We need to get a baseline on the true community infection mortality rate by performing experiments immediately. With multiple high-profile confirmed COVID19 infections (including politicians, sports stars and actors), but tiny confirmed community infections in the US, UK and Australia, it is quite possible that the reported community infection rate is much, much lower than the true rate. The community rate could be 10X, 100X or maybe even 1000X the currently reported levels due largely to the lack of testing.

Currently our diagnoses have focused on sections of the community who are not truly representative of the broader community. These people include:

• High risk of infection (such as those who have travelled especially those on cruise ships who are typically older)
• High risk of death (such as those in nursing homes)
• Highly important people (such as politicians)
• High resource people

Tests have been very hard to get for people in the US, with almost all people being denied a test, and in Australia doctors have been advised to focus tests on those who have travelled. With limited resources, this is obviously going to improve the outcomes of those who are at risk, but it also denies scientists, politicians and community accurate information on COVID19 mortality.

The bias in testing will have important implications for the mortality rate of the infection, and hence the desired intervention strategy. If the community infection rate is far higher than reported (say 10X higher), the COVID19 mortality rate in Australia could be up to 10x lower (0.3% instead of 3%). If the community infection rate is 100X lower, the mortality rate could be up to 100X lower at only 0.03% (see note [1]).

It's important to note here that I am NOT saying that the mortality rate IS lower. I'm saying that it COULD be lower and this is why we should immediately try to get a baseline on the community infection rate, then do follow-up to track community-wide individuals to adjust our interventions appropriately.

Getting this data is critical to human life. As infections increase, more medical resources will be dedicated to COVID19 while other important issues take a back seat, such as elective (but important) operations, access to life saving hospital resources, etc. The community impact is also large at the moment, and I do not fault the federal government for acting decisively in a low-information environment, but it must be said that the priority should be towards shifting to a high-information environment, such that interventions can be wound back or tailored to better combat the epidemic.

My proposed experiment would be to swab around 1000 people at random, including both people who do and don't show some symptoms, recruited as closely to random as possible, preferably identified via phone interview. After initial testing, these individuals should be followed up to determine subsequent mortality rates and to get a better picture about community infection rates. Now is the perfect time to perform this experiment, as the reported infection rate is low but the statistical power is likely to increase to useful levels in the coming days.

In my PhD in mathematical modeling of HIV, one of the things that frustrated me the most in terms of modelling the real infection levels and HIV mortality was the limitations of the data. Data always has a bias, but good science tries to reduce this bias as much as possible. Australia, with its high resources and small population, is in an ideal position to perform such an experiment. Australia should repeat the world lead it had in the HIV epidemic and be world leaders on accurate COVID19 data collection.

About the author
Dr James Jansson
PhD Mathematical modelling of HIV @ UNSW 2010-2015
https://scholar.google.com/citations?user=isxX2WEAAAAJ

Notes
1: There is a lead time from infection date to death. Increased testing rate will lead to a longer time between between confirmed infection and death. This should be accounted for in any model created.