Embedded Files
Mitigation Strategies in Public Health Crises:
Mitigation Strategies in Public Health Crises:
Evaluating the successes and failures in the COVID-19 pandemic
Evaluating the successes and failures in the COVID-19 pandemic
Phase I. Introduction
Phase I. Introduction
Project Pitch, Literature Review, and EDA
Project Pitch, Literature Review, and EDA
Project Aim
Project Aim
Explore COVID-19 global datasets with mitigation and mobility information to gain insights on the pandemic and mobility trends and impact of mitigation efforts.
Visualize the trends and patterns with animated graphs and charts for easy understanding.
Provide recommendations on the most or least effective mitigation efforts for public education and policy guidelines.
Rational
Rational
Over 10 million confirmed cases and 500,000 deaths worldwide since the outbreak of “pneumonia of unknown cause” in Wuhan, China on December 31, 2019 [1,2].
Numerous mitigation efforts employed across the globe since WHO declared the outbreak a pandemic on March 11, 2020 [3,4].
CDC didn’t recommend wearing masks in public until April 3,2020 [5,6].
For individuals worldwide, the start of population quarantine and process to test and treat have been widely different [7].
Lack of uniformed mitigation guidelines by WHO, CDC, and federal governments has caused public confusion and fear [8,9].
Research Questions
Research Questions
What mitigation impacts the number of COVID-19 confirmed cases positively or negatively?
Is that impact significant enough to recommend to policy makers?
Can mobility data show how mitigation is tightly followed in each countries?
Is there any correlation between mobility trend and the number of COVID-19 confirmed cases?
Datasets
Datasets
1. Global COVID-19 confirmed cases by date: Provided by the Johns Hopkins University, COVID-19 confirmed cases were tracked in 266 countries. Data is accumulated from 1/22 with continuous update. There are total 188 countries/regions with spatial coordinates.
2. COVID-19 Mitigation effort by governments: Provided by the University of Oxford, 8 containment and closure policies, 4 economic policies, 5 health system policy by 173 countries, coded data collected from 1/1 with continuous update. There are around 30,000 rows and 41 columns.
3. Mobility data during pandemic: Provided by Google, mobility was tracked by 6 categories in 132 countries, recorded positive or negative based on baseline from 2/19 with continuous update. There are around 500,000 rows and 11 columns.
4. Countries of the world: Originated from CIA, the dataset contains 20 columns of information such as population, population density, and GDP in 227 countries. This data will be paired with other datasets.
Related Work and Literature Review
Related Work and Literature Review
At the early stages of COVID-19 outbreak, containment strategies were put into place to slow the spread. Travel restrictions and population quarantine were common approaches [7]. As the virus continued to spread and reach pandemic levels, epidemiology experts emphasized that containment alone would not help; the world needed to use mitigation strategies [11]. Public health and epidemiology experts looked to past viral outbreaks for insights - Spanish influenza (1918), SARS (2002), H1N1 (2009), MERS (2012). Many of the mitigation strategies employed basic hygiene efforts and social distancing with variable outcomes [13-15].
Today, governments across the world are testing and evaluating their own mitigation strategies using these insights and adapting to their context. The goal is to "flatten the curve" before they reach critical points in their healthcare supply chains [16]. There are many variables in play - prevalence, population mobility, public health perception, and testing and treatment supplies are highly variable between countries. As mitigation strategies, the implementation of public health hygiene and social distancing have also varied significantly between countries and regions [17]. Understanding and visualizing the data regarding the spread of COVID-19 and mitigation strategies across the world will give insights into the successes and limitations of the strategies employed and potentially the impact of certain variables on outcomes.
Methods Overview
Methods Overview
Python/ Jupyter notebook will be used for this project unless stated otherwise.
Create a choropleth map visualizing the total number of COVID-19 confirmed cases in a world map. Hover information was added to display country name and the total number of COVID-19 confirmed cases. Fig. 1(a).
Create a choropleth map visualizing prevalence rates of COVID-19 confirmed cases in a world map. Fig. 1(b).
Check if there is any trend in the total number of COVID-19 confirmed cases during each mitigation in each country. Show the trend in a graph. Fig. 5.
Check if there is any trend in the total number of COVID-19 confirmed cases related to mobility pattern in each country. Show the trend in a graph. Fig. 6.
Calculate the statistical significance of the trend.
Preliminary Exploratory Data Analysis
Preliminary Exploratory Data Analysis
Choropleth maps show Log10 scale of COVID-19 total confirmed cases in 188 countries.
Figure names have clickable links for a HTML page that has mouse over text information.
From Fig. 1(a), US, Russia, Brazil, and India have highest number of confirmed cases.
Date columns in confirmed cases data were converted to rows and merged to countries of the world data by country and date.
17 countries with different names in population data was manually corrected for merge.
7 countries/ territories/ cruise ships with missing population data were manually added.
From Fig. 1(b), Qatar, Chile, United Arab Emirate, Oman, Kuwait, and Panama have highest prevalence rates.
The mitigation data contains 13 coded columns and 155 overlap countries from the confirmed cases data.
The mobility data has 135 overlap countries from the confirmed cases data.
Phase II. Body
Phase II. Body
EDA & Model Construction
EDA & Model Construction
All the figures are created with plotly with mouseover text feature to show information. All the figures are linked to HTML pages where they can be shown as interactive format.
Exploratory Data Analysis
Exploratory Data Analysis
Fig 2. COVID-19 Spread Map visualizes the trails of virus where it was started and how it was spread over the world. The number of total confirmed cases are indicated by color and size of the bubbles.
Fig 3. COVID-19 Cases by GDP visualizes the speed of virus spread from x axes with each country's GDP on y axes. It also indicates population by bubble size, and continent by color with the mouseover text feature for information. While countries with higher GDP have spread first, countries with lower GDP also have spread later on, which indicates GDP don't impact on the spread of COVID-19.
Graph Chart.pptxTable 1. List of 25 countries &
Table 1. List of 25 countries &
External link for HTML graphs
External link for HTML graphs
Exploratory Data Analysis_continued
Exploratory Data Analysis_continued
Fig 4. COVID-19 Trend in 25 countries visualizes the total confirmed cases trend in countries with drop down menu to select each country.
Fig 5. COVID-19 Trends with Mitigation visualizes the total confirmed cases trend in each country with different types of mitigation. By selecting mitigation type, it shows when the mitigation was ordered with what severity and how that impacted the confirmed cases trend.
Fig 6. COVID-19 New Cases with Mobility Trend visualizes the new cases trend with different mobility data. By selecting location, the graph shows how mobility patterns changed in different places and how that impacted the trend of new cases.
Phase III. Conclusion
Phase III. Conclusion
Fig 5. COVID-19 patterns in 25 countries
Fig 5. COVID-19 patterns in 25 countries
As of August 8th, 2020, there has been pattern changes of COVID-19 confirmed cases in some of 25 countries. These patterns can be categorized as three groups.
Very progressive group: US, Brazil, Russia, India, Peru, Chile, Spain, Iran, Mexico, South Africa, Singapore, Vietnam
Somewhat progressive group: UK, France, Germany, Turkey, Sweden, South Korea, Slovenia, Cambodia
Stable group: New Zealand, China, Italy, Taiwan, Laos
Model Construction: Compare 3 Forecast Models with the US dataset (Train: test as 80:20 ratio)
Model Construction: Compare 3 Forecast Models with the US dataset (Train: test as 80:20 ratio)
Model 1. ARIMA
Model 1. ARIMA
Check stationarity by seasonal decompose, rolling statistics graph, and ADF test
Determine ARIMA order (p, d, q) by differencing & autocorrelation, PACF, and ACF: (1, 2, 1)
Forecast, check residual errors, and measure accuracy by RMSE, MSE, mean value, MAPE, correlation, and Min-Max Error: 94% accurate
Result
Result
Comparing predictions using test dataset with 3 different forecast models reveals that LSTM performs the best and Prophet the least.
Fig 7. 3 Forecast Models Comparison with Test Dataset
Fig 7. 3 Forecast Models Comparison with Test Dataset
Model 2. LSTM
Model 2. LSTM
Preprocess with MinMaxScaler & generate time series
Build & fit the model (relu, adam, mse)
Plot the loss and forecast
Measure accuracy by RMSE, MSE, mean valueMAPE, and correlation: 93% accurate
Model 3. Prophet
Model 3. Prophet
Prepare the dataset: change column names and convert the value to log scale
Predict (yhat, yhat_lower, yhat_upper)
Convert back log scale
Plot component plot, forecast plot
Measure accuracy by RMSE, MSE, mean value, MAPE, correlation, and Min-Max Error: 86% accurate
Fig 8. Forecast pandemic trend with LSTM
Fig 8. Forecast pandemic trend with LSTM
All forecasts and dashboard are available GitHub
Forecast Result
Forecast Result
Countries with strong mitigation: confirmed cases stay stable in the future
Countries with weak mitigation: confirmed cases progress in the future
Few countries with dramatic changes: possible super spreaders
Future Directions
Future Directions
What could be done differently?
What could be done differently?
Build more robust LSTM model.
Make a complete loop system to reproduce graphs by country.
What can be done next?
What can be done next?
Build a system to automatically update up to date graphs.
Build a dashboard where all the graphs can be seen in a page.
Forecast before and after mitigation to see the direct impact.
References
References
Gardner L. Mapping 2019-nCoV. Johns Hopkins Whiting School of Engineering| Center for Systems Science and Engineering. 2020.
WHO Director-General's opening remarks at the media briefing on COVID-19 - 11 March 2020. WHO. 2020.
CDC, Implementation of Mitigation Strategies for Communities with Local COVID-19 Transmission.
Hub staff post. Understanding changing guidance on mask use. Johns Hopkins University. 2020.
Geggel L. Everyone should wear face 'masks' in public, CDC now recommends. Live science. 2020.
Blanco A, Bellack M. How countries around the world have tried to contain the coronavirus. The Washington Post. 2020.
Edmondson C, Benner K. After Days of Anxiety and Confusion, Government Workers Told to Stay Home. The New York Times. 2020.
Sun L, Wan W, Abutaleb Y. A plan to defeat coronavirus finally emerges, but it’s not from the White House. The Washington Post. 2020.
Johnson T, Fritz A. You’re under a stay-at-home order? Here’s what that means in your state. The Washington Post. 2020.
Parodi SM, Liu VX. From Containment to Mitigation of COVID-19 in the US. JAMA. 2020;323(15):1441–1442. doi:10.1001/jama.2020.3882
Walensky RP, del Rio C. From Mitigation to Containment of the COVID-19 Pandemic: Putting the SARS-CoV-2 Genie Back in the Bottle. JAMA. 2020;323(19):1889–1890. doi:10.1001/jama.2020.6572
Ahmed F, Zviedrite N, Uzicanin A. Effectiveness of workplace social distancing measures in reducing influenza transmission: a systematic review. BMC Public Health. 2018;18(1):518. Published 2018 Apr 18. doi:10.1186/s12889-018-5446-1
Liang M, Gao L, Cheng C, et al. Efficacy of face mask in preventing respiratory virus transmission: a systematic review and meta-analysis. medRxiv 2020.04.03.20051649; doi: https://doi.org/10.1101/2020.04.03.20051649
Saunders-Hastings P, Crispo J, Sikora L. Effectiveness of personal protective measures in reducing pandemic influenza transmission: A systematic review and meta-analysis. j.epidem. 2017. doi: https://doi.org/10.1016/j.epidem.2017.04.003
Roberts S. Flattening the Coronavirus Curve. The New York Times. 2020.
Bruinen de Bruin, Y., Lequarre, A. S., McCourt, J., et al. Initial impacts of global risk mitigation measures taken during the combatting of the COVID-19 pandemic. Safety science. 2020;128, 104773.
Bricker D. More people say they’re wearing masks to protect themselves from COVID-19 since March. Ipsos. 2020.
Ngo M. Introducing: International Social Distancing Data. Unacast. 2020
Ritchie H, et al. Coronavirus Pandemic (COVID-19). Our World in Data. 2020.
Chimmula V, Zhang L. Time series forecasting of COVID-19 transmission in Canada using LSTM networks. Chaos, Solutions and Fractals. 2020; 135: 109864.
Page updated
Report abuse