On Monday, the Australian government released the disease modelling for the novel coronavirus that it has used to inform its decisions about health system resourcing and societal interventions. In this piece, regular contributor, Dr Neela Janakiramanan, teams up with Dr Freya Shearer, one of the researchers at the University of Melbourne helping to produce these models, and Dr Rebecca Chisholm, an infectious disease modeller from La Trobe University, to explain the purpose, and limits of these models.
The papers released by the Doherty Institute at the University of Melbourne, as well as the ensuing media conference, can be found here.
Over the last few months, many discussions have permeated both the mainstream media and social media as to what impact the novel coronavirus might have on Australia.
Pundits have produced various ‘curves’ to try and illustrate graphically how many people might get infected, or die, from COVID-19. In some cases, these have been simple plots of the data already publicly available, such as the number of infections or the number of deaths. Other curves have been projections, sometimes based on terms we would all be familiar with from grade school such as ‘exponential’ or ‘logarithmic’ and sometimes using terms many of us would have to look up, like a Gompertz curve.
Over the past month, countries have started releasing the modelling that has informed their planning and response to date. Imperial College London released their modelling on March 17, suggesting that the worst-case scenario would be a significant loss of life in the US and UK, and the New Zealand government released their modelling last week. This sparked much speculation in Australia and reinforced how important openness and transparency is to the Australian public.
Now that the Australian models have been released, it is important to understand their scope, and limitations, and how they are different to some of the simpler mathematical calculations we have seen to date.
Infectious disease models can provide insight on a wide range of policy questions that will arise during the course of an epidemic. Each model is developed with a particular focus. In essence, any epidemic has various phases, and no single model can answer all questions that might be relevant for each phase. In the preparedness phase, for example, the model might be designed to show what the health system capacity needs to be in order to meet a projected demand, whereas in later phases, a different model design might help to indicate when certain interventions might be relaxed.
One consideration in the development of an infectious disease model is what happens to individuals as the disease spreads. Before contact with a virus, a certain proportion of the population is susceptible to infection. With a novel disease such as COVID-19, we expect that the majority of the population will be susceptible. Susceptible individuals who are exposed to the virus may then become infected, and if infected, they will either recover or die. Those who recover may be immune to re-infection.
One challenge, especially when dealing with a completely novel infectious disease, is that at the start of the epidemic, some of the disease characteristics are either unknown or poorly understood. It may not be clear how many who are infected will develop symptoms, how many will recover or die, and how quickly individuals will move from one of these disease states to the next. Another key uncertainty about this novel coronavirus is the extent to which infected individuals with few or no symptoms are able to spread the disease.
Some of the things that are often considered to be inherent to the virus are actually an interplay between the virus itself, the population it is infecting, and the health system capability within that population. For example, the average number of people who might contract the virus from any given infected person (referred to as the R0) or the proportion of people who develop disease and die from it, are actually specific to each new population in which the virus circulates.
Some of this disease-specific information can be estimated using data from places that have been impacted by the new disease first, with the caveat that the greater the impact of the disease in these locations, the less likely the data coming out of an overwhelmed health system will be entirely accurate or complete. Even with high-quality overseas data, there is no guarantee that the disease will behave the same within our population, so while we have access to information from overseas, it cannot be assumed that these characteristics are accurate for an Australian context.
On other matters, researchers must make more significant assumptions. At present, it is assumed that those who recover from this novel coronavirus disease have immunity, at least for some period of time, but this may not be scientifically demonstrable until months, and maybe years go by and such immunity is proven (or disproven) by the fullness of time.
It is important to note that all models are limited by the information and knowledge that is used to construct them, and that this uncertainly is higher with a novel infection.
The modelling results released by the University of Melbourne yesterday are based on preparedness scenarios, meaning that the model was being used to project future plausible scenarios. The different scenarios were constructed based on the researchers’ current understanding of the infectiousness and clinical course of disease of this novel coronavirus. This was derived mostly from overseas experience. As such, they are not predictions of what will occur in Australia.
These models were also designed to be useful during the ‘preparedness phase’ – which allowed health departments and health services to prepare for possible scenarios by elevating the capacity of health systems to meet the curve, and for jurisdictions to implement concurrent interventions to decrease and/or delay the peak of the curve.
Importantly, these models have not been calibrated to the current data on what is unfolding right now, after various interventions have taken place.
The current disease patterns hopefully suggest that the social distancing interventions have had some effect at suppressing disease transmission. The next phase of modelling will involve making predictions based on the epidemiological data of what is actually happening in Australian State and Territories now and will require a different modelling approach.
In a time of great uncertainty, with unknown medical risks and great social upheaval, it is natural for people to want transparency so that we can be reassured that our collective sacrifices are both necessary and enough. There is also a pervasive desire to know, as much as possible, what the future might hold.
The models released yesterday are only a start. They were designed specifically, with what data was available, to help Australia prepare, and they are not a current prediction of what the future looks like for Australia.
The team at the University of Melbourne, and researchers at many other institutions, will continue preparing models to help inform Australia’s response to future stages of this epidemic, using up-to-date data as it becomes available.