At least half of the global population currently lacks access to essential health services, challenging the ability of development stakeholders to achieve Sustainable Development Goal 3 over the next 10 years. Many of the hardest-to-reach populations live in rural areas lacking detailed maps, without which health care workers might not even know that these communities exist.
“It sounds almost obvious, but we can’t get a health worker in a community if we don’t know that community exists,” said Avi Kenny, former director of research, monitoring, and evaluation at nonprofit organization Last Mile Health, adding that census data in low- and middle-income countries is often not that good.
The need to fill in data gaps and create reliable maps calls for new methods to track both access to health services and the spread of diseases. Satellite imagery and data hold promise as key tools for this, with satellites’ ability to rapidly cover remote areas and see things that people can’t always detect on the ground.
According to Einar Bjorgo, director of satellite analysis and applied research at the United Nations Institute for Training and Research, the institute’s Operational Satellite Applications Programme, or UNOSAT, has supported the World Health Organization in rapid response situations on numerous occasions.
“We can’t get a health worker in a community if we don’t know that community exists.”— Avi Kenny, former director of research, monitoring, and evaluation, Last Mile Health
He cited the polio outbreak in Bihar, India, a decade ago. “Our WHO colleagues called us because they needed to quickly organize a vaccine campaign but had no maps,” Bjorgo said. Within a few days, UNOSAT was able to create up-to-date printed maps of the area for the WHO team to take to Bihar. The team then stuck them together with Scotch tape to help the campaign plot out the quickest routes to reach people.
“It worked out well. Since then, there was not one polio case reported there,” Bjorgo said, with India declared polio-free in 2014.
UNOSAT has since provided satellite imagery for WHO in other disease mapping campaigns, such as the Ebola crisis. Maps have been used to identify the locations of treatment centers, roads, and routes to communities, as well as by Médecins Sans Frontières to help plan facilities based on village locations and their layouts.
Bjorgo said there is additional scope for using satellites in longer-term health and disease tracking and eradication efforts — something that UNITAR has been looking into.
“In terms of the Sustainable Development Goals, we are looking now to see if we can come up with various datasets that can be fed to the countries, including in the health sector,” he said.
But Bjorgo said one of the challenges is ensuring that data is sufficiently detailed to be relevant in local contexts. “They shouldn’t be just global models; you have to be able to zoom down and make them useful also at the subnational level,” he said.
Such initiatives are particularly pertinent following the launch of the UNITAR-hosted Global Surgery Foundation in January, recognizing the urgent need to scale up access to safe and affordable surgical care to help achieve universal health coverage and meet the SDGs. Though the program is in its early days, Bjorgo said UNITAR will be looking at how satellite information can be combined with the foundation’s work to improve access.
Translating satellite images into useful data
Other organizations have also been looking into blending satellite data with technologies such as machine learning algorithms to further boost effectiveness.
In Liberia, Last Mile Health recently worked with the U.S.-based Arnhold Institute for Global Health to examine the potential for mapping remote communities using a novel method that combines analysis of publicly available satellite images with machine learning to aid health service planning.
Last Mile Health had already been working in partnership with Liberia’s Ministry of Health since 2016 on scaling up access to community health workers across Liberia, requiring census data on people’s locations. The traditional method, said Kenny — a technical adviser at the organization — involved people riding around on motorcycles armed with GPS devices and Android phones to map roads and count households they passed as a proxy for population size.
While this method worked relatively well, isolated clusters of communities were often missed, Kenny said. So the team trained a machine-learning algorithm to detect buildings in satellite images. It then used the algorithm to find buildings in areas that Last Mile Health had already surveyed on the ground in Liberia’s Rivercess county.
The method correctly detected 75% of registered communities and, crucially, found an additional 167 building groupings, suggesting the presence of communities that had not previously been identified.
Kenny noted that by combining machine learning with traditional ground-based methods, the team was able to significantly boost the richness and quality of data.
While Kenny said he thinks the type of method used in the study shows real promise, it’s uncertain how Last Mile Health might harness such technology in the future. “[The organization does] want to take advantage of emerging machine-learning technology to improve the quality of care for patients, but what that looks like is an open question,” Kenny said.
Others are using satellite data to make strides in tracking the spread of infectious diseases and utilizing maps showing accessibility to health facilities that can feed into this tracking.
A global network of researchers at the Malaria Atlas Project, or MAP, has been doing this since the initiative launched in 2006. MAP, which is based out of Australia’s Telethon Kids Institute and the University of Oxford’s Big Data Institute, provides maps for WHO’s annual “World Malaria Report” as well as mapping support for NGOs, national malaria programs, ministries of health, and other third parties.
Through satellite imagery, the initiative captures global environmental conditions that influence malaria transmission — including land surface temperatures, precipitation, vegetation cover, and population trends — collected from sources such as the Moderate Resolution Imaging Spectroradiometer, the Landsat program, and the Global Urban Footprint initiative. This is combined with a variety of other data from national surveys and surveillance systems to help map malaria.
Among its initiatives, the organization last year succeeded in producing maps of the distribution of malaria worldwide between 2000 and 2017 to chart change over time, compared with previous global maps that had focused on a single year.
According to Harry Gibson, senior business analyst for malaria and geospatial data at MAP, this could help follow trends of where things are improving and where they are not. This could, for example, help give a sense of how well interventions such as supplying bed nets or insecticides may have worked and where efforts need focusing.
MAP’s aim is to gain increasingly high-resolution information over time to better predict malaria burden, Gibson said. In a blog post last year, Bill Gates — whose foundation funds MAP — wrote of the initiative: “We now have data-rich maps with pixels that are just 5 km square. Instead of blanketing entire regions with bednets and other anti-malaria measures, health officials can target efforts where they will do the most good.”
Gibson, meanwhile, said that there is still much room to improve satellite data on population distributions to get a better picture of the proportions of infected people rather than just numbers of cases — though this is also advancing through initiatives he cited, such as Global Urban Footprint and Facebook’s High Resolution Settlement Layer.
“There’s no such thing as a definitive global map of where people live, but a lot of people are doing interesting work with satellite data to produce better maps of this,” Gibson said.
While there is great potential for satellites to provide increasingly high-resolution health insights, those active in this area stress a need to put that data into the right context for each specific location.
For example, Kenny pointed out that it would be difficult to apply the same method used in Last Mile Health’s study to urban areas, where it’s harder to equate buildings to population size because there may be large, abandoned constructions.
He also highlighted a need to make satellite-based technology systems accessible to people without technical backgrounds.
“What would be ideal is if there was some sort of a website where somebody who doesn’t have a technical background can just go and look at a map and zoom in on a country or a district … and view data that’s coming out of this [type of] algorithm,” he said.
Bjorgo emphasized a similar point, adding that UNITAR has a role to play in training and building capacity development on the ground. Furthermore, he said, there is a need for in-country health professionals to receive ready-processed information rather than having to become “image analysis experts.”
“Often, you can’t solve the problem only with the imagery,” he added. “You need to solve a problem by combining the imagery with other information.”
Visit the Data for Development series for more coverage on practical ways that satellite data can be harnessed to support the work of development professionals and aid workers. You can join the conversation using the hashtag #DataForDev.