Frequently Asked Questions

  1. How reliable are the data?
  2. How have you accounted for excess deaths not-related to COVID-19?
  3. Why are the estimated infections so much higher than the reported cases?
  4. Why are the reported cases increasing in my country but the estimated infections are falling?
  5. What have you assumed about comorbidities?
  6. Why are most of the model parameters based on UK data?
  7. Can heterogeneity reduce the spread of the infection and so result in lower final attack rates?
  8. Do your projections change if you modelled sub-nationally?
  9. How should a national response respond in light of these findings and in the context of the secondary impact of COVID-19 interventions on other health systems?
  10. The website and projections look different to when I last viewed the website?

1. How reliable are the data?

We use data on excess mortality to calibrate our model, this allows us to account for under ascertainment of COVID-19 related deaths. This data taken is taken from The Economists Excess Deaths Tracker and World Mortality Dataset. Where this data is unavailable we use data from the The Economists Excess-Deaths Model, which uses machine learning methods to predict excess-mortality. Though these methods and the sparsity of mortality data outside of HIC & UMICs the estimates are comparable to estimates made by the WHO.

We also present and contrast reported COVID-19 deaths and cases. These are taken from the Johns Hopkins Coronavirus Resource Centre.


3. Why are the estimated infections so much higher than the reported cases?

To estimate the total number of infections, we use detailed information collated from ongoing epidemics to “back-calculate” from the reported deaths. This includes information on the proportion of infections that require hospitalisation as well as distributions for the time from infection to onset of symptoms (i.e. the incubation period), from onset of symptoms to hospitalisation, and from hospitalisation to outcome.

A significant proportion of infections are likely to be either completely asymptomatic, or sufficiently mild as to not seek care. From our early analysis of data from China, we estimate this to be approximately 40-50% of all infections.

Of those infections that are symptomatic and would seek care, the surveillance underway in each country is likely to pick up a fraction of these. In countries that are testing widely in the community, we would expect this fraction to be much higher than in countries that are focusing testing in hospitals. The difference in the estimated infections and reported cases should be interpreted with this in mind; however, the scale gives a sense of the likely potential for ongoing transmission in the community.


4. Why are the reported cases increasing in my country but the estimated infections are falling?

We fit our transmission model to the reported deaths in a country, as we believe this is the most accurate indication of the progression of the epidemic, with the under reporting of cases likely to be far greater than the under reporting of deaths. The level of under reporting, however, is likely to change over time. Testing capacity will be very limited at the beginning of the epidemic with only the most severe infections and deaths likely to be detected. As the epidemic progresses, testing capacity is likely to increase. Consequently, most countries will observe an increase in reported cases over time that may not be in proportion to the reported deaths over time. In these settings, where daily reported cases are increasing but daily reported deaths remains constant, we predict the difference to be due to the increasing testing capacity and may not be indicative of an increasing epidemic.


5. What have you assumed about comorbidities?

Our current model is based on estimates of severity based on data from China and Europe. These estimates of the infection fatality ratio are then incorporated into the model and modified as a result of the different demography, social contact patterns and health systems in each country. As a result, the “average” IFR can range from 0.5% through to 1.5% depending on these factors.

At the current time we do not explicitly take into account the role of comorbidities in each country, although they are implicit in the underlying IFR patterns obtained from China and Europe. We are actively working to improve inputs for currently identified co-morbidities.


6. Why are most of the model parameters based on UK data?

The model parameters were initially based on an early analysis of data from China and have more recently been updated based on data from the epidemic in the UK. The reason for using these parameters is that we have extensively analysed data in the UK and believe it represents the best current estimate of the key epidemiological parameters driving the transmission and severity of the disease.

As data emerge from other countries, we will continue to review these parameters and move to geography-specific parameters if differences emerge.


7. Can heterogeneity reduce the spread of the infection and so result in lower final attack rates?

Heterogeneity is known to slow the spread of respiratory pathogens, thereby resulting in a “less-peaked” epidemic. Whilst in some cases it is possible that this heterogeneity could result in lower overall attack rates, this is not often observed for highly transmissible respiratory pathogens.


8. Do your projections change if you modelled sub-nationally?

There are many aspects of populations that vary at a smaller spatial scale and could therefore differ between sub-national units. Our outputs are therefore representative of “average” patterns for a given country. Whilst sub-national estimates can help to determine local responses, we note that for the most part respiratory pathogens tend to spread quite rapidly across populations and so differences between sub-national areas are likely to primarily relate to the timescale of the epidemic rather than the overall attack rate. Differences in underlying demography and risk factors could however give rise to different overall mortality rates at the sub-national level.


9. How should a national response respond in light of these findings and in the context of the secondary impact of COVID-19 interventions on other health systems?

Our outputs are provided to help support decision makers to interpret their current epidemiological situation and to provide short-term projections of the likely trajectory of the COVID-19 epidemic. They are, however, only one input to be considered prior to deciding on the most appropriate actions to take in response to this epidemic. It is important to note that our model (for many countries) is reliant on estimates of excess-mortality and we do not estimate any indirect mortality that could occur as a result of either limited healthcare capacity or the wider response taken. Equally, it is likely that any actions taken to mitigate the spread of SARS-CoV-2 virus could impact in other ways that could impact both the health and livelihoods of populations. All of these factors should be considered in tandem to decide the appropriate national response.


10. The website and projections look different to when I last viewed the website?

We are actively involved in updating our methods and the scenarios that we are modelling. Consequently, the website will change as we try to provide a more useful tool for the global community, which more accurately captures the dynamics of the COVID-19 pandemic. To see how the website has changed over time, please view the News page for more information.