Figure 1. Causal Bayesian network model for learning population COVID-19 infected and death rates.

University of Michigan - Ann Arbor


- The COVID-19 testings tend to focus on already hospitalized people, not accounting for the people with mild symptoms and the asymptomatics leading to selection bias.  
- The differences in death rates across countries may not be attributed to clinical, demographic, and environmental factors.  
- Together with random sampling/testing, the Bayesian Network model allows causal explanations from data for better predictions.  


Summary of "COVID-19 infection and death rates: the need to incorporate causal explanations for the data and avoid bias in testing"

- The model (Figure 1) demonstrates that the COVID-19 death rate is not only affected by the observed variable (confirmed cases), but also influenced by other factors such as sampling methods, testing accuracies, and the definition of COVID-19 deaths.  
- In the causal model, dependencies are expressed in arrows.
- (e.g.) 'population proportion with COVID-19 who die' is affected by both 'population demographics and environmental factors' and 'Quality of healthcare.' 
- (e.g.) Similarly, 'Number of reported COVID-19 positive reported as dead'  is dependent on (affected by) 'Population proportion with COVID-19 who die', 'Death Reporting Policy,' 'Number tested positive with symptoms who do not have COVID-19', 'Number tested positive with symptoms who have COVID-19,' and the 'Number tested without symptoms who test positive.'


How Causal model (Bayesian Network) can better explain CFR

- The differences in death rates across countries may be attributed to clinical, demographic, and environmental factors.  
- While the above is true, the differences may also be to do with the sampling methods used, death reporting policies (deaths "because of" COVID-19 vs. deaths caused by something else while also having COVID-19), etc.  
- Thus, the differences in death rates across countries may not be necessarily much to do with how well/poorly one is managing the virus.  

Potential causes of the differences in death rates

Norman E. Fenton, Martin Neil, Magda Osman & Scott McLachlan
Journal of Risk Research, DOI: 10.1080/13669877.2020.1756381
Received: April 7, 2020
Accepted April 10, 2020
Published online: April 24, 2020
https://www.tandfonline.com/doi/full/10.1080/13669877.2020.1756381


COVID-19 infection and death rates: the need to incorporate causal explanations for the data and avoid bias in testing

Misleading death rate calculation: At the time of the writing, the UK's confirmed COVID-19 cases were 33,722, and of which, 2,921 people died, which resulted in the case fatality rate (the number of reported deaths per number of the reported case) of about 9%.  

The problem of selection bias: Most of the tests were performed against hospitalized individuals. Thus, the above death rate does not account for people with COVID-19 who were not hospitalized and did not die.  If they have mild symptoms or asymptomatic, they may not visit hospitals in the first place.  

Learn Before

Related