Dataset Information

Using Satellite Images and Deep Learning to Identify Associations Between County-Level Mortality and Residential Neighborhood Features Proximal to Schools: A Cross-Sectional Study.

ABSTRACT: What is the relationship between mortality and satellite images as elucidated through the use of Convolutional Neural Networks? Background: Following a century of increase, life expectancy in the United States has stagnated and begun to decline in recent decades. Using satellite images and street view images, prior work has demonstrated associations of the built environment with income, education, access to care, and health factors such as obesity. However, assessment of learned image feature relationships with variation in crude mortality rate across the United States has been lacking. Objective: We sought to investigate if county-level mortality rates in the U.S. could be predicted from satellite images. Methods: Satellite images of neighborhoods surrounding schools were extracted with the Google Static Maps application programming interface for 430 counties representing ~68.9% of the US population. A convolutional neural network was trained using crude mortality rates for each county in 2015 to predict mortality. Learned image features were interpreted using Shapley Additive Feature Explanations, clustered, and compared to mortality and its associated covariate predictors. Results: Predicted mortality from satellite images in a held-out test set of counties was strongly correlated to the true crude mortality rate (Pearson r = 0.72). Direct prediction of mortality using a deep learning model across a cross-section of 430 U.S. counties identified key features in the environment (e.g., sidewalks, driveways, and hiking trails) associated with lower mortality. Learned image features were clustered, and we identified 10 clusters that were associated with education, income, geographical region, race, and age. Conclusions: The application of deep learning techniques to remotely-sensed features of the built environment can serve as a useful predictor of mortality in the United States. Although we identified features that were largely associated with demographic information, future modeling approaches that directly identify image features associated with health-related outcomes have the potential to inform targeted public health interventions.

SUBMITTER: Levy JJ

PROVIDER: S-EPMC8602058 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:Objective To examine associations between neighborhood built environment (BE) variables, residential property values, and longitudinal 1- and 2-year changes in body mass index (BMI). Methods The Seattle Obesity Study III was a prospective cohort study of adults with geocoded residential addresses, conducted in King, Pierce, and Yakima Counties in Washington State. Measured heights and weights were obtained at baseline (n = 879), year 1 (n = 727), and year 2 (n = 679). Tax parcel residential property values served as proxies for individual socioeconomic status. Residential unit and road intersection density were captured using Euclidean-based SmartMaps at 800 m buffers. Counts of supermarket (0 versus. 1+) and fast-food restaurant availability (0, 1–3, 4+) were measured using network based SmartMaps at 1600 m buffers. Density measures and residential property values were categorized into tertiles. Linear mixed-effects models tested whether baseline BE variables and property values were associated with differential changes in BMI at year 1 or year 2, adjusting for age, gender, race/ethnicity, education, home ownership, and county of residence. These associations were then tested for potential disparities by age group, gender, race/ethnicity, and education. Results Road intersection density, access to food sources, and residential property values were inversely associated with BMI at baseline. At year 1, participants in the 3rd tertile of density metrics and with 4+ fast-food restaurants nearby showed less BMI gain compared to those in the 1st tertile or with 0 restaurants. At year 2, higher residential property values were predictive of lower BMI gain. There was evidence of differential associations by age group, gender, and education but not race/ethnicity. Conclusion Inverse associations between BE metrics and residential property values at baseline demonstrated mixed associations with 1- and 2-year BMI change. More work is needed to understand how individual-level sociodemographic factors moderate associations between the BE, property values, and BMI change. Highlights • Strong, inverse cross-sectional relationships between the built environment, residential property values (a proxy for individual socioeconomic status), and measured BMI were observed.• Measures of the built environment and residential property values showed modest and inconsistent associations with 1- and 2-year BMI change.• There was suggestive evidence that age may moderate the association between urban density and 1- and 2-year BMI change while education may moderate the association between residential property values and 2-year BMI change.

Project description:BackgroundEpidemiological studies have identified associations between air pollution and green space access with type 2 diabetes in adults. However, it remains unclear to what extent associations with greenness are attributable to air pollution exposure.ObjectivesWe aimed to investigate associations between long-term exposure to air pollution and satellite-derived greenness with insulin resistance in adolescents.MethodsA total of 837 participants of two German birth cohorts (LISAplus and GINIplus) were included in the analysis. Generalized additive models were used to determine the association of individual satellite-derived greenness defined by the Normalized Difference Vegetation Index (NDVI), long-term air pollution exposure estimated by land-use regression (LUR) models with insulin resistance (HOMA-IR) in 15-year-old adolescents. Models were adjusted for study area, cohort, socioeconomic, and individual characteristics such as body mass index, physical activity, and smoking.ResultsIncreases of 2 SDs in nitrogen dioxide (NO2; 8.9 ?g/m3) and particulate matter ? 10 ?m in diameter (PM10; 6.7 ?g/m3) were significantly associated with 11.4% (95% CI: 4.4, 18.9) and 11.4% (95% CI: 0.4, 23.7) higher HOMA-IR. A 2-SD increase in NDVI in a 1,000-m buffer (0.2 units) was significantly associated with a lower HOMA-IR (-7.4%; 95% CI: -13.3, -1.1). Associations tended to be stronger in adolescents who spent more time outside and in those with lower socioeconomic status. In combined models including both air pollution and greenness, only NO2 remained significantly associated with HOMA-IR, whereas effect estimates for all other exposures attenuated after adjustment for NO2.ConclusionsNO2, often considered as a marker of traffic, was independently associated with insulin resistance. The observed association between higher greenness exposure and lower HOMA-IR in adolescents might thus be attributable mainly to the lower co-exposure to traffic-related air pollution.CitationThiering E, Markevych I, Brüske I, Fuertes E, Kratzsch J, Sugiri D, Hoffmann B, von Berg A, Bauer CP, Koletzko S, Berdel D, Heinrich J. 2016. Associations of residential long-term air pollution exposures and satellite-derived greenness with insulin resistance in German adolescents. Environ Health Perspect 124:1291-1298;?http://dx.doi.org/10.1289/ehp.1509967.

Dataset Information

Using Satellite Images and Deep Learning to Identify Associations Between County-Level Mortality and Residential Neighborhood Features Proximal to Schools: A Cross-Sectional Study.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets