Which early indicator allows for a better understanding of the evolution of the COVID-19 epidemic in France?

1 Introduction

The COVID-19 epidemic started in China in late 2019 in Hubei Province, then it spread around the world and became a pandemic [1]. A first wave reached France in late winter and continued in early spring 2020. After a lull at the end of spring following lockdown, since the end of July the epidemic has been restarting to form a second wave. This second wave is less strong than the first, in the sense that the evolution of the number of cases is less rapid. But we do not yet know the consequences in the medium term. In order to analyse this new phase of the epidemic we have data provided by Sante Publique France [2]. The weekly bulletin of Sante Publique France provides us with the the following indicators: new hospital admissions, new intensive care admissions and deaths.

From the beginning of the second wave, which can be estimated from the end of July to the beginning of September, these quantities provided us with relevant and consistent indicators to monitor the evolution of the epidemic: we could deduce a doubling of the quantities considered every 21 days. However, for the past month, there has been a discrepancy between the evolution of the rate of positive tests and the numbers of hospital admissions and deaths: while these numbers double every 16 days, in recent weeks the rate of positive tests has hardly changed at all. We will therefore try to explain this discrepancy. One possible explanation is the time lag between the infection dates of asymptomatic young people and the hospital admission dates of older people who develop severe forms, which might suggest that the epidemic is coming to an end. We will show that this is not the case and that the positive tests rate is no longer a relevant indicator and will seek to propose a new indicator.

In section 2, based on the available indicators, we will put in the discrepancy between the positive tests rate evolution and the number of new hospital admissions. In section 3, we will use additional data based on age structure and positive tests and will propose a new indicator. To conclude, we will summarise the information we have been able to extract from the data.

2 Analysis of available indicators

Every week, Sante Publique France publishes a newsletter providing, among other things, the following indicators: positive tests rate, numbers of new hospital admissions, new intensive care admissions and deaths. We are interested in the second wave of the epidemic that began around end of July. In Table 1 we reproduce the data provided by Sante Publique France.

The positive tests rate has been increasing exponentially from the end of July to the beginning of September, but in the last few weeks, this rate growth has slowed. This could indicate a slowing down of the epidemic. To find out wheher it is the case or not and to better understand the rate evolution, we make a preliminary evaluation of the local k parameter of the exponential describing the evolution of the indicators of interest and the corresponding doubling times. This will allow us to detect break points in the epidemic process. We give in Table 2 the growth rates over 3 weeks: for the positive tests rate and the numbers of hospital admissions, intensive care admissions or deaths, we give an estimate of the associated doubling time:

Reading the results confirms for the positive tests rate, a first phase between the end of July and the end of August. There is indeed an exponential growth with a doubling time of approximately 20 days followed by a period of stagnation. Furthermore, we note that :

• - while the rate had an exponential growth the numbers of new admissions and deaths increased only marginally. This very weak evolution made some observers say that there was no cause for concern and that the increase in the rate was anecdotal because it only affected asymptomatic young people. This is not the case because it means forgetting the existence of time lags, which we is described in the following.

• - while in recent weeks the positive tests rate has stagnated, the number of new admissions follows an exponential trajectory with a doubling time significantly lower than the rates in the previous phase: 15 days instead of 20 days. This is evidence that the epidemic is far from over.

To confirm this discrepancy between the past doubling time of the positive test rate and the current doubling time of new admissions we operate regressions of the log of positive tests rate, number of hospital admissions, intensive care admissions and deaths. We obtain the following results (Table 3) for the estimates of the k parameter of the exponential with the associated doubling time:

The doubling time for the positive tests rate is 21 days while the doubling time for new hospital admissions and intensive care admsissions is 16 days. The result concerning the growth rate for the number of deaths is not very significant, it will be necessary to wait another week to confirm the value obtained for the doubling time.

But then how can we explain the discrepancy in behaviour between the evolution of the positive tests rate and the numbers of new admissions or deaths? To answer this question, we consider the evolution of the positive tests rate by age group.

3 Analysis of the age structure of positive tests rates

Table 4 shows for each week of the positive tests rate for different age groups [3]. Figure 2 gives the contour plot of the positive tests rate.

From the Table 4 or the associated Figure 1, we notice that travelling waves start from the age group 20-29 years old: for example the rate around 2% of the age group 20-29 years old in week 30 spreads to the age group 30-39 years old in week 31, … until the age group 70-79 years old in week 35. This phenomena is observed in the following weeks for higher rates. We deduce that the epidemic move from the age group 20-29 years old to the other class at a speed of 10 years per week until the age group 70-79 years old. Hence we have a first result:

• Download figure
• Open in new tab

Figure 1: Contour plot of Positive tests rate for the different age groups

• Download figure
• Open in new tab

Figure 2: Overall positive tests rate, Positive tests rate for the 70+ age group, Number of hospital admissions

“The latency time between the infections of young individual (generally asymptomatic) to older individuals that may have a severe form of the disease and may be hospitalized is close to 5 weeks.”

From :

• - the vast majority of those who test positive are young people who will not develop serious forms of the disease.

• - the test detects very recent infections

• - hospitalisations mainly concern individuals of more than 70 years old

it can be deduced that the positive tests rate at a given date will only lead to an increase in the number of new admissions 5 weeks later. This mini-conclusion is consistent with the facts we observed earlier: the exponential phase of the positive tests rate starts in week 29 and that of new admissions only in week 33 or 34.

We still have to explain the second phase where there is a discrepancy: the positive detections rate stagnates while the numbers of new hospital admissions is growing exponentially. The answer lies in the process of the spread of the virus in the age groups. Indeed, the overall positive tests rate is an aggregation of positive test rates by age group and the structure of these rates in the age groups changes significantly over time. At the beginning of the second wave, it is mostly young people who have been contaminated and the more the weeks go by, the more older people are becoming infected. Since the end of August, the positive tests rate among older people have been increasing much faster than among young people: the positive tests rate in the 70-79 (resp. 89-89) age group was multiplied by 3 (resp. 3.4) between weeks 34 and 38, whereas it was only multiplied by 1.55 for the 20-29 age group. So even though the overall rate does not increase much anymore, the number of individuals who will develop severe forms is increasing very significantly. Thus if the overall positive tests rate provided relevant information until the end of August, since the beginning of September the overall positive tests rate has reached a plateau and no longer provides relevant information on the current state of the epidemic. In order to know where we are in the evolution of the epidemic, it is therefore necessary to observe the rates of positive tests among the population at risk of developing a serious form, namely the age group of 70 years and over. To this end, we have calculated an average positive tests rate in this population, taking into account the numbers in the three age groups: 70-79, 80-89, 90 and over. We obtain the following evolution of the rates for the 70+ age group:

As before, we calculate local doubling times: :

We detect two phases, the first with a doubling time of around 25 days and the second with a doubling time of around 16 days. Once again we perform linear regressions and we get:

Figure 2 shows the evolution of the overall positive tests rate (with a fitted logistic curve), the positive tests rate for the 70+ age group and the number of hospital admissions. The difference in behaviour between the evolution of the overall positive tests rate and the rate for 70+ age group is obvious. Moreover in the last weeks, the evolution of the number of hospital admissions is quite similar to the evolution of the positive tests rate for the 70+ age group.

Results of Table 7 thus confirms on the one hand that the doubling time is currently 16 days for positive test rates of the age group 70 and over, new admissions to the hospital or in intensive care and deaths.

On the other hand, it confirms that the positive tests rate in the age group 70 and over is currently an early indicator of the epidemic. These results will have to be confirmed by the data of the following week.

4 Conclusion

We notice a discrepancy between the dynamic of the positive tests rate and the dynamics of the numbers of hospital admissions, intensive care admissions and deaths. These three quantities double every 15 days. The positive tests rate appears to have slowed grown over the past two weeks. We show that this apparent discrepancy is explained by the changing age structure of the population of individuals who test positive. We highlight the existence of a latency time of about 5 weeks between the positive detection rate and the number of new hospital admissions in the phase of the exponential dynamics of the positive tests rate. Thus if the overall positive tests rate provided relevant information until the end of August, since the beginning of September the evolution of the overall positive tests rate has reached a plateau and no longer provides relevant information on the current state of the epidemic. However, the positive tests rate in the 70+ years old group is a relevant and early indicator of the epidemic. Furthermore, we have shown an identical doubling time of around 16 days for the following indicators: the positive tests rate for the 70+ age group, the number of new admissions to hospital, intensive care unit admissions and deaths.