BackgroundAdolescents living in agricultural areas are at higher risk of secondary pesticide exposure; however, there is limited evidence to confirm exposure by pesticide drift for greenhouse floriculture, like rose production.Methods525 adolescents (12–17, 49% male) living in Pedro Moncayo, Ecuador were assessed in 2016. Urinary concentrations of creatinine and pesticide biomarkers (organophosphates, neonicotinoids, and pyrethroids) were measured using mass-spectrometry. Home distance to the nearest greenhouse and surface area of greenhouses within various buffer sizes around the home were calculated. Linear regression assessed whether home distance and surface area of greenhouses was associated with creatinine-adjusted metabolite concentration, adjusting for demographic, socioeconomic, and anthropometric variables. Geospatially weighted regression (GWR) was conducted, adjusting for similar covariates. Getis-ord Gi* identified hot and cold spots using a 1994 m distance band.ResultsThe associations between residential distance to greenhouses and urinary pesticide metabolites differed by metabolite type. The adjusted mean concentrations of OHIM (neonicotinoid) were greater (p-difference = 0.02) among participants living within 200 m (1.08 ug/g of creatinine) vs > 200 m (0.64 ug/g); however, the opposite was observed for 3,5,6-Trichloro-2-pyridinol (TCPy, organophosphate; 0-200 m: 3.63 ug/g vs > 200 m: 4.30 ug/g, p-diff = 0.05). In linear models, greater distances were negatively associated with para-nitrophenol (PNP, organophosphate; percent difference per 50% greater distance [95% CI]: − 2.5% [− 4.9%, − 0.1%]) and somewhat with 2-isopropyl-4-methyl-6-hydroxypyrimidine (IMPy, organophosphate; − 4.0% [− 8.3%, 0.4%]), among participants living within 200 m of greenhouses. Concurring with the adjusted means analyses, opposite (positive) associations were observed for TCPy (2.1% [95%CI 0.3%, 3.9%]). Organophosphate and pyrethroid hotspots were found in parishes with greater greenhouse density, whereas neonicotinoid hot spots were in parishes with the lowest greenhouse density.ConclusionWe observed negative associations between residential distance to greenhouses with OHIM, PNP and to some extent IMPy, suggesting that imidacloprid, parathion and diazinon are drifting from floricultural greenhouses and reaching children living within 200 m. Positive TCPy associations suggest greenhouses weren't the chlorpyrifos source during this study period, which implies that non-floricultural open-air agriculture (e.g. corn, potatoes, strawberries, grains) may be a source. Further research incorporating diverse geospatial constructs of pesticide sources, pesticide use reports (if available), participant location tracking, and repeated metabolite measurements is recommended.

BackgroundEmissions from wildfire plumes are composed of modified biomass combustion by-products, including carcinogens. However, studies of the association between wildland fires (WF; includes wildfires, prescribed burns, and resource management fires) exposure and lung cancer are scant. We evaluated geographic patterns in these exposures and their association with lung cancer mortality (LCM) rates across the conterminous United States (US).MethodsWe extracted data from the Monitoring Trends in Burn Severity program (1997–2003) and derived county-level exposure metrics: WF density by area, WF density by population, the ratio between total burned land area and county area, and the ratio between total burned land area by population. We obtained sex-specific, county-level LCM rates for 2016–2020 from the National Center for Health Statistics. Counties with fewer than 10 cases were suppressed. To account for cigarette smoking, we first modeled residual values from a Poisson regression between cigarette smoking prevalence and sex-specific, age-adjusted LCM rates. We then used Lee’s L statistic for bivariate spatial association to identify counties with statistically significant (p < 0.05) associations between WF exposures and these residuals. In a sensitivity analysis, we applied a false discovery rate correction to adjust for multiple comparisons.ResultsWe observed geographic variation in bivariate associations between large WFs and subsequent LCM rates across US counties while accounting for ever cigarette smoking prevalence. There were positive (high WF exposures and high LCM rate) clusters for males and females in counties within the mid-Appalachian region and Florida, and modest differences across WF metrics in the cluster patterns were observed across the Western US and Central regions. The most positive clusters were seen between WF density by area and LCM rates among women (n = 82 counties) and a similar geographic pattern among men (n = 75 counties). Similar patterns were observed for males and females in the western US, with clusters of high WF exposures and low LCM rates. After adjusting for multiple comparisons, a positive cluster pattern among both sexes persisted in Kentucky and Florida with area-based exposure metrics.DiscussionOur analysis identified counties outside the western US with wildfires associated with lung cancer mortality. Studies with individual-level exposure-response assessments are needed to evaluate this relationship further.

Generative AI is rapidly establishing itself as a key member of the GeoAI battery of methods, models and tools in use today in various health applications. This paper is the first in an Int J Health Geogr two-article series (2025) on the ‘Generative Revolution’. It is meant to serve as a brief introduction to the second article entitled ‘The Generative Revolution: AI Foundation Models in Geospatial Health—Applications, Challenges and Future Research’.

In an era of rapid technological advancements, generative artificial intelligence and foundation models are reshaping industries and offering new advanced solutions in a wide range of scientific areas, particularly in public and environmental health. However, foundation models have previously mostly focused on understanding and generating text, while geospatial features, interrelations, flows and correlations have been neglected. Thus, this paper outlines the importance of research into Geospatial Foundation Models, which have the potential to revolutionise digital health surveillance and public health. We examine the latest advances, opportunities, challenges, and ethical considerations of geospatial foundation models for research and applications in digital health. We focus on the specific challenges of integrating geospatial context with foundation models and lay out the future potential for multimodal geospatial foundation models for a variety of research avenues in digital health surveillance and health assessment.

BackgroundPast epidemiological studies, using fixed-site outdoor air pollution measurements as a proxy for participants’ exposure, might have suffered from exposure misclassification.MethodsIn the MobiliSense study, personal exposures to ozone (O3), nitrogen dioxide (NO2), and particles with aerodynamic diameters below 2.5 μm (PM2.5) were monitored with a personal air quality monitor. All the spatial location points collected with a personal GPS receiver and mobility survey were used to retrieve background hourly concentrations of air pollutants from the nearest Airparif monitoring station. We modeled 851,343 min-level observations from 246 participants.ResultsVisited places including the residence contributed the majority of the minute-level observations, 93.0%, followed by active transport (3.4%), and the rest were from on-road and rail transport, 2.4% and 1.1%, respectively. Comparison of personal exposures and station-measured concentrations for each individual indicated low Spearman correlations for NO2 (median across participants: 0.23), O3 (median: 0.21), and PM2.5 (median: 0.27), with varying levels of correlation by microenvironments (ranging from 0.06 to 0.35 according to the microenvironment). Results from mixed-effect models indicated that personal exposure was very weakly explained by station-measured concentrations (R2 < 0.07) for all air pollutants. The R2 for only a few models was higher than 0.15, namely for O3 in the active transport microenvironment (R2: 0.25) and for PM2.5 in active transport (R2: 0.16) and in the separated rail transport microenvironment (R2: 0.20). Model fit slightly increased with decreasing distance between participants’ location and the nearest monitoring station.ConclusionsOur results demonstrated a relatively low correlation between personal exposure and station-measured air pollutants, confirming that station-measured concentrations as proxies of personal exposures can lead to exposure misclassification. However, distance and the type of microenvironment are shown to affect the extent of misclassification.

In this paper, we define the optimization of health station locations as a clustering problem. We design a robust algorithm for the problem using a pre-calculated overhead graph for fast distance calculations and apply a robust clustering algorithm called random swap to provide accurate optimization results. We study the effect of three cost functions (Euclidean distance, squared Euclidean distance, travel cost) using real patient locations in North Karelia, Finland. We compare the optimization results with the existing health station locations. We found that the algorithm optimized the locations beyond administrative borders and strongly utilized the transport network. The results can provide additional insight for the decision-makers.

BackgroundTick-borne encephalitis (TBE) is the most serious tick-borne viral disease in Europe. Identifying TBE risk areas can be difficult due to hyper focal circulation of the TBE virus (TBEV) between mammals and ticks. To better define TBE hazard risks and elucidate regional-specific environmental factors that drive TBEV circulation, we developed two machine-learning (ML) algorithms to predict the habitat suitability (maximum entropy), and occurrence of TBEV (extreme gradient boosting) within distinct European regions (Central Europe, Nordics, and Baltics) using local variables of climate, habitat, topography, and animal hosts and reservoirs.MethodsGeocoordinates that reported the detection of TBEV in ticks or rodents and anti-TBEV antibodies in rodent reservoirs in 2000 or later were extracted from published and grey literature. Region-specific ML models were defined via K-means clustering and trained according to the distribution of extracted geocoordinates relative to explanatory variables in each region. Final models excluded colinear variables and were evaluated for performance.Results521 coordinates (455 ticks; 66 rodent reservoirs) of TBEV occurrence (2000–2022) from 100 records were extracted for model development. The models had high performance across regions (AUC: 0.72–0.92). The strongest predictors of habitat suitability and TBEV occurrence in each region were associated with different variable categories: climate variables were the strongest predictors of habitat suitability in Central Europe; rodent reservoirs and elevation were strongest in the Nordics; and animal hosts and land cover contributed most to the Baltics. The models predicted several areas with few or zero reported TBE incidence as highly suitable (≥ 60%) TBEV habitats or increased probability (≥ 25%) of TBEV occurrence including western Norway coastlines, northern Denmark, northeastern Croatia, eastern France, and northern Italy, suggesting potential capacity for locally-acquired autochthonous TBEV infections or possible underreporting of TBE cases based on reported human surveillance data.ConclusionsThis study shows how varying environmental factors drive the occurrence of TBEV within different European regions and identifies potential new risk areas for TBE. Importantly, we demonstrate the utility of ML models to generate reliable insights into TBE hazard risks when trained with sufficient explanatory variables and to provide high resolution and harmonized risk maps for public use.

BackgroundWith China becoming an aging society, the number of elderly care institutions (ECIs) is continuously increasing in response to the growing population of older persons. However, regional disparities may lead to an uneven distribution of ECIs, which could affect equity in care. This study identified the limiting factors in the development of ECIs across different regions, thereby promoting equity in accessing care for the older population.MethodsThis study utilised point-of-interest data on ECIs in China from 2018 to 2022. The spatiotemporal distribution of ECIs and the causes of disparities were assessed along four dimensions—economy, population, society, and environment—using research methods such as the standard deviation ellipse, rank-size rule, and multiscale geographically weighted regression.ResultsThere were significant differences between the ECIs of the eastern and western regions in China. The eastern region had a denser distribution and higher concentrations in primary cities. The proportion of the older population, regional economic development, and household income are crucial for a balanced distribution of ECIs, whereas the environmental impact is relatively minor.ConclusionsThe number of ECIs in China continues to increase, but improvements in regional disparities remain insignificant. The construction of ECIs is influenced by various factors; in underdeveloped regions, government initiatives are crucial for promoting equity in care for older persons.

BackgroundAddressing key behavioral risk factors for chronic diseases, such as diet, requires innovative methods to objectively measure dietary patterns and their upstream determinants, notably the food environment. Although GIS techniques have pushed the boundaries by mapping food outlet availability, they often simplify food access dynamics to the vicinity of home addresses, possibly misclassifying neighborhood effects. Leveraging Google Location History Timeline (GLH) data offers a novel approach to assess long-term patterns of food outlet utilization at an individual level, providing insights into the relationship between food environment interactions, diet quality, and health outcomes.MethodsWe leveraged GLH data previously collected from a sub-set of participants in the Washington State Twin Registry (WSTR). GLH included more than 287 million location records from 357 participants. We developed methods to identify visits to food outlets using outlet-specific buffer zones applied to the InfoUSA data on food outlet locations. This methodology involved the application of minimum and maximum stay durations, along with revisit intervals. We calculated metrics from the GLH data to detect frequency of visits to different food outlet classifications (e.g. grocery stores, fast food, convenience stores) important to health. Several sensitivity analyses were conducted to examine the robustness of our food outlet metrics and to examine visits occurring within 1 and 2.5 km of residential locations.ResultsWe identified 156,405 specific food outlet visits for the 357 study participants. 60% were full-service restaurants, 15% limited-service restaurants, and 16% supermarkets. Mean visits per person per month to any food outlet was 12.795. Only 8, 10 and 11% of full-service restaurants, limited-service restaurants, and supermarkets, respectively, occurred within 1 km of residential locations.ConclusionsGLH data presents a novel method to assess individual-level food utilization behaviors.

BackgroundMalaria is a major public health issue in Nekemte City, western Ethiopia, with various environmental and social factors influencing transmission patterns. Effective control and prevention strategies require precise identification of high-risk areas. This study aims to map malaria risk zones in Nekemte City using geospatial technologies, including remote sensing and Geographic Information Systems (GIS), to support targeted interventions and resource allocation.MethodsThe study integrated environmental and social factors to assess malaria risk in the city. Environmental factors, including climatic and geographic characteristics, such as elevation, rainfall patterns, temperature, slope, and proximity to river, were selected based on experts' opinions and literature review. These factors were weighted using the analytic hierarchy process according to their relative influence on malaria hazard susceptibility. Social factors considered within the GIS framework focused on human settlements and access to resources. These included population density, proximity to health facilities, and proximity to roads. The malaria risk analysis incorporated hazard and vulnerability layers, along with Land use/cover (LULC) data. A weighted overlay analysis method combined these layers and generate the final malaria risk map.ResultsThe malaria risk map identified that 18.2% (10.5 km2) of the study area was at very high risk, 18.8% (10.9 km2) at high risk, 30.4% (17.8 km2) at moderate risk, 19.8% (11.5 km2) at low risk, and 12.6% (7.3 km2) at very low risk. A combined 37% (21.4 km2) of Nekemte City was classified as at high to very high malaria risk, highlighting key areas for intervention.ConclusionsThis malaria risk map offers a valuable tool for malaria control and elimination efforts in Nekemte City. By identifying high-risk areas, the map provides actionable insights that can guide local health strategies, optimize resource distribution, and improve the efficiency of interventions. These findings contribute to enhanced public health planning and can support future regional malaria control initiatives.