Scientists in China and in the US wanted to “go over the limitations of epidemiological models”, and used a method resorting to artificial intelligence to project the evolution of covid-19 in the world.
With an “active public health intervention” on the ground as of March 17, and with 245 confirmed cases last Monday, Portugal could reach a peak epidemic already by next Monday, March 23, with a total of 1026 confirmed cases. With these calculations, outbreak control could be a reality already by May 5, and after 65 days of epidemic we’d have a total of 2655 cases of covid-19 in the country. These are the results of an exercise carried out by an American scientist, specifically adapted to Portugal. The team of specialists from China and the US presented a series of calculations and scenarios in a pre-publication of an article published last Monday. However, the study quickly got outdated and was surpassed by reality. Portugal was mentioned, along with 29 other countries, but the only presented scenario was for a public health intervention starting on March 9, (which obviously didn’t happen).
In response to a request from Público, Momiao Xiong, researcher at the Biostatistics Department of the Texas School of Public Health in Houston, updated the data. The new data takes Monday’s total of confirmed cases, (245) as a starting point, and it assumes a public health intervention starting from March 17. (As we know, it came into effect on March 16 when all schools closed).
The information sent to Público by Momiao Xiong says the peak would be on March 23, with 205 new confirmed cases to be registered on that day. The outbreak control would be at the start of May, at which time we’d have a total of 2655 covid-19 cases. But how did the researcher get to this?
“Assessing the effect of the public health intervention in covid-19’s global dissemination trajectory”, that’s the title of an article published on Monday on the MedRxiv platform, which gathers a series of pre-publications on the pandemic that’s haunting the world. The authors from Shanghai University and from Houston, Texas present an analysis that aims to “overcome the limitations of epidemiological models”, and sits on a method that resorts to artificial intelligence to project the evolution of covid-19 in the world. There are various scenarios conditioned to a more or less aggressive intervention in many countries.
Eliminating 94% of the cases
Scientists present estimates for two scenarios: one reflecting the effects of a limited public health intervention and another showing the result we’d have if an “active intervention”, (restricted social interaction and quarantine) had started already on March 9. Today, to get the math straight, we can’t just push calculations a few days ahead. These complex models have to take into account a series of variables and parameters that, among other things, are influenced by forecasts of contagion by an infected individual.
As explained, the method developed is “a method of modified auto-codifiers (MAE) to forecast the trajectory of covid-19 dissemination in affected countries, under different intervention strategy deadlines”. The MAE models “allow to insert information on the interventions, to investigate the impact of the intervention on the size, duration and time of the virus outbreak, and to recommend specific timings for interventions”.
The approach wants to answer two important questions. “The first one is whether there needs to be a wide-reaching public health intervention without resorting to medication. The second question is on the importance of the timing of an intervention”, they explain. And they answer: “Our analysis shows public health interventions should be carried out as early as possible”.
The authors of the article conclude that on a global scale, a delay of 4 weeks, (after March 8) in the adoption of protection measures might lead to an increase in the cumulative total number of deaths, from 7174 to 133.608, (using a average of 3.4% of the mortality rate), and it would mean the outbreak would only be controlled by August 22, and not by June 25.
“After an active intervention, the final number of predicted cases in the world, (once the dissemination of covid-19 is through) would be reduced from 3929641 cases to 211 thousand, which is to say, 94,6% of the potential cases would be eliminated and the duration would go from 215 days to 157 days, with the final date of August 22 shifting to June 25”, the authors write.
Enough time has passed between the execution of this model and the publication of the article for the data to become outdated. But the main idea is this: it's extremely important to flatten the curve with an active public health intervention.
To get an idea of the impact an active public health intervention can have, Momiao Xiong shared with Público other much less optimistic scenarios for Portugal. So, if the public health action, (quarantine) only started on March 23, we’d have a peak on March 30 and a total of 14,825 cases, with a likely end on June 2. Lastly, if the intervention started only on March 30, the peak would be on April 3 and the total number of cases would reach 59,674 by June 22.