Globally we’ve all had to undergo the Covid lockdown in one way or another, yet there has been a significant difference in how each country has faced the difficulty of slowing the spread of Covid.
It’s a tough balance to strike between creating an enforced, effective social distance measure to flatten the curve, whilst trying to reduce the damage of a country’s economy, job losses and other effects such as mental health.
Flattening The Curve
In the below visuals we can observe different countries weekly new cases over time, adding in a countries lockdown date, highlighting the ‘main’ lockdown period in red light. We then take the number of weekly cases at the fist lockdown date to give us the baseline for each country. This is then compared to the number of weekly cases 21 days later (decided by the ‘days after lockdown introduction’ parameter).
We can take the ratio cases at the introduction of lockdown vs maximum number of cases before 21 days later. This gives us a metric for how much the virus has spread in each country since lockdown was introduced, which we can use to compare different countries ‘lockdown effectiveness’.
Lastly, we can compare this to unhindered, exponential growth. This is done by taking the number of weakly cases and doubling it every 7 days (set by the ‘expected double period (days)’ parameter).
This Allows for a rough estimate of how effective a countries lockdown restrictions were at limiting the spread of coronavirus, as in theory, after 14 or so days, countries should begin to see the curve being flattened. We can also see the period in which measures start to become effective by varying the ‘Days After Lockdown Introduction’ parameter.
As we know, lockdown restrictions seem to have an impact on limiting the spread of the virus – infection rates trail off roughly two weeks after the lockdown is introduced.
However when comparing the lockdown effectiveness between countries, there does not seem to be an obvious relation between a countries initial lockdown effectiveness vs their resulting a total number of cases or total number of deaths.
We can compare this with a countries lockdown ‘lateness’. This is measured by how late they implement their national lockdowns (dotted reference line), when compared to the day they reach 50 weekly cases per 10,000 (solid reference line)
The graph below displays scatter plots with best fit lines for the Ration of Cases at the beginning of lockdown vs 21 days later, the lockdown ‘lateness’ compared with the date of 50 Weekly Cases/10,000, and the total cases and deaths per 10,000.
As presented in the correlations formulae, the only relationship which shows a p value of below 0.05 is the relation between lockdown ‘lateness’ and the total cases per capita. When the weekly cases/10,000 reference is at 50 cases, the p value is 0.04, with an R^2 of 0.36. This is a relatively strong univariate relationship, especially when comparing with other COVID analysis, which often struggle to find correlations between COVID data and single metrics.
The analysis implies that it is generally the data of action taken, rather than the lockdown effectiveness which has a stronger impact on cases per capita, but that neither has a large impact on deaths.
However the limitation of this analysis has been recognised. First of all, this is a small selection of Western European countries, and these trends may be different when viewed for different regions. The limited selection of countries was made as the lockdown dates has to be manually verified in the dataset provided.
The analysis also doesn’t factor its aspects such as the discrepancy in testing, the different ways a country records a ‘COVID death’ and other factors. However its interesting to observe that, for these selected nations, the timeliness of a lockdown seems to make more impact than the ‘flatness’ of the curve they produce, when comparing total number of cases (there was no correlation observed when comparing the ratio of cases and the lockdown lateness).
Working in partnership with NHS Digital and Tableau, over the last two months TrueCue have played a leading role in developing a new dashboard that allows the public to track the progression of Covid-19 throughout communities in England. Find out more here.
Andrew is an Analytics Consultant with a keen interest in machine learning. He uses his degree in Data Science to build advanced analytics solutions - using creative approaches to leverage data in order to solve problems and create business value.
He thoroughly enjoys working with others to unlock insights hidden in data and always finds it fascinating what truths can be revealed through the power of analytics.