Introduction
To conduct a study analyzing under-resourced populations in a dense, dynamic city like Lagos, understanding where these populations reside, and their proximity to telecommunications and health system access alike, is vital. Understanding location and travel accessibility affects how a person can travel to access primary healthcare – as areas that are further away from a facility strongly benefits from a telehealth initiative.
Geographically identifying under-reached populations and applicable population segments to target interventions, understanding where current health system delivery gaps in facility and service availability and poor geographic/physical access to health facilities and low performance. The Lagos Innovation portfolio, a geospatial analysis performed by members of the Geospatial Insights Support Team (GIST) from DevGlobal and Dev-Afrique Development Advisors, aimed to analyze and identify underserved communities who could potentially receive telehealth interventions for an interventional research study conducted by Solina Centre for International Development and Research (SCIDaR) and VillageReach.

Methods
Geospatial datasets were procured from GRID3, OpenStreetMap, and OpenCellID, while tabular datasets were acquired from the Nigeria Federal Ministry of Health, the Lagos Bureau of Statistics, and the Lagos Primary Healthcare Board.
GRID3 population density data from 2021 (quantified as high, medium, and low), which demonstrates the distribution and intensity of human activity across all 20 LGAs, was overlayed with the coordinates of existing health facilities obtained from the Lagos Primary Healthcare Board. The existing health facilities were mapped with their primary, secondary, and tertiary-level healthcare facility levels. Additional information, such as their operational status, as defined by the Nigeria Federal Ministry of Health’s Master Facility Lists and Health Facility Registry, public/private status, and clinical staffing levels, were also incorporated for each facility.
A 3km buffer was created around each active healthcare facility to show the estimated coverage of healthcare services, a more conservative estimate compared to the World Health Organization’s 5 kilometers optimal distance for healthcare accessibility. As a result, areas outside the 3km buffer were staked as potential intervention locations for the telemedicine study. These areas of potential intervention were further overlayed with Google Earth imagery and OpenStreetMap road network data to precisely map areas of considerable development and human activity.
For each LGA in Lagos state, a set of socioeconomic indicators – such as gender (male/female), income level (low or high), education level (elementary or advanced levels), housing type (based on building size, rental/ownership costs and luxurious building features), and employment type (high/low) - were obtained from the Lagos Bureau of Statistics, categorized and appended to each LGA. Potential service points were then overlaid on top of the LGAs with appended socioeconomic indicators.
Additional analyses were performed to analyze the effects of travel time and cell signal catchment areas. Travel time catchments were determined in Python using OpenStreetMap road networks (including routing information), and the osmnx, networkx (to Python package for the creation, manipulation, and study of the structure, dynamics, and functions of complex networks.), and sklearn Python modules to (1) download, model, analyze and visualize street networks and incorporate speed, travel time, and routing, (2) create, manipulate, and model the complex road network, and (3) analyze the projected graph network. The travel time catchment area, determined as the number of intersections that could be traversed within a 20-minute time, was determined for both driving (using a 30km/hr. speed) or walking (using a 5km/hr. speed) scenarios. These catchment areas were overlaid on the circular buffer areas. Areas selected were typically those with high measures of poverty (e.g., low income, poor housing type and/or employment type) that were outside of the healthcare-accessible buffer areas.
Another auxiliary analysis, to map cell signal catchment areas, was performed to help guide potential telemedicine initiatives. Tower location coordinates were obtained from OpenCellID, which contained information about the tower location, the type of transmission (2G, 3G, 4G, and 5G), and other indicators. Each tower location had a buffer mapped around it, equivalent to the radius (“range”) of transmission. The individual tower buffers were merged together for each type of transmission, using dissolve in QGIS, to create a zone for each transmission type.

Results
A total of 26 potential service points were identified as underserved areas across different local government areas (LGAs) and local council development areas (LCDAs) in Lagos. These locations were found in Ikorodu (6), Ibeju Lekki (5), Badagry (2), Ojo (2), Alimosho (2), straddling Ojo and Alimosho (1), Ifako Ijaye (1), Amuwo Odofin (2), straddling Oshodo-Isolo/Ikeja (1), Lagos Island (1), Eti Osa (1), and Epe (1) LGAs.
When analyzing the distribution of primary healthcare facilities, including community health centers, rural health clinics, paediatrics practices, family medicine practices, etc., about 70 percent of the total healthcare facilities in Lagos state were made of such facilities, with more primary healthcare facilities occurring in higher-population density areas LGAs (e.g., Alimosho, Badagry, Eti-Osa, Kosofe, Ikorodu, and Ojo). Additionally, it was found that private HCFs were more prevalent than public HCFs in the higher-density locations. Our results also suggested a low population of clinical staff occurred across the state.
When using calculated travel time from each underserved point, it was found that the travel time-specific catchment areas often went further outside of the circular buffer zone used to signal healthcare accessibility. When incorporating the travel time buffers into the analysis, the number of communities captured as part of each catchment area were often much higher than those captured solely by the circular healthcare accessibility buffer.
The majority of cell signal network types encountered in underserved communities are 2G, according to an analysis of cell signal range and LGA data, with 3G and 4G services available in more population-dense areas of the state.

Discussion
By utilizing a hybrid approach of mapping and spatial overlay of local health facilities, telecommunications access, road networks, travel time catchments and various socioeconomic indicators, our team was able to identify a shortlist of 26 underserved settlements across 11 LGAs in Lagos state that could be potentially be best serviced by a telehealth intervention.
After these initial results were disseminated by the GIST in May 2023, these results were passed to SCIDaR and VillageReach for a telemedicine research intervention study, “Implementation Research to Test Equitable Digital-First Care in Lagos”. Taken together, these results could be taken as a first step improving equitable access to primary healthcare for underserved populations in Lagos State.