ADS & AMDS Webinar | Different Scenarios for Exit Strategies in COVID-19

To kick off the first in the Amsterdam Data Science & Amsterdam Medical Data Science webinar series on Data & COVID-19, Edwin van den Heuvel, professor in statistics TU Eindhoven and Bert Slagter, founder Procurios, talked about the predictions and changes in spread of the virus and the possible scenarios for exit strategies.

 

Check out the recording of this webinar.

Edwin van den Heuvel, professor of statistics at TU Eindhoven talked about predictions and changes in the spread of COVID-19 in several countries.

Key Takeaways

Flattening of the curve. Using two different statistics, a turning point (flattening of the curve) was observed on 31st March in the Netherlands. However, the flattening of the curve didn’t behave as expected compared to China. Using a generalized logistic curve (specifically here: Poisson) did improve the predictions.

Contact-Rates. After estimating this model, profiles are given to countries to determine data-driven change points in daily contact-rates. These change points can then be compared to governmental measures (e.g. lockdown or banning events), which should then be parallel. This will then help to analyze to see what governmental restrictions are (most) effective. Three interesting conclusions from analysing this is that:

  1. Closing schools and banning events seem to have a direct effect;
  2. A lockdown is not necessarily effective without police enforcement;
  3. Closing restaurants did not show a clear impact.

Bert Slagter, founder of Procurios discussed the different scenarios for Exit Strategies in COVID-19

Key Takeaways

Data and ICU Capacity. With the aim of helping healthcare professionals have more insight on the required capacities, a very simple model with clever extrapolation of the data series was used. This was due to the lack of (reliable) data. However, the fit of the model did quite well and the values from academic papers were used to cross check. Finally, the data on ICU-admissions was used as reference data.

A Model to Explain. The model estimated the required ICU-capacity to be 1400-2000 beds around April 4. It turned out to be 1313 patients on April 7.The goal was not to make a perfect model or to outsmart the government institutions. The goal was to inform the general public, to educate and help people understand what was happening:
A virus spread is exponential in nature.

  • Effects of exponential growth (very different from linear growth);
  • Presence of 2-3 weeks lag between interventions and effects;
  • Lockdown necessary to avoid ICU overcrowding.

Reopening the Economy. When plans were made by the government about reopening the economy, the model insights counter argued. The Dutch approach is to mistake absence of evidence with evidence of absence. The approach should be to keep the reproduction number at one. However, this is also dependent on the behaviour and environment of humans. Therefore, it would require a balancing act of increasing and decreasing measures. The policy should be to decrease the daily number of infections to 10-100’s instead of 1000’s.

Check out the Amsterdam Economic Board’s article on the webinar.

Read More