Spatial heterogeneity can lead to substantial local variations in COVID-19 timing and severity

Loring J. Thomas, Peng Huang, Fan Yin, Xiaoshuang Iris Luo, Zack W. Almquist, John R. Hipp, Carter T. Butts

Introduction

Standard epidemiological models often assume that populations within a city or county mix uniformly. These "well-mixed" models (such as the standard SIR—Susceptible, Infected, Recovered framework) imply that any two individuals in a metropolitan area have a roughly equal chance of interacting. However, sociological research consistently demonstrates that human interaction is deeply tied to geography; individuals are far more likely to interact with immediate neighbors than with those across town. This paper investigates whether local spatial heterogeneity significantly impacts disease dynamics or if the "well-mixed" assumption remains a reliable approximation for urban modeling.

Methodology

The researchers simulated COVID-19 spread across 19 U.S. cities using a network diffusion approach designed to isolate the effects of geography.

  • Spatial Bernoulli Models: Contact networks were constructed using block-level U.S. Census data. Connections were established based on "Spatial Interaction Functions" (SIFs), where the probability of a tie between two individuals decays as a power law relative to the distance between them.
  • Disease Simulation: Infection spread was modeled as a continuous-time Poisson process. To isolate the pure effect of spatial network structure, the authors intentionally omitted social distancing measures, lockdowns, and demographic variables like age-based susceptibility.
  • Impact Assessment: The study mapped infections to local hospitals using Voronoi tessellations to estimate healthcare strain and tracked "social exposure"—the timeline of when individuals first encountered morbidity or mortality within their personal networks.

The "Smoothness Illusion" and Local Variance

A primary finding of the study is the "smoothness illusion." While aggregate, city-wide infection curves often appear smooth and follow standard SIR predictions, this macro-level view masks chaotic and disjointed local outbreaks.

Figure 1: Infection curves for Seattle. The red line represents total city infections, while the grey lines represent individual census tracts, illustrating the massive variance in local peak timing and severity.

Because people interact primarily within local clusters, the disease spreads rapidly within a neighborhood but may stall at its boundaries. Transmission to adjacent areas requires "bridging ties"—rare long-distance interactions—which creates significant time lags between local outbreaks. Consequently, peak infection days across census tracts within the same city can vary by weeks or even months.

Statistical Evidence of Asynchrony

The lack of synchrony between local neighborhoods is statistically profound. Across the 19 simulated cities, the mean correlation of infection curves between different tracts was only approximately 0.2.

Figure 4: Distribution of correlations between infection curves (left) and the proportion of variance explained by the first eigenvector (right) across the studied cities.

Principal Component Analysis (PCA) further supported this finding, showing that the first principal component accounted for only about 35% of the variance in local trajectories. This indicates that local curves are highly distinct and only weakly related to the city-wide average.

Implications for Public Health and Logistics

This spatial heterogeneity has massive implications for resource management and public risk perception:

  • Hospital Strain: Spatially homogeneous models predict that all hospitals in a city will surge simultaneously. In contrast, this spatial model shows some hospitals reaching maximum capacity within the first month while others remain largely empty for nearly a year.
  • Risk Perception: Because the disease hits different areas at different times, "on the ground" experiences vary wildly. Residents in an untouched tract may perceive the pandemic as overblown, while a neighborhood just miles away faces a crisis.
  • Social Exposure: In cities like Baltimore, residents in the urban core were exposed to morbidity in their personal networks much faster than those on the periphery, creating a stark divide in the psychological burden of the pandemic.

Limitations and Future Directions

The study utilized a "no-mitigation" scenario, excluding the effects of mask mandates, behavioral changes, or demographic vulnerabilities. Additionally, the hospital model assumed patients strictly visit the nearest facility, ignoring real-world load-balancing.

Future modeling that incorporates spatial heterogeneity could significantly improve healthcare logistics. By recognizing that hospitals peak at different times, authorities could implement dynamic resource sharing and patient transfer arrangements more effectively than current spatially homogeneous models allow.

Summary updated

Abstract

Standard epidemiological models for COVID-19 employ variants of compartment (SIR or susceptible-infectious-recovered) models at local scales, implicitly assuming spatially uniform local mixing. Here, we examine the effect of employing more geographically detailed diffusion models based on known spatial features of interpersonal networks, most particularly the presence of a long-tailed but monotone decline in the probability of interaction with distance, on disease diffusion. Based on simulations of unrestricted COVID-19 diffusion in 19 US cities, we conclude that heterogeneity in population distribution can have large impacts on local pandemic timing and severity, even when aggregate behavior at larger scales mirrors a classic SIR-like pattern. Impacts observed include severe local outbreaks with long lag time relative to the aggregate infection curve, and the presence of numerous areas whose disease trajectories correlate poorly with those of neighboring areas. A simple catchment model for hospital demand illustrates potential implications for health care utilization, with substantial disparities in the timing and extremity of impacts even without distancing interventions. Likewise, analysis of social exposure to others who are morbid or deceased shows considerable variation in how the epidemic can appear to individuals on the ground, potentially affecting risk assessment and compliance with mitigation measures. These results demonstrate the potential for spatial network structure to generate highly nonuniform diffusion behavior even at the scale of cities, and suggest the importance of incorporating such structure when designing models to inform health care planning, predict community outcomes, or identify potential disparities.
BibTeX
@inproceedings{thomas2020spatial,
  author = {Loring J. Thomas and Peng Huang and Fan Yin and Xiaoshuang Iris Luo and Zack W. Almquist and John R. Hipp and Carter T. Butts},
  title = {{Spatial heterogeneity can lead to substantial local variations in COVID-19 timing and severity}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2020},
  doi = {10.1073/pnas.2011656117},
}
Figure