Exploring the spatiotemporal heterogeneity in the relationship between human mobility and COVID-19 prevalence using dynamic time warping

TL;DR

This study investigates the complex spatiotemporal relationship between human mobility and COVID-19 cases across U.S. counties using dynamic time warping, revealing significant heterogeneity in how mobility influences disease spread over time and space.

Contribution

It introduces a novel application of dynamic time warping to analyze spatiotemporal heterogeneity in mobility-COVID-19 relationships at the county level.

Findings

Mobility-COVID-19 correlation varies across counties and over time.

Dynamic time warping identifies specific periods and locations where mobility impacts infection rates.

Significant heterogeneity exists in the mobility-infection relationship across different regions.

Abstract

Understanding where and when human mobility is associated with disease infection is crucial for implementing location-based health care policy and interventions. Previous studies on COVID-19 have revealed the correlation between human mobility and COVID-19 cases. However, the spatiotemporal heterogeneity of such correlation is not yet fully understood. In this study, we aim to identify the spatiotemporal heterogeneities in the relationship between human mobility flows and COVID-19 cases in U.S. counties. Using anonymous mobile device location data, we compute an aggregate measure of mobility that includes flows within and into each county. We then compare the trends in human mobility and COVID-19 cases of each county using dynamic time warping (DTW). DTW results highlight the time periods and locations (counties) where mobility may have influenced disease transmission. Also, the correlation between human mobility and infections varies substantially across geographic space and time in terms of relationship, strength, and similarity.