Modelling Group : Leading Indicators, Commentary and Call for Volunteers

Sources like Johns Hopkins University, Worldometers and so on are being widely utilised to enable analysis. Domestically, the daily reports from the Government Health Department summarise the information available in Australia. Sites like www.cov19data.com.au are aggregating this and state data to build time series.

The group’s key contribution has been via Matthew Tiong’s re-calibration of a model from NMG Consulting first constructed to model the pandemic in South Africa to the Australian situation, and more detail is available here.

While the group is happy to provide a sounding board if people wish to explore the validity of models built, the key area the group thinks is helpful is to try and get ahead of the second wave via “early warnings” (see article on Douglas Isles’ thinking on this).

If there are volunteers interested in getting involved, the challenges ahead are to try and get better leading indicators, and to transmit these into the community (as opposed to transmitting the virus!) In order to maintain this, building appropriate “systems” could also be useful.

At this stage there are some simple resources we can track, or work from overseas we can replicate. The goal of having an “early warning system” is to have the public and authorities ready to restrict activity if it seems that the virus is spreading again quickly.

The two key challenges are the lack of reliable and consistent data globally, and the lag from infection to confirmed cases (in the event a person is tested).  Given the rapidity of an outbreak, trying to get ahead of confirmed cases would be useful. The low number of deaths in Australia present an additional modelling challenge. While contract tracing exists and is a big step forward from the first wave, it still relies on cases being identified.

Early on in the first wave, Douglas Isles identified the problem of community infection being underestimated. While it subsequently transpired that Australians adapted more quickly to the risks than was clear at the time, the case remained valid, that there were considerably more carriers than the reported number of confirmed cases. This was a good example of applying actuarial skills quickly to a problem and making an impact on the debate. 

Australia, by remaining closed despite the low rates of infection, is again positioning itself to be a follower rather than a leader in setting policy. This gives us the clear advantage of being able to track what is going on in the rest of the world, and learning from their errors. This week’s prompt response by South Korea to an increase in cases, provides a case study for our leaders to examine.

If we can measure the extent to which the relaxing of restrictions leads to a pick up in cases, we can use this to inform policy.  It is likely given what we are doing at present that Alex’s report (below) will inform which countries are at risk. Reconciling this to changes in restrictions would be a helpful task for volunteers – we have limited resources to do this at present. For now, Germany looks to be seeing an elevated rate of transmission compared with countries at a similar stage, and hence we should watch this closely.

Current sources/ topics

Alex Stitt has for several weeks been compiling a weekly report which looks at rates of active cases spreading. This provides us with data on the spread of the virus, in a digestible form, but remains conditional on data release. It is keeping us informed on the spread of the disease with the first wave now centred on Latin America and Emerging Markets, and will increasingly point to second wave emergence. Alex’s work is published weekly by the Institute as part of a broader weekly round up. The most recent is article is here.

A key data point we are now looking at, based on Douglas Isles’ method, introduced in late April is the daily rate of new confirmed cases over active case populations. This is seen as a measure of the spread of the virus’ spread from current to new carriers.

To try and get ahead of potential spread we can look at mobility. We know that as the population stopped moving the spread of the disease reduced. We can assume that as it picks up, the spread will increase but, armed with new found knowledge and rules around social distancing, hopefully at a lower rate for the same amount of movement. It seems inevitable there will be second waves – the degree is less clear.

Here the key sources are Citymapper (timely) –  and Google Mobility (broader). There will be nuances here – perhaps tracking car accidents could provide an additional indication of changes in habits of travel eg less concentrated rush hour. The relationship with case transmission, to date, is shown here, by Douglas Isles.

Understanding ICU capacity would allow for forecasts to be made about how many days we are from hitting capacity if we know current daily admissions, rates of change thereof, average time in ICU and total spare capacity, allowing for any seasonal expected use (in winter particularly). Working with health authorities to assist with presentation of the data in a useful way would be worthwhile. Currently the government health department shows the number of ICU’s in use, but gives no indication of capacity and hence of risk. Matthew Tiong’s model will be expanded in this context. 

The Germans have some good presentation of data in this area here.

Martin Mulcare has identified that the similarity of transmissions makes influenza data helpful. He has highlighted the existence of flu surveillance from NSW Health – while useful, it is currently not very timely with March the latest report published, as at 1 June. The link is here.

Another service in this arena which is perhaps worthwhile is Flu Tracking – one can provide their own survey data every week, encourage others to do so, and read their reports.

Finally, Martin also brought to the group’s attention that the world’s most profitable company may also be giving its own assessment of risk via the timing of its closing and re-opening of its stores around the world. With all the data being tracked from its watches and phones, perhaps, while it feels unscientific and lacking in true actuarial method, we should all be keeping a lazy eye on the actions of Apple.

The group would be delighted to hear of more ways we could get ahead of the next wave, and volunteers to help with tracking, displaying and communicating these indicators would be welcomed.

It would also be helpful where people are working on this in a commercial capacity, and are able, that they come forward and reduce risks of duplication of effort, and increase any collaboration benefits that can be achieved. If you are interested, please get in touch via email – COVID19@actuaries.asn.au.

Current members of modelling group

• Mark Prichard (Chair)
• Douglas Isles*
• Julia Lessing*
• Jennifer Lang*^
• Martin Mulcare
• Alex Stitt
• Matthew Tiong
• Raymond Yeow

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.