Evaluation of Various Health Interventions to Curb the Spread of COVID-19 in the United States of America

Authors: Zhouxuan Li, Kai Zhang, Tao Xu, Deng HW, Eric Boerwinkle, Xiong M

Journal: Epidemiology International Journal. Article type: Short Communication. Published: 2020-08-12. DOI: https://doi.org/10.23880/eij-16000156

Abstract

As of July 9, 2020, the cumulative cases of Covid-19 in the US passed 3,055,491, including 132,310 deaths, causing a serious public health crisis. There is an urgent need to curb the spread of Covid-19. In the absence of vaccines and effective medication, non-pharmaceutical interventions are the only option to mitigate the spread of Covid-19. To accurately estimate the potential impact of different non-pharmaceutical measures on containing Covid-19 is crucial for planning the most effective interventions to curb the spread of Covid-19. We applied time series regression models to the surveillance data of lab-confirmed Covid-19 cases in the 10 states of the US up to June 22, 2020 in order to evaluate the contributions of the Google mobility indexes, the rate of the virus test and protest to the number of the new cases of Covid-19. We found that the reason for the well-controlled spread of Covid-19 in NY and CT was more likely due to strong health interventions and reason for the less well-controlled spread of Covid-19 in CA, TX, FL and GA was due to weak health interventions. VT, NH and WV were less affected states.

Keywords: COVID-19, Public Health Interventions, Time Series, Transmission Dynamics, Control of the Spread

Introduction

As of July 9, 2020, the number of laboratory confirmed cases of Covid-19 in the US passed 3,055,491, including 132,310 deaths, causing a serious public health crisis. There is an urgent need to curb the spread of Covid-19 [1]. In the absence of vaccines and effective medication, non-pharmaceutical interventions are the only option to mitigate the spread of Covid-19 [2]. Some studies have shown that moderate interventions could reduce the size of the epidemic, but more intensive intervention measures would be required to curb the spread of Covid-19 [3,4]. To accurately estimate the potential impact of different non-pharmaceutical measures on containing Covid-19 has been crucial for planning the most effective interventions to curb the spread of Covid-19. We applied the time series regression models [5] to the surveillance data of laboratory confirmed Covid-19 cases in 10 states in the US up to June 22, 2020, to evaluate the contributions of the Google mobility indexes, the rate of the virus testing and the number of attendees in protesting to the number of the new cases of Covid-19.

Methods

Consider a time series regression model2 where

$$y_t = \beta_0 + \beta_1 x_t + e_t$$

and

$$y_t = \beta_0 + \sum_{j=1}^{k} \beta_j x_j^t + e_t$$

where $y_t$ is the number of new cases of Covid-19, $x_t$ and are predictors including six Google mobility indexes, the rate of the virus testing and the number of attendees in the protest, and k=8.

The coefficient of determination, or R2 between the number of new cases of Covid-19 and the Google mobility indexes, the rate of the virus testing and the number of attendees in the protest is used to measure the proportion of variation in the number of new cases of Covid-19 that is explained by the Google mobility indexes, the rate of the virus testing and the number of attendees in the protest. The simple R2 is also equal to the square of the correlation between the number of new cases of Covid-19 and the individual intervention factor. The Bartlett’s test is used to test for correlation between two time series [6].

Data Collections

Data on the number of confirmed and new cases of Covid-19 in 10 states in the US from March 5, 2020 to June 22, 2020 were obtained from the John Hopkins Coronavirus Resource Center (https://coronavirus.jhu.edu/MAP.HTML).

Results

Five less controlled states (CA, AZ, FL, TX and GA) and five well controlled states (NY, CT, NH, VT and WV) were selected for the analysis. Table 1 summarized individual R2 of six Google mobility indexes, the rate of testing and the number of attendees in the protest and their total R2 in 10 states. The P-values for testing their association with the number of new cases were summarized in Table 2. The total R2 in the 10 states varied from 0.2610 (VT) to 0.9416 (WV). Eight factors explained equal or more than 50% of the variation in the number of new cases in six states: AZ, CA, NY, NH, TX and WV. The top 5 states with the highest R2 for the rate of virus testing were AZ (0.6579), CA (0.6120), WV(0.5464), TX (0.4254) and NH (0.3788), which was also implied by the association tests (AZ (P-value < 1.04E-08), CA (P-value < 2.0E-16), WV(P-value

Discussion

We showed that the rate of testing was associated with most ten states. We found that the spread of Covid-19 in NY and CT being well controlled was due to strong health interventions and the spread of Covid-19 in CA, TX, FL and GA being less controlled was due to weak health interventions. VT, NH and WV were less affected states.

Acknowledgement

HW Deng was partially supported by NIH grants U19AG05537301 and R01AR069055. Momiao Xiong was partially supported by NIH grants U19AG05537301. Thank Sara Barton for editing the manuscript.

References

1. Callaway E (2020) Time to use the p-word? Coronavirus enters dangerous new phase. Nature.

2. Irfan U (2020) The math behind why we need social distancing, starting right now. Vox.

3. Johansson MA, Apfeldorf KM, Dobson S, Devita J, Buczak AL, et al. (2019) An open challenge to advance probabilistic forecasting for dengue epidemics. PNAS 116(48): 24268-24274.

4. Davies MG, Kucharski AJ, Eggo RM, Gimma A, Edmunds WJ (2020) Effects of Non-Pharmaceutical Interventions on COVID-19 Cases, Deaths, and Demand for Hospital Services in the UK: A Modelling Study. Lancet Public Health 5(7): 375-385.

5. Hyndman RJ, Athanasopoulos G (2018) Forecasting: principles and practice. O Texts.

6. Brockwell PJ, Davis RA (2016) Introduction to Time Series and Forecasting. Springer Texts in Statistics.

Total and individual R2 between the number of new cases and the Google mobility indexes, rate of the test and the number

P-values for testing the significance of Google mobility indexes, test rate and the number of attendees. Retail: Retail

References

  1. Callaway E (2020) Time to use the p-word? Coronavirus enters dangerous new phase. Nature.
  2. Irfan U (2020) The math behind why we need social distancing, starting right now. Vox.
  3. Johansson MA, Apfeldorf KM, Dobson S, Devita J, Buczak AL, et al. (2019) An open challenge to advance probabilistic forecasting for dengue epidemics. PNAS 116(48): 24268-24274.
  4. Davies MG, Kucharski AJ, Eggo RM, Gimma A, Edmunds WJ (2020) Effects of Non-Pharmaceutical Interventions on COVID-19 Cases, Deaths, and Demand for Hospital Services in the UK: A Modelling Study. Lancet Public Health 5(7): 375-385.
  5. Hyndman RJ, Athanasopoulos G (2018) Forecasting: principles and practice. O Texts.
  6. Brockwell PJ, Davis RA (2016) Introduction to Time Series and Forecasting. Springer Texts in Statistics.

PDF of this article

Related articles