Uncovering who is furthest behind
Two decades of maternal and child nutrition data across Cambodia, Lao PDR, and Viet Nam — harmonised into one pipeline that shows whether progress is reaching the people national averages hide.
National averages mask who progress reaches
Alive & Thrive needed to understand how maternal and child health and nutrition outcomes had changed over two decades across Cambodia, Lao PDR, and Viet Nam — and, critically, whether that progress was reaching the most vulnerable populations.
Improvements in antenatal care, breastfeeding, or dietary diversity at the country level may not reflect the reality for ethnic minorities, rural communities, or the poorest households. But the data sat in dozens of separate survey files, collected with different instruments and variable definitions over more than 20 years.
A reusable evidence base — not a one-off report — that A&T could extend to new countries and survey rounds without rebuilding the analysis from scratch.
One config per survey, five equity lenses on every indicator
A3DI built a reproducible, configuration-driven analysis pipeline in Python to harmonise 15+ rounds of MICS and DHS surveys across the three countries, then disaggregate every indicator across five stratifiers and across time.
{
"country": "VNM",
"survey_year": 2021,
"survey_type": "MICS",
"indicators": {
"excl_bf": "BD2",
"cont_bf": "BD3",
"mdd_ch": ["food_grp_1", "food_grp_2", "food_grp_3"],
"mmf_ch": "BD7",
"mad_ch": ["mdd_ch", "mmf_ch"]
},
"equity_stratifiers": {
"wealth_q": "windex5",
"region": "HH7",
"residence": "HH6",
"eth_hoh": "ethnicity",
"mother_edu": "welevel"
}
}Each country-year maps raw survey variables to standardised indicator definitions through a JSON config, not hardcoded logic — so Viet Nam’s 2000 MICS and Cambodia’s 2021 DHS run through the same codebase. Every statistic is weighted for the complex survey designs, and disaggregated by wealth quintile, region, residence, ethnicity of household head, and mother’s education.
Progress was real — but deeply uneven
The gap national averages hid
Lao PDR · between ethnic groupsSchematic of the pattern; 35pp is the real finding — the breastfeeding/dietary-diversity gap between ethnic groups. The national average sits between the two.
The analysis produced a comprehensive equity profile for each country. Wealth-driven disparities persisted across most indicators; access to services like antenatal care and institutional delivery had expanded substantially, but the quality of care within them improved far less.
Regional and ethnic disparities remained substantial — particularly in Lao PDR, where gaps in breastfeeding and dietary diversity between ethnic groups exceeded 35 percentage points. A&T used the findings to inform country strategies, advocacy with government and donors, and regional knowledge products. Because the pipeline is configuration-driven, it extends to new countries or survey rounds without rebuilding — a reusable asset, not a one-off deliverable.
When national averages show progress, disaggregated analysis is the only way to know whether it reaches the people who need it most.
Need to know who your averages are hiding?
I build reproducible, configuration-driven pipelines that turn scattered survey rounds into equity evidence your team can extend.