In a week, the situation in hospitals in some provinces could…

During the annual symposium of the University Laboratory of the ULB, devoted this year to COVID-19, epidemiologist Marius Gilbert carried out an intervention on the analysis and communication of epidemiological data, for political decision-making purposes. A reflection, as a scientist, on what the scientific community knows, understands, but which is not easy to understand for the citizen, the journalist, the politician, in times of global epidemic.

Among the points put forward, that of understanding the evolution of the epidemic in the near future, based on extrapolation from current figures, and therefore estimates of future situations based on the current situation. And the ULB epidemiologist’s estimates are quite alarming regarding hospitalizations: within 14 days, the hospital incidence in Belgium would exceed that reached during the first wave. For some provinces, such as Hainaut and Liège, this would take barely 7 days: the hospital incidence per province would exceed the national hospital incidence of the peak of the first wave.

Even if the hospitals are better prepared than in the first wave, this situation would still put the care services in great difficulty.

►►► Coronavirus in Belgium: “These measures are the result of our collective failure”, believes Marius Gilbert

Before analyzing these estimates a little more, it is important to introduce scientific concepts to fully understand the evolution of an epidemic, which is developed in the following paragraph. If you want to see the figures and conclusions immediately, follow this link.

Speed ​​and acceleration of an epidemic

Marius Gilbert recalled an important concept in epidemiology: the difference between what he calls the speed of an epidemic, and its acceleration.

• The speed of an epidemic is the number of cases (or deaths, hospitalizations, etc.) per unit of time, in the case of covid, the number of contaminations / hospitalizations / deaths per day.
• Acceleration is the growth in the number of cases per day, therefore the number of cases that we have in more, or less, compared to the previous day. On a graph showing the number of cases per unit of time, the acceleration is quite simply the slope that the curve will take: the steeper it is, the stronger the acceleration (or deceleration).

All the non-medical measures taken to fight against the coronavirus, such as wearing a mask, social distancing, the closure of certain sectors only act directly on the acceleration of the epidemic, that is to say its increase day by day. . With this consequence: regardless of the speed of the epidemic, that is to say the number of cases observed each day, the measures to be put in place to manage its acceleration are the same.

Concretely, this means, that there are 30 cases or 1000 cases daily, as long as there are no vaccines and effective treatments against covid-19, the preventive measures must always be present to prevent the disease. ‘epidemic to resume, therefore the acceleration to increase. Maintaining a stable or decreasing speed (therefore almost zero or negative acceleration) means continuing to apply preventive measures. We could take the metaphor of a campfire in a forest: whether it is in the state of glowing embers, or with very bright flames, it is always necessary to be vigilant to prevent the fire from spreading, as long as the ‘we have no water to extinguish it completely.

The daily ratio as an indicator of acceleration

To assess this acceleration, several indicators are possible. The best known is the reproduction rate Rt, which is the number of people that a positive person will infect. While the concept is easy to understand, it is very difficult to calculate, and depends on the model and the parameters used. For example, Sciensano uses hospitalizations to calculate the national Rt, but for provinces, calculates an Rt based on contamination. It is therefore an indicator that is difficult to compare.

►►► The real-time evolution of the coronavirus epidemic in Belgium in figures and graphics

Another indicator that is easier to understand and calculate, points out Marius Gilbert, is the daily ratio. This is a number, which multiplied by the number of daily cases / hospitalizations / deaths, will make it possible to estimate the figure for the next day, and the days after. The epidemiologist estimates this multiplication factor from the observed acceleration during the last 14 days (by doing a linear regression from the data in logarithmic form). And given the current figures, this method makes it possible to estimate when we will arrive at a situation similar to the first wave (with some uncertainty, since the acceleration can change in the meantime).

Regarding contaminations, a comparison with the first wave is impossible, given the differences between the two situations: the testing and tracing capacities are much greater, cases with little symptomatic or asymptomatic can be tested (unlike in the spring when only cases the most severe were counted), and as a result of improved testing capacities, the positivity rate is also very different.

In order to be able to compare more effectively with the wave of March / April, we should therefore rather look at hospitalizations. Marius Gilbert thus estimated the evolution of the hospital incidence, therefore the number of hospital admissions per 100,000 inhabitants, for Belgium and its provinces, and compared them to the national hospital incidence of the peak of March / April, in 7 and 14 days. Obviously, the further into the future we go, the more uncertain these estimates are.

If the acceleration, therefore the daily ratio, remains the same as at present, in two weeks, counting from October 16 (date on which the daily ratio was calculated), the hospital incidence of the country would reach 50 per 100,000 inhabitants, therefore 5,500 hospitalizations for the 11 million inhabitants of Belgium. We would arrive at a rate of 785 new hospitalizations per day. That is to say well beyond the figures of the first wave.

For the provinces of Hainaut and Liège, this symbolic threshold would even be exceeded before 7 days.

The provinces of Namur, Luxembourg, and West Flanders would probably also pass this threshold in 7 days, with some uncertainty. And there would only be Flemish Brabant, the province of Antwerp, and Limburg that would still be below the threshold of the first wave after 14 days.

How to understand these graphics?

• Gray bars on the left are weekly hospital incidence, as of October 15, 2020 for 14-day estimates, and October 16 for 7-day estimates.
• The black squares are the estimate of the situation at 7 and 14 days, according to the graph. The horizontal bars around this black square represent the variance around this estimate, or the uncertainty around the estimate. The larger the horizontal line, the greater the uncertainty.
• The vertical dotted line is the national hospital incidence at the peak of the first wave.