Fatality Rate Models

Fatality Rate Models

Introduction

Fatality rate models categorize people by age or health and then use data to estimate the number of fatalities that will result from the pandemic. These models rely on data and predictions from other models to make estimates of fatalites. Simple models, such as these, are far more accurate than individuals who make up numbers out of whole cloth.

If there were no differences in fatality rates across groups, the total number of fatalities could be estimated by multiplying the fatality rate by the size of the population.However, COVID-19 does not effect every age group equally. Older people are more likely to suffer fatalities. The chart shows fatality rates and percentage of individuals exposed. These initial fatality rates werebased on the figure to the right, which details data from China. As more data arrives, experts will be able to make more precise estimates of the total.

To estimate fatalities, the population within each category is multiply its fatality rate. For example, there are approximately 40 million Americans between the ages of 40 and 50. If all of them were to contract sympotamic COVID-19, the expected number of fatalities in this age group would be 40,000,000 x (0.004) = 160,00 fatalities, an enormous number.

The number of people in each age category is known. Experts have constructed a variety of scenarios (all based on models) that produce different predictions for:

• how many people in each age group contract COVID-19

• the fatality rate for each group

How experts arrive at these two numbers differs. Estimates of the number of people in each group who will contract the virus come from models calibrated to data. The fatality rates for each group are derived from data.

Next, we describe how we can improve each estimate.

Diverse Models and Model Ensembles

To make more accurate predictions, experts relay on a variety of models. Diverse models lead to more accurate predictions. Furthermore, within a given class of models, such as the SIR or SIS models that we describe on this site, experts construct ensembles with different parameter settings to establish bounds.

Category Granularity

As more data arrives, fatality rate models can divide the population into finer categories. In addition to age, people can be categorized by overall health, weight, blood pressure, and other attributes that correlate with fatality. The accuracy of these more granular models increases with data.