632,908 Cases in Greece 

COVID-19 Projections for Greece

Introduction | Videos | Analysis | Evaluation | Policies | Disclaimer | Contact

Introduction

In this section, we give a short summary about our projections and analysis for Greece.

NOTE: We have updated again our projection. New first projection day for Greece is September 15, 2020.

If there will be any more updates, we will inform you on this page. You can view projections of other countries here.

Videos

These videos are an animation of confirmed cases and deaths with embedded the measures taken from country authorities. We believe that with these timelapses it is more easy to understand the magnitude and the speed of spreading of COVID-19. These videos will not be updated daily, but whenever is necessary. To see the projections scroll up to the top of this page.

Analysis

Our analysis for COVID-19 progression in Greece is based on multiple parameters. Some of them are the Mobility and Activity Index (MAI), number of daily tests, positivity rate. These parameters are presented in the figures below.

Τhe Mobility and Activity Index (MAI) is based on the data provided from Google COVID-19 Community Mobility Reports. In this dataset Google provides mobility trends for Grocery & pharmacy(grocery markets, food warehouses, farmers markets, specialty food shops, drug stores, and pharmacies), Parks (local parks, national parks, public beaches, marinas, dog parks, plazas, and public gardens), Transit stations (public transport hubs such as subway, bus, and train stations), Retail & recreation (restaurants, cafes, shopping centers, theme parks, museums, libraries, and movie theaters), Residential (places of residence) and Workplaces (places of work).

Τhe Mobility and Activity Index (MAI) is created by applying Principal Component Analysis (PCA) in 5 out of 6 trends, we left out Parks as they highly collerated with seasonality. By definition the Mobility and Activity Index (MAI) is a weighted sum function of the mobility trends and shows the mobility and economic activity of each country.

For Greece the first principal component (PC1) of PCA can explain 80.25% of the total variance.

Mobility and Activity Index (MAI) = 0.4*T(Retail & Recreation) + 0.16*T(Grocery & Pharmacy) + 0.32*T(Transit) + 0.23*T(Workplaces) - 0.11*T(Residential)

The baseline is calculated by Google as the median value, for the corresponding day of the week, during the 5-week period Jan 3–Feb 6, 2020. After lockdown of March in Greece we see a dramatic drop of the MAI at a level of -75.5%. In May some measures were lifted and the MAI gained 25%. During June and July tourism was open and by the end of July Greece gained all economic activity. In August, MAI was above baseline and touched again the baselined at 15 of August. In summary, Mobility and Activity Index for Greece gained 75% from the support level and has a 0% percent to achieve pro-coronavirus levels.

Another figure that can explain qualitative the COVID-19 curves by its quantity over time is the daily tests performed. Data for testing are available from Our World in Data via Github here.

As it can be observed the number of tests followed the increase of cases. In August testing capacity is plateauing at a level of 10000 - 16000 daily tests.

We combined the data of daily tests with the daily confirmed cases to produce a figure of positivity rate. This rate is a good indicator about our control over the COVID-19 pandemic. A positivity rate under 1% is considered safe zone, a positivity rate above 20% is considered hot zone. WHO threshold for re-opening is 5% for 14 consecutive days.

We observe that from May and on the positivity rate drop nearly to zero levels. This was a good indicator that test were enough and the prevalence of the coronavirus slow. From mid June and after we see a steady increase of positivity rate from 1% to 2% in late August. Now is around 3%. This indicates that COVID-19 in Greece is under control for now, but if we see a sharp increase in this indicator reaching a level of 5% then new measures must be imposed and the capacity of testing must be increased.

The aforementioned information were derived from the data and from this point on we analyze our assumptions, choices and conclusions based on our simulation model. The results of which are diplayed as projections in the first figures of this webpage.

The basic assumption is that the reported data are not showing the reality of COVID-19 spread. This is based on the fact that there is no efficient testing on the whole population. For our analysis, only daily deaths can give a good estimation of the disease progression. This is the reason why we do not calibrate the model to the data, as we want the model to capture what happens in reality, without overfitting.

The initial population of Greece that has been incorporated into the model is 10,423,056.

The first case reported on 27/02/2020 and the first death reported two days later on 11/03/2020.

With this in mind, we assume that around 65 exposed persons entered Greece between 05/02/2020 and 16/02/2020.

In our projection model season has an effect of 47%.

The reaction time of people to the measures was 5 days, on average.

The confirmed cases now are around 20% of the real number.

We estimate that until 27/09/2020 a 1.02% percent of the total population of Greece has been infected and 0.76% percent of the total population of Greece has been recovered.

Overall, the situation in Greece is not under control, after the loosening of measures,opening the tourism and now schools it has become very fragile. Our projection shows a scenario where the measures that imposed in August made people to follow the rules for personal protection measures at a level of 70%, then with the schools opening this reaches 65% and a general lockdown at around 10th of November will bring this number to 85%.

More details about the model can be found here.

Please visit the official website of Greece National Public Health Organization for official information.

Evaluation of the projections models

We use as evaluation metrics the mean absolute percentage error (MAPE) and median absolute percentage error (MdAPE) on the cumulative cases and deaths, respectively. We made this decision in order to allow comparability of our models with other forecasting models for COVID-19. See details of performance of other prediction models here.

The overall performance of our projections is obtained by a weighting average of the metrics of each model. As weights the normalized to one number of days is used.

The partial and overall Absolute Percentage Errors (APEs) of our projections of COVID-19 for Greece with the confirmed cumulative cases it shown in the next table. Lower value is better performance.

The partial and overall Absolute Percentage Errors (APEs) of our projections of COVID-19 for Greece with the confirmed cumulative deaths it shown in the next table. Lower value is better performance.

Policies

The measures taken by Greece authorities by date (Day/Month/Year) are:

Level is in ordinal scale and higher value means more strict measures. Detailed explanation about the table can be found here.

Disclaimer

This analysis contains Work in Progress (WIP) results. No rights can be derived from any of the results.

The projection that is visualized is only one simulation run and is not forecast neither future prediction.

We are not directly or indirectly owned or managed by any governmental organization, public charity, non-profit, or private foundation. Covid-Analysis is not affiliated, associated with, or in any way has any official status with the United Nations or any other official organization.

The project is self-funded. Your contribution is important for the operating expenses of this site and for coffee, our basic fuel. Every donation will help us continue this effort.

Contact

If you find this analysis useful please do not hesitate to contact us at: info@covid-analysis.com