Targeted placement of government health clinics is required to provide widespread access to primary health care services, and achieve the Sustainable Development Goals (SDGs) in many developing countries.
New data sources such as mobile phone data and geographic information systems (GIS) can be used as a public good to inform government policy decisions on health clinic locations. This approach is detailed in the new DIAL publication, Using Mobile Phone Data to Make Policy Decisions, that investigates how new data sources can optimize health facility placement
Malawi MNO Data for Health Clinics
In Malawi, an estimated 8 million people – 45% of the population – live more than 5 km from a health facility, reducing their access to essential health care services. Malawi’s population is projected to grow from 18 million in 2018 to 22 million in 2023. Without action, the number of Malawians without access to a health facility would grow to 9.7 million by 2023.
The Malawi Ministry of Health created a Capital Investment Plan (CIP) that proposes to build 900 new health facilities between 2020 and 2023. These clinics should be strategically located to ensure that 95% of the population lives within 5 to 6 km of a health facility by 2023.
The Malawi Ministry of Health (MoH) and Malawi Communications Regulatory Authority (MACRA), worked with DIAL, Cooper/Smith, Infosys, and a Mobile Network Operator (MNO) data to understand population density and migration patterns of people to get a clearer picture of population density and movement patterns throughout the country.
Anonymized call data records (CDRs) and unique call density were used as a complement to national census statistics to better reflect both seasonal migration patterns and long-term urbanization, without the need for expensive on-the-ground surveys.
The optimized analysis was integrated into the CIP as a technical appendix, informing the allocation of new health clinics. The model used a ranked list of priority locations for new health posts to ultimately reduce the proportion of Malawians without access to health care from 45% to just 5% by 2023.
5 MNO Data Analysis Lessons Learned
The process of acquiring the data, analyzing it and integrating it into country systems generated a number of lessons learned and recommendations, notably:
- Define a specific, demand-driven use case to structure the analytical process within a realistic time horizon that allows for unforeseen delays.
- Emphasize country-level buy-in from the very beginning of the process to ensure that the use case is policy-relevant and that the research is in full regulatory compliance with regards to data encryption and user confidentiality.
- Engage with private-sector partners on the potential value-add from both a CSR and business development perspective. Mobile network operators are looking to engage with development partners to expand the use of their data products, but expectations must be managed to allow for differing perspectives.
- Bring together a broad-based analytical team of researchers to tackle the many technical challenges inherent in preparing, cleaning and analyzing multiple datasets and bringing them together to deliver relevant insights.
- Engage continuously with technical counterparts to ensure the relevance of analytical products, laying the groundwork for integrating these products into country systems.
2 Limitations of MNO Data
While MNO call record data can benefit government policy makers, even if its data from only one of two principle telecom providers in a country, there are still two limitations on CDRs for development usage:
Gender
The DAIL model was gender blind by construct, since the data was stripped of identifying characteristics. However, there is a significant global gender gap in mobile ownership. Any model that relies on MNO data will likely over-represents the movements of men, who usually have greater access to mobile phones compared to women, who are more likely to visit health facilities, particularly for prenatal care.
Mobile Coverage
MNO data does not reach to the 5% of zones where there is no mobile coverage in Malawi. Population movement and growth in these zones can be inferred based on observed patterns in adjacent areas with mobile coverage, but since these are the most remote areas, these inferences might not provide a complete picture.
One critical thing I observe this piece is missing in terms of Lessons Learned : challenging regulatory environments.
This is very relevant (especially as we are seeing a trend of regulatory changes in East Africa, with governments making changes to data localisation that actually make it very challenging for private sector actors.)
And, the literature is out there (In fact, you have written on this before =-)
ICTWorks had a post the other week which specifically mentions this: >> Using Mobile Network Operator Data for COVID-19 Response >> ‘Considerations for Getting Started’
>> Is there political support for using MNO data?
>> Are there data sharing agreements in place?
This post also refers us to DIAL’s paper (2018), ‘Unlocking MNO data to enhance public services and humanitarian efforts’ . Quite relevant to what you are publishing here, and it specifically cites the topic:
Among practitioners, vs academics, Lessons Learned are what make things useful to anywhere else.
Many thanks for sharing this activity in Malawi and the actions of Malawi Ministry of Health.
1. https://digitalimpactalliance.org/wp-content/uploads/2018/02/DIAL_D4D-Report_2018.pdf
2. http://www.ictworks.org/mobile-network-operator-data-covid-19-digital-response/
Thanks for sharing! I’m curious to know how many of these colds pots of populations uncovered were unknown previously and only detected with this method? And was model validated by looking at the real situation on the ground?
Thanks,
Karin