Population density: not a primary factor in Covid-19

From a demo­graph­ic stand­point, what can we say about the Covid-19 crisis?

In my book Serons-nous sub­mergés ?, which was pub­lished in Octo­ber 2020, I stud­ied the way the first wave of the pan­dem­ic unfold­ed, day after day, in four coun­tries – France, Switzer­land, Italy and Spain. I stud­ied them at the lev­el of regions, coun­ties and provinces. The results showed that geo­graph­ic – rather than social – fac­tors were key for dynam­ics of the epi­dem­ic. Cer­tain­ly, as many stud­ies have shown, immi­grants and low-income peo­ple have a high­er risk of dying from Covid-19 than the rest of the pop­u­la­tion. Not because of where they’re from or how much they earn, how­ev­er. Rather, it is because of their prox­im­i­ty to the virus. These pop­u­la­tions work in hos­pi­tals, at super­mar­ket check­outs, as deliv­ery per­sons or taxi dri­vers. That is why they are more impact­ed. The virus can­not dis­tin­guish between an immi­grant and a non-immi­grant, nor can it tell how much mon­ey you make. Rather, it goes straight to the near­est per­son, and we can learn a lot more about who the near­est per­son will be through geography.

We often think of Covid-19 as an “urban virus”. How­ev­er, your study found that pop­u­la­tion den­si­ty was not one of the key fac­tors in its spread. Why is that?

When you com­pare Coro­n­avirus deaths between 1^st March and 15^th May 2020 with the den­si­ty of French départe­ments, there are huge dif­fer­ences. Over that peri­od, the mor­tal­i­ty rate in Ter­ri­toire de Belfort (1.18 per 1000 res­i­dents) was 170 times high­er than that of Ariège (.007 per 1,000). The con­trast between a map of mor­tal­i­ty rate and a map of pop­u­la­tion den­si­ty shows no cor­re­la­tion (see below). 

Den­si­ty is not a fac­tor in the large-scale spread of the virus. What mat­ters is the loca­tion of clus­ters, which are ini­tial­ly linked to just one per­son. The more peo­ple are con­t­a­m­i­nat­ed before the infect­ed per­son realis­es, the hard­er it is to con­tain the epi­dem­ic. This is what hap­pened with the sec­ond wave, but the dif­fer­ences in mor­tal­i­ty remain con­sid­er­ably dif­fer­ent – some­times as much as ten­fold high­er. And we see that “patient zeros” appear in both the coun­try­side and the city includ­ing large towns such as Mul­house and Ajac­cio, but also small­er ones like Auray, Creil or even a vil­lage in Savoie, Les Contamines-Montjoie. 

Hence, while den­si­ty mat­ters as the epi­dem­ic spreads, the ini­tial impact is min­i­mal. The only thing that we can say is that patient zero is a trav­eller. So, they often appear near big inter­na­tion­al hubs (Gene­va, Milan, Rois­sy Air­port near Paris, New York, etc.). But they also trav­el out of the city, which is where the clus­ter devel­ops, i.e. Crépy-en-Val­ois, La Bastide-Mon­tjoie, Bergame. This is actu­al­ly what hap­pened at the start of the AIDS epidemic. 

So to con­trol the spread one must first con­trol people’s move­ments – hence the lock­down. After expo­nen­tial growth of cas­es in the first clus­ters (Mul­house, Auray, Milan, etc.), the epi­dem­ic in the first wave was con­tained. For exam­ple, it prac­ti­cal­ly didn’t get into the Loire at all, nor into Andalu­sia or South­ern Italy. It also explains why deaths were thir­ty times high­er in Milan than in Naples. It’s a reminder of how, dur­ing the last out­break of the bubon­ic plague in France (Mar­seille, 1721), lines of sol­diers were deployed to pre­vent the epi­dem­ic from spread­ing beyond the Provence region.

Let’s go back to social cri­te­ria. Which did you select for your study?

In each of the coun­tries includ­ed in our study, fig­ures for Covid-19 deaths were com­pared with four indi­ca­tors: den­si­ty, pover­ty, the pro­por­tion of immi­grants in the pop­u­la­tion, and the pro­por­tion of peo­ple over 70 years old. The geo­graph­ic dis­tri­b­u­tion of these four fac­tors gave us no indi­ca­tion of which seg­ment of the pop­u­la­tion would be impact­ed. The maps speak vol­umes. This is due to the fact that the first wave was con­tained in the four coun­tries studied.

What about the sec­ond wave?

Para­dox­i­cal­ly, the sec­ond wave orig­i­nat­ed at the end of the first lock­down. The num­ber of dai­ly cas­es was very low at the end of June. But peo­ple in the 15–49 age brack­et account­ed for two thirds of new cas­es. Many of them passed by unde­tect­ed, as they were asymp­to­matic. With peo­ple trav­el­ling for the sum­mer hol­i­days, the virus spread all over France. Old­er peo­ple, par­tic­u­lar­ly grand­par­ents, were also infected.

As such, in Octo­ber France found itself with a large num­ber of clus­ters. With the excep­tion of Mayenne, these could not be con­tained as seen in Ajac­cio, Auray and Les Con­t­a­mines-Mon­tjoie. As a result, the epi­dem­ic spread pret­ty much every­where, as it had done in the two big clus­ters in the first wave in Creil and Mulhouse.

Because of more wide­spread safe­ty mea­sures and improved health­care, the virus did not spread very quick­ly. The spread of Covid-19 across almost the entire coun­try has pro­duced new social dif­fer­ences con­nect­ed to what hap­pened dur­ing the first wave – cer­tain groups are more care­ful than oth­ers, as shown by debates about mask-wear­ing. Nev­er­the­less, region­al dif­fer­ences are not insignif­i­cant – between coastal Brit­tany and the Lyon region, the rate of cas­es and the mor­tal­i­ty rate varies by a fac­tor of ten.

Just as we saw in the first wave, the sec­ond lock­down (pre­vent­ing peo­ple from trav­el­ling) will prob­a­bly main­tain these dif­fer­ences, once the wave has been stopped. Like oth­er con­tact epi­demics through­out his­to­ry, con­trol­ling people’s mobil­i­ty remains an essen­tial fac­tor for con­trol­ling the epi­dem­ic. We had bet­ter remem­ber that if we want to pre­vent a third wave.

Interview by Clément Boulle