset index: - overestimation of economic resources available for child health - understimation mortality inequalities
Tanja AJ Houweling
i. Definition & lived reality Household as process
DHS is static: no info on household formation & dissolution Household economic status: associated with household composition – e.g. poor households in Jakarta send family members
away to live in other households as part of survival strategy. Richer households are able to take up others (distant relatives, domestic workers, etc).
– Research DHS SSA: nuclear families: poorer, lower educated adults, children worse health, lower immunisation coverage and ORT use
Need indepth research – processes that may influence child health
Tanja AJ Houweling
ii. Intra-household diversity
Economic status measured at household level -> all household members attributed same economic statusViz. definition social class in Britain: occupation male household head to determine social class all household membersdon’t take intra-household diversity into account
Intra household allocation: – Unequal allocation, for example along gendered lines – Allocation for child health: maternal education & income ownership
important role
Need to look at what goes on within the household
Tanja AJ Houweling
iii. Excluded groups
Sampling households/dwellings:
Socially excluded groups: homeless, street children, refugees, prisoners, mobile groups: not in surveys – > perhaps underestimation of
mortality inequalities, depending on relative size of this excluded group
Tanja AJ Houweling
iv. Fit for purpose?
Data and methods (incl definitions) should be fit for question asked (*)Need comparable definition of household across countries & time periods to answer some important questions for which DHS has been set upInequalities: broad patterns, key findings not explained by definition problemsSpecific detailed patterns, outliers, perhaps explained? Other research questions: standardized definition as in DHS less useful.
(*) Measurement and Evidence Knowledge Network CSDH
3. IMPLICATIONS
Tanja AJ Houweling
i. Implications for surveysNeed more detailed & accessible documentation: definitions + general and country specific implications of definitions
– definition: rural/urban
Excluded groups: esp. in countries where relative size of these groups is large (monitoring)
– Targeted studies that suplement the surveys?– Information should be publicly acessible, along side larger survey
results.
Measure wealth at level of mother in addition to hh wealth??– importance of money in hands of mother
Surveys: distinction residential unit & cooking unit?
Tanja AJ Houweling
ii. Implications for interventions
Reaching the poor & targetting: Who is poor? Who is not?
Intra-household allocation– Be careful with targetting: problems with identifying
poor women & children – Measure economic status at level of the mother?
Tanja AJ Houweling
iii. Implications for research: the need for more inter- and multidisciplinary research
Household – strong argument multi and interdisciplinary research:
Social epidemiology: broad patterns/general trends. National level generalizability. household as black box. Certain questions can only be answered by unpicking this black box (Anthropology)– Who pools resources for daily needs? – Dynamics of household composition (formation, change, dissolution of
households) as consequence of social change (economics; disease (HIV/AIDS)
– Explaining outliers & irregular patterns. Due to definition household? – In which countries is DHS definition of household especially a problem?
Also other concepts western bias: e.g. socio-economic status. Need to deconstruct this concept, in context of low/middle income countries