How are Countries Controlling COVID-19’s First Wave?
The pandemic resulting from the novel SARS-Cov-2 virus has been an unprecedented event, having dire consequences for the health and economies of countries on a global scale. As we now know the virus operates in cyclical patterns of infection known as waves, it is important to study the nature of these epidemiological events using trusted statistical methods. In this analysis, we focus on the first wave of Coronavirus in 5 countries. This aggregation is important, as we have now seen that the heterogeneity of initial response and policies put in place by individual countries has caused extremely different trajectories in terms of disease spread and virological evolution.
The countries chosen for this analysis were Canada, Australia, Germany, Finland, and Israel. We want to compare the pandemic’s nature in Canada with other countries that have a similar GDP/Capita, with a clearly identified first wave, and accurately recorded recovered and death numbers. Choosing countries that are similar to Canada in terms of economic standing can help us to use it as a controlling factor. Then, between these countries, we will compute the number of people infected during the first wave, as well as how fast COVID spread within each country based on trusted methods used by epidemiologists. Finally, based on our analysis, we will examine different Covid-19 prevention measures to see which countries were more successful in flattening the curve during the first COVID wave.
The following graph shows daily cases reported in each of the 5 countries and using the drop-down menu, we can see each countries’ daily cases and the ending time of the first wave.
The data were trimmed to include only the timeline for the first wave, specific for each country. The first wave was identified by looking at the curve of the active cases in the graph below. The first wave’s ending date can be identified when the number of active cases is close to 0. As the first wave comes to an end, people have either recovered or died; thus the number of active cases decreases as the rate of increase in new daily cases decreases simultaneously.
A table displaying the duration, starting date, and ending date of each country’s first wave is shown below. Australia and Israel are having the shortest duration, followed by Germany, Canada, and Finland.
Now we will be looking at multiple ratios to further determine how countries are controlling the first COVID wave.
Point Prevalence Rate:
In order to further examine how affected a population is and to measure the intensity of the disease beyond simple visualizations, we used the prevalence and incidence rates. The prevalence rates tell us how many people are being infected at a given point in time and the incidence rates tell us how fast the disease is spreading.
Starting with the point prevalence (PP), we can see from the graph below how each country is affected by COVID’s first wave given the population size at all points in time. Israel had the highest PP ratio, but it dramatically decreased around day 85. Canada’s PP ratio was high from day 100 until day 177, then it had sharply decreased afterward. Germany’s PP ratio peaked on day 76 and slowly reduced afterward. Finland and Australia were the least affected.
Period Prevalence Rate:
Now looking at the period prevalence in the month of February, the PP rate is the highest for Germany, followed by Australia, and Israel. This means that those countries were the most affected given their respective population size in the month of February. Germany’s PP rate was close to 0.000094%. The rate represents the percentage of active cases per population in February.
One month later, all the PP ratios had increased. 0.086% of the population was newly infected in Germany, 0.066% for Israel, 0.026% for Finland, 0.023% for Canada, and 0.018% for Australia. Within the 5 countries, Australia had a relatively lower PP rate.
Looking at the whole first wave period, The percentage of newly active cases per population is the highest for Canada, followed by Germany, and Israel.
Incidence Rate:
Now looking at the incidence proportion, this ratio indicates the proportion of the at-risk population that became new cases during a given period of time. In February (the beginning of the pandemic), given the size of the at-risk population, Germany has the worst incidence ratio of 0.000085%. The ratio represents the percentage of the population that became new cases in the month of February.
One month later, all the incidence rates had increased with the highest rate in Germany, followed by Israel, Finland, Canada, and Australia.
Looking at the whole first wave period, 0.30% of the at-risk population became new cases in Canada, followed by Germany and Israel, with an incidence rate of 0.23% and 0.20% respectively.
Hazard Rate:
The hazard rate is a modified version of the incidence rate. It calculates the instantaneous rate of change of the disease at time t. Looking only at the first wave data, from the hazard rate plots we are able to see the likelihood of contracting the disease changes over time. For instance, we can see that in Australia, the population was generally much less likely to contract the disease. In addition, we can see that the peak hazard rate lasted for a very limited number of days compared to countries like Canada, whose hazard rate stayed at higher values for many days.
Mortality Rate:
Mortality is an important metric to measure across Countries, as it communicates how deadly the disease is. We can measure this through the case-fatality ratio and outcome-fatality ratio. Looking at the case-fatality ratio, which tells us how many deaths resulted from the number of cases on a given day, we again observe the superiority of Australia. Even at the peak, the CF value stayed below 4 and tapered off faster than any other country. Other countries also seemed to reach the peak of CF slower than Australia but had difficulty reducing this rate.
The outcome-fatality ratio differs from the above in that it tells us how many deaths resulted from all completed (resolved) cases on a given day. We can see that Australia and Israel were the most successful in containing deaths in this manner, while Canada’s OF value had the highest peak out of all countries.
Reproductive Rate:
From the graph above, we can see the number of susceptible, infected, and removed people for each country. To measure how a disease spreads, we need to know how fast it reproduces.
R-nought represents the disease’s reproductive rate and is a good indicator to tell us how the epidemic will play out if we do not do anything. This number represents the average number of subsequent people directly infected by a single case. If R-nought is greater than one, then 1 person will infect more than 1 person and the disease will become an epidemic. A value equal to 1 indicates the disease will become endemic, and a value smaller than 1 indicates that the disease will eventually die out. The scientific community originally carried out a prediction of R-nought between 1.4 to 5.7 for COVID-19.
Beginning of pandemic (before any intervention was taken):
We can observe that from February 22 to 28, the R-noughts are above 1 for Israel, Canada, and Germany, and exactly 1 for Australia and Finland. Based on the data, COVID is more infectious in countries with an R-nought above 1. R-nought will not only depend on the nature of the disease but also the way we interact, the environment, and the hygiene conditions. This rate can be properly reduced using proper measures and bring R-nought below 1.
After Intervention:
Threes months after the beginning of the COVID spread, countries have taken preventive measures to decrease the infection rate, such as social distancing, and wearing masks. We can observe that from May 1 to 7, the R-noughts have dropped relative to the graph above. Australia, Germany, and Israel have an R-nought of less than 1, meaning that the disease will eventually die out. Canada and Finland have an R-nought close to 1, so their measures were less effective.
Reproductive Rate During First Wave:
Here is a graph illustrating the R-nought during the first wave. The R-nought for the 5 countries is in between 1 to 1.12.
Which Country Was More Successful in Flattening the Curve?
Actual VS. Predicted
The dotted lines represent the actual number of people infected and the light blue line represents the predicted infection curve. We can see that the actual numbers for all countries are lower than the predicted infection numbers.
Now let's take a look at the comparison between real and worst-case (WC) curves in logarithm. Infection curves are lower than the WC prediction.
Measuring the Flattening Effectiveness
To measure the curve-flattening effectiveness, we have computed serval measures in the table.
The peak reduction ratio (PRR) measures the difference in the peaks of the actual and predicted graph. The ratio goes from 0 to 1, the closest to 0, the least effective the efforts were. The closest to 1, the more effective the effect was. So all 5 countries were able to reduce by more than 90% the peak via interventions.
The peak delay days (PDD) measures by how much we delayed the peak. The larger PDD it is, the more we managed to delay the peak. Canada, Germany, and Israel were successful in delaying the peak. By doing that, we can have more time to improve treatments, find vaccines and increase hospital capacity.
The case reduction ratio (CRR) measures the number of cases that were reduced during all epidemics. If CRR is close to 1, we were able to successfully flattened the epidemic. If CRR is substantially smaller than 1, we still contained the epidemic. So for all the countries, the epidemic is still not at its end.
Overall, all the countries were able to reduce the spread rate using different measures, some were also able to delay the peak to gain more time. However, based on the CRR measures, Covid-19 is still not ending. Countries need to continuously monitor their COVID’s prevention measures.
Apart from controllable factors used in each country in order to reduce the spread of COVID, we will also be investigating any external factors that might be correlated with the infection rate.
What Are the External Factors Affecting COVID’s Numbers?
Weather and COVID
Several people have suggested that the COVID-19 will go away on its own in the warmer weather. Some have even suggested that the experience with SARS in 2003 provides evidence for this assertion, arguing that seasonal cyclicity is a ubiquitous feature of acute infectious diseases, which is also commonly observed in respiratory viral diseases.
In the following, we will investigate the effects of several meteorological factors on daily new cases and daily new deaths of COVID-19. We choose to focus on 4 main factors which are: temperature, wind, and relative humidity. A number of epidemiological studies demonstrated that various contagious respiratory illnesses transmission is strongly modulated by temperature and humidity.
The wind is a crucial factor in the transmission of respiratory infectious diseases and it may modulate the dynamics of various vectors and pathogens. Because of uncertainties surrounding how low the virus can travel and transmit through aerosols, looking at wind can provide interesting insights.
In order to overlay daily case data for each country with daily weather features, we chose the most populated city from each chosen country as a proxy for that country’s weather. We obtained weather data from VisualCrossing.com
We chose countries that had relatively differing levels of aggregated weather metrics to conduct our analysis. We can see that overall temperature, humidity, and wind speed differ in each country, with Australia and Finland representing windy, warmer climates, Canada and Israel representing cooler, less windy climates, and Germany standing in the middle.
Relationship Between Weather and COVID
Looking at the overall correlations between weather features and new daily cases and deaths, we can see that cases and deaths are positively correlated with temperature, with deaths having a stronger magnitude of correlation. Wind speed and relative humidity, on the other hand, had a negative correlation with cases and deaths, with wind speed having a stronger negative correlation and humidity’s correlations being insignificant. This is interesting, as it is potential evidence debunking the fact that the disease would “die out” with warmer weather.
Drilling down on Temperature, we can see from the plot that higher points of temperature were associated with higher daily case numbers and deaths.
Looking at individual observations, we also see that higher wind speeds are associated with a lower number of deaths and new daily cases.
Similarly, higher levels of humidity were associated with a higher number of deaths and daily new cases.
Compare Top Correlations Between Countries
We will now compare the most relevant correlations involving cases, deaths, humidity, temperature, and wind speed across each country. In the below bar charts, only the correlations that had a p-value of over 0.05 are observed for each country and are presented in descending order.
For Canada, we can see that Temperature and total are extremely highly correlated, followed by temperature and deaths. Interesting to note, is that the positive relationship between cases/deaths and temperature was actually greater in magnitude compared to wind speed and temperature and relative humidity and wind speed.
For Finland, the temperature had a slightly lesser correlation with total new cases and deaths relative to Canada. The relationship to wind speed and deaths/cases was also significant, unlike Canada.
Germany had similarly ranked correlations compared to Finland.
Australia displayed interesting results, as the temperature was negatively correlated with deaths and new cases, as well as wind speed and humidity. We have seen in previous plots that Australia seemed to combat the virus much more effectively compared to the four other countries seen. Therefore, these weather results are curious. However, we cannot make certain conclusions based on this, as we know that correlation is not synonymous with causation.
Lastly, in Israel, we see that the relationship between temperature and deaths/cases was less pronounced compared to other countries. However, that of relative humidity and deaths was much more pronounced.
In conclusion, although temperature might represent one of the factors that influence COVID-19 prevalence, there are other important factors that might have worsened the situation in countries that were heavily invaded by the pandemic. We must be wary of making conclusions solely based on correlation, as the meteorological variables alone cannot explain most of the variability in the disease transmission and its mortality. In future studies, it would be interesting to control for the general behavior of disease transmission. By accounting for confounding variables such as these, it might be easier to attribute case and death evolution with weather features.
Datasets and Code Repository:
https://github.com/BeiqiZh/Covid_First_Wave
References:
https://www.worldometers.info/world-population/population-by-country/
https://zpremma2453.medium.com/is-weather-a-good-marker-for-covid-19-b2092af8a423
https://www.visualcrossing.com/weather/weather-data-services#/editlocations
https://www.kaggle.com/vishalvjoseph/weather-dataset-for-covid19-predictions
https://en.wikipedia.org/wiki/List_of_countries_by_GDP_(nominal)_per_capita
https://www.worldometers.info/world-population/population-by-country/
