Inside The DHS Program: Q&A with Gisèle Dunia

Name: Gisèle Dunia

Position title: Senior Advisor for Capacity Strengthening

Languages spoken: French, English, Swahili, Lingala, and Haitian Creole

When did you start at The DHS Program? March 2019

Favorite DHS survey cover: I prefer it when we have an image that represents the country. For example, the report for the 2007 Democratic Republic of the Congo DHS had an okapi on the cover, and you find okapis only in DRC.

What is your role at The DHS Program? As the Senior Advisor for Capacity Strengthening, I oversee the implementation of strategies to strengthen host country individual and institutional capacity, working with different technical teams.

My work involves assessing survey implementing agencies’ capacity at the beginning and at the end of a survey and working on capacity strengthening activities to improve and sustain institutional capacity. Capacity strengthening activities are either survey-related or competency-based trainings. Our training opportunities are offered both online, on The DHS Program Learning Hub, and in-person during national and regional workshops.

Another way of strengthening capacity at the country level is by collaborating with consultants. The DHS Program has been using south-to-south consultants for several years. Consultants help build and reinforce capacity in host countries and across regions. Under DHS-8, we are designing a certification program for these consultants, streamlining processes to equip them with skills to better support survey implementation and dissemination.

What work are you most proud of? I have designed and facilitated several capacity strengthening activities in the past. In my work now at The DHS Program, I’m no longer in front of people facilitating trainings. I am mostly behind the scenes. I am very much involved in the design process, making sure that we have the right tools to facilitate engaging trainings. I am proud of the way I’ve been able to help technical teams design trainings, and I trust them to successfully run the show.

I am most proud of completely designing the DHS-8 Global Capacity Strengthening Strategy within my first six months at a program that has so many components as The DHS Program.

What’s your favorite trip to date? So far, my second trip is my favorite one. In December, I went to Madagascar to conduct a capacity assessment of the Institut National de la Statistique (INSTAT), the implementing agency for the forthcoming fifth Madagascar Demographic and Health Survey. While there I pilot-tested our updated Capacity Assessment Tools, which I used to assess INSTAT’s current capacity. Based on the results, I shared with INSTAT a list of capacity strengthening activities that I think would benefit them, like how they can restructure the way they work so that whatever capacity is built during the DHS survey process can be managed and shared throughout INSTAT to build long-term institutional capacity.

For more information about The DHS Program’s capacity strengthening approaches, visit our website.

Luminare: Geospatial Modeling for Locally Available Data

This blog post is part of Luminare, our blog series exploring innovative solutions to data collection, quality assurance, biomarker measurement, data use, and further analysis.

Demographic and Health Surveys (DHS) collect nationally representative data and data representative at the first subnational administrative level (ADMIN 1). The 2016 Ethiopia DHS was designed to produce representative estimates for nine regions and two administrative cities. The 2014 Kenya DHS produced estimates for eight regions (formerly provinces). In addition to national-level indicators, STATcompiler also presents subnational data, as shown in the map of childhood stunting at the ADMIN 1 level in Ethiopia and Kenya.

Childhood Stunting by Subnational Level, 2016 Ethiopia DHS & 2014 Kenya DHS

Click the map to explore childhood stunting using STATcompiler.

National and ADMIN 1 data help countries track their progress towards achieving the Sustainable Development Goals, for instance. However, as countries decentralize their health service delivery systems, local health officials increasingly need local data. In Ethiopia, nine regions are further divided into zones and special districts (ADMIN 2). In Kenya, eight regions are further divided into counties.

One option to get data representative at the ADMIN 2 level is to increase the survey sample size, requiring more time and more money. Another option is to produce spatially interpolated maps, which use Bayesian geospatial modeling techniques to predict indicator values at non-surveyed locations.

The DHS Program’s Geospatial team assembled data for 12 geospatial covariates, such as elevation, precipitation, and population density. These covariates are related to and can partially explain variation in health indicators of interest, allowing for more accurate predictions across the map.

Next, the Geospatial team imported georeferenced cluster data points from the 2016 Ethiopia DHS and 2014 Kenya DHS. (Did you know? You can download shapefiles or geodatabases of georeferenced data for most DHS surveys from the Spatial Data Repository.)

Using the geospatial covariates and survey data, the Geospatial team employed a new modeling approach–a stacked ensemble model–which combines multiple models. This increases predictive power and captures the potentially complex interactions and non-linear effects among the geospatial covariates. Three sub-models were fit to the health indicator data using the geospatial covariates as exploratory predictors. The prediction surfaces generated from the sub-models were then used in the final Bayesian geospatial model, producing 5 X 5 km pixel-level mean estimates of health indicators with associated uncertainty.

Childhood Stunting by 5 X 5 km Pixel, 2016 Ethiopia DHS & 2014 Kenya DHS

Modeled surface maps available from the Spatial Data Repository.

Pixel-level estimates were then used to calculate population-weighted averages to aggregate estimates to the ADMIN 2 level. For Ethiopia, this produced estimates of childhood stunting by zone, and in Kenya, estimates by county.

Childhood Stunting by ADMIN 2 level, 2016 Ethiopia DHS & 2014 Kenya DHS

Health system program managers in Ethiopia and Kenya can now use these zonal- and county-specific estimates to make decisions and manage locally administered health programs to address childhood stunting in their areas.

The DHS Program will continue exploring model-based geostatistics as a feasible, reliable, and cost-effective way to produce local data for local needs.

Read the full report, Interpolation of DHS Survey Data at Subnational Administrative Level 2.

Explore available spatially modeled map surfaces of DHS indicators on the Spatial Data Repository.