Background: Attempts to quantify effect sizes of non-pharmaceutical interventions
(NPI) to control COVID-19 in the US have not accounted for heterogeneity in social or
environmental factors that may influence NPI effectiveness. This study quantifies national
and sub-national effect sizes of NPIs during the early months of the pandemic in the US.
Methods: Daily county-level COVID-19 cases and deaths during the first wave (January
2020 through phased removal of interventions) were obtained. County-level cases,
doubling times, and death rates were compared to four increasingly restrictive NPI
levels. Socio-demographic, climate and mobility factors were analyzed to explain and
evaluate NPI heterogeneity, with mobility used to approximate NPI compliance. Analyses
were conducted separately for the US and for each Census regions (Pacific, Mountain,
east/West North Central, East/West South Central, South Atlantic, Middle Atlantic and
New England). A stepped-wedge cluster-randomized trial analysis was used, leveraging
the phased implementation of policies.
Results: Aggressive (level 4) NPIs were associated with slower COVID-19 propagation,
particularly in high compliance counties. Longer duration of level 4 NPIs was associated
with lower case rates (log beta -0.028, 95% CI -0.04 to -0.02) and longer doubling
times (log beta 0.02, 95% CI 0.01–0.03). Effects varied by Census region, for example,
level 4 effects on doubling time in Pacific states were opposite to those in Middle
Atlantic and New England states. NPI heterogeneity can be explained by differential
timing of policy initiation and by variable socio-demographic county characteristics that
predict compliance, particularly poverty and racial/ethnic population. Climate exhibits
relatively consistent relationships across Census regions, for example, higher minimum
temperature and specific humidity were associated with lower doubling times and higher
death rates for this period of analysis in South Central, South Atlantic, Middle Atlantic,
and New England states.
Conclusion and Relevance: Heterogeneity exists in both the effectiveness of NPIs
across US Census regions and policy compliance. This county-level variability indicates
that control strategies are best designed at community-levelswhere policies can be tuned
based on knowledge of local disparities and compliance with public health ordinances.
