Strengthening policy coding methodologies to improve COVID-19 disease modeling and policy responses: a proposed coding framework and recommendations

Abstract

Background: In recent months, multiple efforts have sought to characterize COVID-19 social distancing policy responses. These efforts have used various coding frameworks, but many have relied on coding methodologies that may not adequately describe the gradient in social distancing policies as states “re-open.” Methods: We developed a COVID-19 social distancing intensity framework that is sufficiently specific and sensitive to capture this gradient. Based on a review of policies from a 12 U.S. state sample, we developed a social distancing intensity framework consisting of 16 domains and intensity scales of 0–5 for each domain. Results: We found that the states with the highest average daily intensity from our sample were Pennsylvania, Washington, Colorado, California, and New Jersey, with Georgia, Florida, Massachusetts, and Texas having the lowest. While some domains (such as restaurants and movie theaters) showed bimodal policy intensity distributions compatible with binary (yes/no) coding, others (such as childcare and religious gatherings) showed broader variability that would be missed without more granular coding. Conclusion: This detailed intensity framework reveals the granularity and nuance between social distancing policy responses. Developing standardized approaches for constructing policy taxonomies and coding processes may facilitate more rigorous policy analysis and improve disease modeling efforts.