地理空间分析在健康不平等研究中的进展与挑战
Progress and remaining priorities in geospatial analysis of health inequalities
地理空间分析技术的进步使得健康结果和服务能够以高分辨率进行空间和时间映射,揭示了国家层面数据掩盖的地方差异。研究发现,虽然高分辨率地图能够显示疾病负担、干预覆盖率和其他健康结果,但分辨率的提升也带来了准确性的挑战,且细粒度的位置信息可能引发隐私问题。
Health outcomes, services and their drivers can now be mapped at fine spatial and temporal resolution, driven by growth in data, computing power and geospatial methods. Here we discuss what the field has achieved, where estimates can outpace their use, and what actions could close the gap.
Health outcomes vary from place to place, yet national averages can mask this. A country on track to meet coverage targets frequently contains areas where large populations remain unreached. Making these variations visible and usable is the work of spatial epidemiology, a field at the intersection of health and geography1,2. Over recent decades, the field has been reshaped by wide access to georeferenced household surveys, geocoded health facility lists and satellite imagery, alongside advances in spatial and spatiotemporal statistics3, computing power and, more recently, geospatial artificial intelligence (GeoAI). We mainly focus on low- and middle-income countries (LMICs), where the burden of disease is highest and inequalities within countries are widest.
The field’s clearest achievement has been high-resolution mapping of disease burden, intervention coverage and other health outcomes, reflecting a shift from describing where disease occurs to generating predictions where no data were collected. These continuous estimates can expose local stagnation that national progress can conceal. For instance, continental modelling showed where malaria transmission fell as control scaled up and where it did not4. Small-area estimation of under-five mortality across 35 African countries revealed which subregions lagged furthest behind national progress, and which areas were catching up5. In Kenya, combining environmental suitability modelling with geostatistical mapping found only 4% of the landmass suitable for podoconiosis, with more than 60% of cases in just three of 47 counties, a pattern that favours locally targeted interventions over nationally uniform programmes6. Beyond mapping levels, space-time models now track how patterns shift and where new hotspots emerge, and link disease to climate, population and environment; they are increasingly used to inform elimination efforts and service planning. Much of this work depends on open, high-resolution gridded population data, which provide the denominators and the population counts against which nearly every estimate is measured, though their accuracy varies7.
Geographic accessibility mapping has advanced in parallel. Global travel-time surfaces now estimate journeys to cities, healthcare, schools and other services, exposing inequalities in who can reach them8. More detailed and localised analyses account for traffic congestion, time of day and travel mode, drawing on routing services, vehicle GPS, mobile phone records and crowdsourced road networks that mirror how people actually reach services9. Several overlapping vulnerabilities, such as poverty, urban deprivation and travel time to care, now can be examined together and tracked over time through high-resolution gridded estimates of wealth, population and urbanisation, linking access to who is being left behind. In sub-Saharan African cities, residents of informal settlements faced longer walking times to facilities than those in formal areas10.
Despite this progress, several weaknesses remain. First, resolution can be mistaken for accuracy. Methods are often used because they are feasible, not because they suit the question at hand; as such, analytic tractability can quickly outrun appropriate data usage, and models at times are pushed further than the data can support. In the Demographic and Health Surveys (DHS), the data source behind most mapping in LMICs, the recorded locations of survey clusters are deliberately shifted to protect confidentiality. Accordingly, the data cannot reliably support estimates below a certain geographic scale, however geospatially fine the model. Changing the unit of analysis can do more than add error; a prevalence estimated at a point, on a grid and across a district are three different quantities, not one quantity measured three ways. Second, the finer the location, the easier it can become to re-identify the people behind an estimate. Third, covariates do not always help; in fact adding them can even make the estimate less certain when data are sparse and clustered, especially, in low-prevalence settings11. Fourth, uncertainty is usually calculated but then set aside: decision-makers see the headline estimates but not how uncertain they are, which is vital for judging how far the estimates can be trusted.
There is also a deeper gap. Geospatial estimates are not frequently checked against independent data; a model tested against its own source survey only shows it agrees with itself. Moreover, even robust estimates do not always change decisions. Malaria programme managers in three African countries reported that their main obstacles to using geospatial estimates for decision-making involved interpreting and trusting the maps, not producing more of them12.The field now can make maps faster than people have learned to use and act on them.
GeoAI is increasingly pitched as a fix. It can find structure in data that older methods miss, and it helps most where conventional covariates are weakest, such as rapidly changing places. But to date GeoAI primarily changes how the maps are made, not if and how people use them. And GeoAI poses a new risk: when data and methods behind their production become even less visible to end-users, trust can further diminish in parallel13.
None of these challenges is new and similar solutions have been offered before: put modellers inside ministries, connect local universities to disease control programmes, fix population denominators on which estimates rely, among others1. Nonetheless the field is still grappling with a more philosophical tension: when does finer resolution estimates stop helping decisions and actually start hindering them. For instance, continental-scale mapping initiatives continue to expand4 while the underlying data have become precarious14; all the while, how exactly geospatial analyses themselves have changed health outcomes remains largely unknown2. On its own, adding computing power and model complexity into increasingly weak data foundation is unlikely to produce desired outcomes for the geospatial field. Elevating four key priorities below is therefore vital to catalysing future progress (Table 1).
The first priority is data, which sets the limit of what can be mapped. Most analyses in LMICs rely on household surveys, routine health records and geocoded lists of facilities. The DHS in particular provides the georeferenced survey locations most mapping depends on. Its brief suspension after the withdrawal of USAID funding was a warning of how fragile that foundation could be14; access has been restored, but a source now negotiated country by country can no longer be assumed as a stable, standard input. Routine health records are collected continuously and geolocated by administrative area, nonetheless, they often experience challenges with incompleteness, unreliable population denominators and unclear catchment areas. The gaps disproportionately affect the people already hardest to reach; data on conflict, displacement, remoteness and informal settlements, which mark out the most vulnerable, are among the least used, even as climate change and conflict reshape where people live and where disease risk occurs15. These are the people health systems most need to reach, yet they are often the least visible in the same data used to guide that reach. Methods alone cannot fill gaps for measurements never taken.
The second is open science, which would strengthen the validation and reuse of current estimates and methods. Analytic code and modelling inputs like covariate layers and facility datasets are not always shared, and written descriptions alone do not lend themselves to reproduction. Sharing modelling pipelines would raise rigour and save effort, letting people build on existing work instead of recreating the same foundations. Reporting standards already exist, such as Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER) and Spatial Lifecourse Epidemiology Reporting Standards (ISLE-ReSt); stronger incentives for their consistent adoption could help galvanise meaningful change in open science.
The third is capacity: shifting away from a primary focus on methods and moving toward an operating model where local teams and system routinely produce, maintain and use geospatial estimates. Advanced modelling is largely concentrated in institutions in the Global North, even though continental mapping initiatives are mostly focused on LMICs. Such gaps are shaped by unequal access to computational infrastructure and sustained funding for capacity-building. Current incentives reward publication over the slower, longer-term work of embedding these tools into health departments; furthermore staff cannot be expected to learn without designated time and resources1. Work that changes practice is built with the people who will ultimately use it; as such strengthening local ownership and interpretation is a critical need going forward12.
The fourth priority is evaluation. Studies typically end in recommendations, but whether these recommendations have traction or impact beyond analysis often remains unknown. Regular evaluation does not need to be resource-intensive, especially if approaches such as contribution analysis and rapid evaluation are accepted as credible alternatives to formal impact evaluation. Meaningfully supporting and acting on evaluations can require social capital and political will exceed the financial requirements. Two shifts could help. First, existing country review processes could be leveraged to expand lighter-touch evaluations of geospatial analyses, for example, annual microplanning exercises for routine immunisation programmes. Second, reframing evaluation as a vital input into health system strengthening, resilience and sustainability could lower the sociopolitical stakes for governments, programme managers and funders alike.
Geospatial analysis can now locate health inequalities with precision. What it has yet to do is turn that precision into decisions and programmes that reduces these inequalities. Investment in data, open science, capacity and evaluation is what would move the field beyond producing evidence. Success should be judged not by the resolution of the map, but by whether it changes decisions and, ultimately, outcomes.
Data availability
No new data were generated or analysed in this Comment. All sources discussed are cited in the reference list.
References
Tatem A. J. Innovation to impact in spatial epidemiology. BMC Med. 16 https://doi.org/10.1186/s12916-018-1205-5 (2018).
Morrison, C. N. et al. Defining spatial epidemiology: a systematic review and re-orientation. Epidemiology 35, 542–555 https://doi.org/10.1097/EDE.0000000000001738 (2024).
Diggle P. J., Giorgi E. Model-Based Geostatistics for Global Public Health (Chapman and Hall/CRC, 2019).
Bhatt, S. et al. The effect of malaria control on Plasmodium falciparum in Africa between 2000 and 2015. Nature 526 https://doi.org/10.1038/nature15535 (2015).
Li, Z. et al. Changes in the spatial distribution of the under-five mortality rate: Small-area analysis of 122 DHS surveys in 262 subregions of 35 countries in Africa. PLoS ONE 14, e0210645 (2019).
Deribe, K. et al. Geostatistical modelling of the distribution, risk and burden of podoconiosis in Kenya. Trans. R. Soc. Trop. Med. Hyg. 117 https://doi.org/10.1093/trstmh/trac092 (2023).
Leyk, S. et al. The spatial allocation of population: a review of large-scale gridded population data products and their fitness for use. Earth Syst. Sci. Data 11, 1385–1409 https://doi.org/10.5194/essd-11-1385-2019 (2019).
Weiss, D. J. et al. A global map of travel time to cities to assess inequalities in accessibility in 2015. Nature 553, 333–336 https://doi.org/10.1038/nature25181 (2018).
Banke-Thomas, A. et al. Geographical accessibility to functional emergency obstetric care facilities in urban Nigeria using closer-to-reality travel time estimates: a population-based spatial analysis. Lancet Glob. Health 12, e848–e858 https://doi.org/10.1016/S2214-109X(24)00045-7 (2024).
Friesen, J., Georganos, S. & Haas, J. Differences in walking access to healthcare facilities between formal and informal areas in 19 sub-Saharan African cities. Commun. Med. 5, 41 https://doi.org/10.1038/s43856-025-00746-5 (2025).
Giorgi, E. et al. Model building and assessment of the impact of covariates for disease prevalence mapping in low-resource settings: to explain and to predict. J. R. Soc. Interface 18, 20210104 https://doi.org/10.1098/rsif.2021.0104 (2021).
Ghilardi, L. et al. How useful are malaria risk maps at the country level? Perceptions of decision-makers in Kenya, Malawi and the Democratic Republic of Congo. Malar. J. 19 https://doi.org/10.1186/s12936-020-03425-z (2020).
Xia, A. et al. Health GeoAI beyond algorithms: embedding equity, accountability, and environmental responsibility. Geogr. Sustain. 7, 100445 https://doi.org/10.1016/j.geosus.2026.100445 (2026).
Khaki, J. J. et al. When health data go dark: the importance of the DHS Program and imagining its future. BMC Med. 23, 241 https://doi.org/10.1186/s12916-025-04062-6 (2025).
Njogu, A. M. et al. Geospatial approaches for mapping zero-dose prevalence and routine childhood immunisation coverage in low- and lower-middle-income countries: a scoping review. Vaccine 89, 128904 https://doi.org/10.1016/j.vaccine.2026.128904 (2026).
Acknowledgements
P.M.M. is supported by Fonds voor Wetenschappelijk Onderzoek (FWO) through a senior postdoctoral fellowship (Grant ID: 1201925N).
Author information
Authors and Affiliations
Department of Public Health, Institute of Tropical Medicine, Antwerp, Belgium
Peter M. Macharia
Population & Health Impact Surveillance Group, Kenya Medical Research Institute-Wellcome Trust Research Programme, Nairobi, Kenya
Peter M. Macharia
Department of Global Health, University of Washington, Seattle, WA, USA
Nancy Fullman
Department of Statistics, Federal University of Technology, Akure, Nigeria
Ezra Gayawan
Authors
- Peter M. Macharia
- Nancy Fullman
- Ezra Gayawan
Contributions
P.M.M. drafted the initial manuscript. N.F. and E.G. both contributed to its revision.
Corresponding author
Correspondence to Peter M. Macharia.
Ethics declarations
Competing interests
P.M.M., N.F. and E.G. served as Guest Editors of the Collection ‘Geospatial analysis for improved understanding of health inequalities’. P.M.M. is an Editorial Board Member for Communications Medicine, but was not involved in the editorial review of, nor the decision to publish this article. The authors declare no other competing interests.
Peer review
Peer review information
Communications Medicine thanks all reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Macharia, P.M., Fullman, N. & Gayawan, E. Progress and remaining priorities in geospatial analysis of health inequalities. Commun Med 6, 575 (2026). https://doi.org/10.1038/s43856-026-01993-w
Received: 29 September 2026
Accepted: 12 October 2026
Published: 03 November 2026
Version of record: 03 November 2026
DOI: https://doi.org/10.1038/s43856-026-01993-w
来源:Nature 健康护理主题 · nature.com