Victoria’s deputy chief health officer has dismissed claims the state has “secret modelling” that shows daily Covid-19 cases rising to 1,100 by the end of next week, saying he has seen no such data from state or federal authorities.
The claim was splashed on the front page of the Australian on Thursday in a story that said the Victorian government’s “own estimates” showed the number was likely to stay above 1,000 for eight days, and “the average number of new cases is not expected to decline until the last week of August”.
The deputy chief health officer, Prof Allen Cheng, who is also a leading epidemiologist, said there was no such government modelling. “I first went to check our internal modelling,” Cheng said. “I also checked with the commonwealth, with Brendan Murphy’s team. We can’t find anything that looks like that.”
Cheng said modelling suggested that stage three and four restrictions in place throughout the state should lead to a decrease in cases in about seven to 10 days. At the end of the lockdown period of six weeks, Cheng said he expected the daily case number would be “substantially less than what it is now”.
It remained unclear where the modelling published by the Australian came from. On Twitter, a user called Paul Sougleris drew attention to similarities with his data. Sougleris is not an epidemiologist, but describes himself as a spatial metadata specialist on his website, which contains models of Covid-19. A section of the website called “stage four modelling” shows the same shape graph with the same data as the one used by the Australian.
Asked if he had a background in modelling or epidemiology, and if he could provide the assumptions or method for the model, Sougleris told Guardian Australia: “I’m just a concerned citizen that felt I needed to graph the numbers that were publicly available. When I saw my graphs, I could see patterns that concerned me so I posted them on Twitter.”
Sougleris said earlier on Twitter: “Apparently I’m the government data modeller.”
I've been outed.— Paul Sougleris (@SougPaul) August 5, 2020
I'm the only one left in Australia who remembers non linear Lorenz statistical analysis "When the present determines the future, but the approximate present does not approximately determine the future"
Apparently I'm the government data modeller#COVID19Aus pic.twitter.com/hthdLBcPHu
“So, like ... who do I invoice?,” he wrote.
Epidemiologists and modelling experts told Guardian Australia the model used by the Australian lacked context. Models that predict case numbers make assumptions about the health system or the behaviour of a population. But even small changes to these elements could drastically change the projections in the model. So governments are more likely to have dozens of different graphs and models to predict infection, bed capacity and deaths, rather than just one.
Ian Marschner, a professor of biostatistics at the University of Sydney, said the model used in the Australian failed to outline what assumptions had been made, or why.
Marschner, who has more than 30 years of experience in biostatistics, clinical epidemiology and public health, said it was likely the model used by the Australian was “extremely pessimistic in its assumptions”.
“It is basically what would happen if stage three and stage four don’t work at all,” he said. “The problem with that is I don’t think anyone thinks there will be no effect from lockdown measures. This is what may happen with no interventions or with interventions just not working at all, which isn’t plausible.”
Associate Prof Michael Brown, from Monash University’s school of physics and astronomy, has expertise in statistical models and said the model immediately raised a number of red flags.
“It has a weird symmetry to it, where the cases fall at the same rate they rise,” he said. “There is no reason for that to happen, and it’s not motivated by science. This modelling doesn’t account for or show how the virus propagates, or how it spreads, or how it spreads given various restrictions.”
He said he was concerned the model could make it to the front page of a newspaper at a time when Victorians were anxious about the virus. “This is a serious crisis,” Brown said. “To put a poorly explained model on the front page of a paper at this time is irresponsible.”
The reporter who wrote The Australian article, Dennis Shanahan, denied Sougleris was his source.
He told Guardian Australia: “Without disclosing any sources I co-operated fully with the Premier’s office in several phone calls this morning.
“I also pointed out to them, as I do to you now, that the graphic image on the front of The Australian was not a reproduction,” he said. “I am aware of all sorts of debates about statistics and have been in contact with some academics today who have views on the Victorian coronavirus case figures. There is debate about the ‘roller coaster’, the three-day peaks, time lags and the effect delays in reporting are having.”
Prof Sally Cripps, a University of Sydney statistician, said: “A good way of determining whether a model is effective or not, is to ask whether it provides the raw data, a description of the mathematical model – including the uncertainty estimates, and a code to run it.”
Prof Tony Blakely, the head of the centre for epidemiology and biostatistics at the University of Melbourne’s school of population and global health, said modelling was already showing that interventions such as masks and stage three lockdowns had resulted in a reduction of cases. Writing in the university’s publication, Pursuit, he said: “A close inspection of the graph in the Australian shows an increasing trend in numbers over the last week that is inconsistent with the (admittedly rollercoaster-like nature) of Victoria’s actual number.
“I would tentatively suggest we’ll probably see another bend in the curve, downwards, 10 days or so after Victoria began its Stage 4 restrictions today.”
Cheng said the daily numbers would already have looked very different if Victoria had not implemented restrictions.
On 25 June the state hit 20 cases and 16 days later, on 11 July, the state had 200 cases. “So what that would imply, if we had continued with that same growth rate, 16 days after that, which would have been 27 July, we would have hit 2,000 cases. And by next week, 12 August, we would have had 20,000 cases,” he said. “Clearly we are not even at 2,000 cases, but certainly the numbers at the moment are certainly too high. We need to get that down.”