Early detection of change patterns in COVID-19 incidence and the implementation of public health policies: A multi-national study

• Main Authors: Steven S. Coughlin, Ayten Yiǧiter, Hongyan Xu, Adam E. Berman, Jie Chen
• Format: Article
• Language: English
• Published: Elsevier 2021-11-01
• Series: Public Health in Practice
• Subjects: COVID-19; Health policy; Incidences; B-spline trend fitting and prediction; Change point models; Confidence intervals
• Online Access: http://www.sciencedirect.com/science/article/pii/S266653522030063X

Objectives: The COVID-19 pandemic caused by the novel SARS-CoV-2 coronavirus has drastically altered the global realities. Harnessing national scale data from the COVID-19 pandemic may better inform policy makers in decision making surrounding the reopening of society. We examined country-level, daily-confirmed, COVID-19 case data from the World Health Organization (WHO) to better understand the comparative dynamics associated with the ongoing global pandemic at a national scale. Study design: Observational study. Methods: We included data from 20 countries in Europe, the Americas, Africa, Eastern Mediterranean and West Pacific regions, and obtained the aggregated daily new case data for the European Union including 27 countries. We utilized an innovative analytic approach by applying statistical change point models, which have been previously employed to model volatility in stock markets, changes in genomic data, and data dynamics in other scientific disciplines, to segment the transformed case data. This allowed us to identify possible change or turning points as indicated by the dynamics of daily COVID-19 incidences. We also employed B-spline regression models to express the estimated (predicted) trend of daily new incidences for each country’s COVID-19 disease burden with the identified key change points in the model. Results: We identified subtle, yet different change points (translated to actual calendar days) by either the mean and variance change point model with small p-values or by a Bayesian online change point algorithm with large posterior probability in the trend of COVID-19 incidences for different countries. We correlated these statistically identified change points with evidence from the literature surrounding these countries’ policies regarding opening and closing of their societies in an effort to slow the spread of COVID-19. The days when change points were detected were ahead of the actual policy implementation days, and in most of the countries included in this study the decision lagged the change point days too long to prevent potential widespread extension of the pandemic. Conclusions: Our models describe the behavior of COVID-19 prevalence at a national scale and identify changes in national disease burden as relating to chronological changes in restrictive societal activity. Globally, social distancing measures may have been most effective in smaller countries with single governmental and public health organizational structures. Further research examining the impact of heterogeneous governmental responses to pandemic management appears warranted.