Creating a verified health facility dataset for spatial analysis: example from Greater Accra, Ghana
- 1University of York
Abstract
Geocoded health facility data are essential for spatial analyses, yet routine health information systems in many low- and middle-income countries often contain inaccurate, incomplete, inconsistent, or outdated facility coordinates. For Greater Accra, Ghana, we created a geocoded health facility dataset aligned with routine maternal health service data from the District Health Information Management System (DHIMS2). We undertook a desk-based, multi-source geocoding and verification workflow of 602 facilities reporting maternal health services between 2020 and 2024. Candidate coordinates were triangulated using national facility records, online mapping platforms, facility websites and social media, directories, photographs, satellite imagery, and local expert knowledge. Coordinates were classified according to the strength and concordance of supporting evidence, with uncertain cases undergoing secondary review. Coordinates were identified for 96.4% of 602 health facilities (n = 580), with 59.3% assigned a high confidence classification (n = 357). Only 22 facilities (3.7%) could not be located with sufficient supporting evidence. Our resulting dataset is made available in a version-controlled openly-modifiable repository for long-term use, permitting routine service utilisation data to be linked with facility locations; thereby facilitating spatial analysis important for equitable health service planning. By documenting our process, this study makes a typically hidden stage of spatial health research transparent and reproducible. Our findings also highlight the need to treat geocoded facility datasets as living, maintainable geospatial health data infrastructure rather than as static research outputs.
Related
- A spatial modelling framework for equitable and efficient elderly care facility allocation in urban China
- A national spatial prioritisation framework for the targeted expansion of essential sickle cell disease services across health-facility catchments in Uganda.
- A multi-omic, spatial, and whole-slide image dataset of lung neuroendocrine tumours from the lungNENomics cohort
- The Role of Community Pharmacies in Creating Harm Reduction Infrastructure to Address the US Opioid Crisis.
- Accountability and equity in primary health care: co-creating user-centered evaluation systems for public-private partnerships
BibTeX
@misc{mroz2026creating,
title = {{Creating a verified health facility dataset for spatial analysis: example from Greater Accra, Ghana}},
author = {E. J. Mroz and E. Afumaa and A. J. Mason-Jones},
howpublished = {medRxiv},
year = {2026},
month = sep,
doi = {10.64898/2026.09.30.26364381},
url = {https://www.medrxiv.org/content/10.64898/2026.09.30.26364381v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Creating a verified health facility dataset for spatial analysis: example from Greater Accra, Ghana
AU - E. J. Mroz
AU - E. Afumaa
AU - A. J. Mason-Jones
PY - 2026
DA - 2026/09/30
PB - medRxiv
DO - 10.64898/2026.09.30.26364381
UR - https://www.medrxiv.org/content/10.64898/2026.09.30.26364381v1
AB - Geocoded health facility data are essential for spatial analyses, yet routine health information systems in many low- and middle-income countries often contain inaccurate, incomplete, inconsistent, or outdated facility coordinates. For Greater Accra, Ghana, we created a geocoded health facility dataset aligned with routine maternal health service data from the District Health Information Management System (DHIMS2). We undertook a desk-based, multi-source geocoding and verification workflow of 602 facilities reporting maternal health services between 2020 and 2024. Candidate coordinates were triangulated using national facility records, online mapping platforms, facility websites and social media, directories, photographs, satellite imagery, and local expert knowledge. Coordinates were classified according to the strength and concordance of supporting evidence, with uncertain cases undergoing secondary review. Coordinates were identified for 96.4% of 602 health facilities (n = 580), with 59.3% assigned a high confidence classification (n = 357). Only 22 facilities (3.7%) could not be located with sufficient supporting evidence. Our resulting dataset is made available in a version-controlled openly-modifiable repository for long-term use, permitting routine service utilisation data to be linked with facility locations; thereby facilitating spatial analysis important for equitable health service planning. By documenting our process, this study makes a typically hidden stage of spatial health research transparent and reproducible. Our findings also highlight the need to treat geocoded facility datasets as living, maintainable geospatial health data infrastructure rather than as static research outputs.
ER -