To overview

New method can improve Covid-19 predictions worldwide

Photo: Peter Thisted Dinesen

Researchers around the world risk misjudging both the worst and the best scenarios concerning COVID-19 and other pandemics. A new study supported by the Carlsberg Foundation shows that when societies need to secure public health, it is precisely the extremes that are crucial to understand.

The study, which has just been published in Nature Physics under the title "Fixed-time descriptive statistics underestimate extremes of epidemic curve ensembles", describes how to understand best and worst-case scenarios in connection with pandemics.

”Keeping track of the worst-case scenario is exactly one of the most important elements when reacting to pandemics - regardless of whether it be in Denmark, the EU, the USA or the WHO. If you have an average estimate for the development of an epidemic, but without knowing for instance how bad it can get, then it is difficult to act politically”, states Professor Sune Lehman from the Department of Applied Mathematics and Computer Science at Denmark’s Technical University (DTU), who is one of the four authors of the article.

“With this study, Sune Lehman and his colleagues have presented a really exciting new idea on how to improve predictions of the development of pandemics. I am pleased that the exceptional Covid-19 funding donated by the Carlsberg Foundation in the spring has resulted in such excellent new knowledge that can be used directly in the battle against both the current corona pandemic and later pandemics. The times we live in show more than ever that our politicians need scientifically sound and well-documented methods and tools when we are facing major challenges such as the current corona pandemic”, states Professor Flemming Besenbacher, Chairman of the Carlsberg Foundation.

Epidemics are unpredictable

Together with researchers Jonas L. Juul, Kaare Græsbøll and Lasse Engbo Christiansen at DTU, Sune Lehmann acts as an advisor to the National Board of Health and Statens Serum Institut during the Corona crisis. Partly based on their own experiences as advisors, they have realised that the existing methods of projecting the development of for instance Covid-19 is difficult in terms of describing the extremes of the predictable development.

Disease outbreaks are basically random processes. Whether an infectious disease infects many people or just a few partly depends on chance. And it is precisely the unpredictability of epidemics which makes it so difficult to make the right decisions in society when the epidemic hits.

How many beds and ventilators will be required? And to what extent can this requirement be reduced by implementing severe restrictions?

No fixed reference points

However, the general unpredictability is just one of many problems in estimating the development of an epidemic.

"It is not just the unpredictability of epidemics that makes it difficult to project their course. It is also our ignorance of the properties of the disease and its prevalence in society at any given time. For example, typically no one knows exactly when an outbreak has started, how many people are infected on a given day, or where the infection is currently spreading. The only thing we know for sure is that when the health authorities discover something, it has been going on for a while”, says Sune Lehmann.

The conventional way of dealing with this lack of information is to build many scenarios almost everywhere in the world based on for instance different dark numbers, start times for outbreaks, etc. You then summarize by looking at each day separately and assessing the “middle” predictions as the most likely outcomes of the day. If most input parameters result in infection numbers of less than 4,000 on Christmas Eve, the assessment is that more than 4,000 new infected is unlikely.

Use whole charts instead of single days

The "day-to-day" way of summarizing the infection prediction is used worldwide, and although the link between the development of the epidemic and specific dates is also useful in some contexts, it systematically excludes data on how severe or mild the epidemic will be.

If all projections for instance predict that the epidemic will peak at 4,000 infected in one day, but none of the charts show it on the same day, then on a given day it will be an extreme and therefore not included in any estimate.

Therefore, the researchers propose to make the summary “chart-based”, so that one looks at an entire simulation instead of assessing which infection rates are probable or unlikely on the individual days. Thus, one must look at whether the whole simulated infection chart is probable or not and on that basis make a summary of the most probable charts for the infection prediction.

"By looking at whole charts instead of a few days, you will get a more realistic idea of ​​how bad the epidemic can become. It is particularly useful when trying to avoid the hospital system becoming overloaded,” says Sune Lehmann.

Read more about research at DTU



To overview