Geographic sensitivity of COVID-19 outcomes: a multiscale analysis of temporal and spatial non-stationarity in the contiguous United States

Abstract Background COVID-19 outcomes varied widely across the United States. Many county-level studies summarize outcomes cumulatively and estimate a single association for each predictor across all counties. We examined temporal changes and spatial variation in these associations. Methods We assembled a dataset covering 3,103 US counties, linking daily reported COVID-19 cases and deaths, January 2020 to March 2023, to 43 socio-environmental and resilience county characteristics checked against primary sources. Deaths were modeled as counts with a population offset using negative binomial regression, cumulatively and within seven calendar windows. Predictive performance was assessed by spatially blocked cross-validation. Spatial models included spatial-lag, geographically weighted (GWR) and multiscale geographically weighted (MGWR) regression; coefficient variability was tested by coordinate permutation. Age composition was assessed by indirect standardization. Results County characteristics predicted cumulative outcomes better than population size alone. Out-of-sample deviance explained for cumulative mortality was 0.455 for the 43 county characteristics against 0.106 for population size alone, and for cases 0.214 against 0.005. The population-size association with mortality changed over time. In the size-only model, the incidence-rate ratio per standard deviation of log population was 1.478 (95% CI 1.410 to 1.550) between January and May 2020, and 0.802 to 0.883 in the six later windows (each P < 0.001). Adjusting for size and all 43 characteristics left the gradient non-significant in six of seven windows. MGWR fitted better than GWR (corrected Akaike Information Criterion 5665.0 against 5770.4) cumulatively and in all seven windows, though neither outperformed the global models out of sample. Of 999 planned coordinate permutations, 998 converged; only ultraviolet exposure (Holm-adjusted P = 0.009–0.018) and household composition and disability (P = 0.016–0.024) exceeded their permutation reference under family-wise correction, and physical inactivity under false-discovery-rate control alone; the intercept and population-size surfaces did not. Conclusions County characteristics predicted cumulative mortality better than population size alone, while the population-size gradient reversed after the first wave. Cumulative estimates can therefore obscure period-specific associations. Local spatial models improved in-sample fit but did not outperform global models in held-out counties.

Authors

Institutions

Publication Details

Journal
BMC Public Health
Published
2026-11-14
DOI
https://doi.org/10.1186/s12889-026-30317-9
Primary Topic
COVID-19 epidemiological studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Geographic sensitivity of COVID-19 outcomes: a multiscale analysis of temporal and spatial non-stationarity in the contiguous United States

David H. An, Ryan Chen¹, Ethan Chen¹
2 links
BMC Public Health
COVID-19 epidemiological studies
article

Geographic sensitivity of COVID-19 outcomes: a multiscale analysis of temporal and spatial non-stationarity in the contiguous United States

David H. An, Ryan Chen¹, Ethan Chen¹
article en

Abstract

Abstract Background COVID-19 outcomes varied widely across the United States. Many county-level studies summarize outcomes cumulatively and estimate a single association for each predictor across all counties. We examined temporal changes and spatial variation in these associations. Methods We assembled a dataset covering 3,103 US counties, linking daily reported COVID-19 cases and deaths, January 2020 to March 2023, to 43 socio-environmental and resilience county characteristics checked against primary sources. Deaths were modeled as counts with a population offset using negative binomial regression, cumulatively and within seven calendar windows. Predictive performance was assessed by spatially blocked cross-validation. Spatial models included spatial-lag, geographically weighted (GWR) and multiscale geographically weighted (MGWR) regression; coefficient variability was tested by coordinate permutation. Age composition was assessed by indirect standardization. Results County characteristics predicted cumulative outcomes better than population size alone. Out-of-sample deviance explained for cumulative mortality was 0.455 for the 43 county characteristics against 0.106 for population size alone, and for cases 0.214 against 0.005. The population-size association with mortality changed over time. In the size-only model, the incidence-rate ratio per standard deviation of log population was 1.478 (95% CI 1.410 to 1.550) between January and May 2020, and 0.802 to 0.883 in the six later windows (each P < 0.001). Adjusting for size and all 43 characteristics left the gradient non-significant in six of seven windows. MGWR fitted better than GWR (corrected Akaike Information Criterion 5665.0 against 5770.4) cumulatively and in all seven windows, though neither outperformed the global models out of sample. Of 999 planned coordinate permutations, 998 converged; only ultraviolet exposure (Holm-adjusted P = 0.009–0.018) and household composition and disability (P = 0.016–0.024) exceeded their permutation reference under family-wise correction, and physical inactivity under false-discovery-rate control alone; the intercept and population-size surfaces did not. Conclusions County characteristics predicted cumulative mortality better than population size alone, while the population-size gradient reversed after the first wave. Cumulative estimates can therefore obscure period-specific associations. Local spatial models improved in-sample fit but did not outperform global models in held-out counties.

BMC Public Health
Harvard University (US)
Good health and well-being
COVID-19 epidemiological studies
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon