MIC CAVAZZINI: There’s a nerdy meme that went round last year showing a page from the classic children’s book, Winnie the Pooh. In the sketch, Pooh bear is strolling out of the woods with little Piglet by his side. There’s snow on the ground and their footprints lead back towards the viewer. Piglet is trailing behind with thoughts on his mind.
“But how will we know if our pandemic guidelines work?” asked Piglet.
“The world will think we overreacted,” said Pooh.
“So even when we’re right, everyone think we’re wrong?”
“Welcome to Public Health,” said Pooh. And Piglet understood.
That’s essentially the theme of today’s episode of Pomegranate Health. I’m Mic Cavazzini and today I’m going to share with you is a snapshot from the Congress of the Royal Australasian College of Physicians last autumn. One of the main themes was forecasting the impacts of the COVID-19 pandemic, and which public health interventions might best mitigate these. I’ve compiled some of those seminars here as an homage to all the physicians, epidemiologists, statisticians and computational modellers who’ve stuck their necks out presenting their findings to government. It's a tough gig, when the nation is watching on and worrying how their lives and their livelihoods will be affected.
First, I want to start by laying out some fundamentals. I’m sure you know by now that the basic reproduction number, R, of the novel coronavirus identified last year was calculated to be around four or less. That means an infected person would pass the virus on to four people, they would each pass it on to four more people and so on, generating exponential growth in case numbers.
Each of the social distancing restrictions we’re now familiar with will nudge that reproduction number down by a fraction. There’s a great figure representing this in the National Plan commissioned from the Peter Doherty Institute for Infection and Immunity. The researchers described four bundles of Public Health and Social Measures of increasing severity. At baseline are density limits of 2 square metres per person indoors and a 70% capacity limit on large sporting venues. The next bundle means there are requirements for record-keeping and COVID-safe plans at all venues, and a few more restrictions on workplaces.
Under stage 3 or moderate lockdowns people are supposed to work from home if they can, cafes and restaurants can only serve seated guests, and concerts and non-essential travel are off the cards. Stage 4 is what we call strict lockdown. That means that only workers from nominated industries can leave the house for work, schools are closed, and outdoor exercise can only be done in pairs. In the Doherty models curfews are an assumed part of this bundle.
Taken together the three top tier bundles only lower the reproduction number by about one point. A far greater impact is made by an effective testing and contact tracing regime. That doesn’t just rely on laboratory assays for SARS-CoV-2, but also a highly coordinated program to follow up positive cases, interview them about their movements, and get each of these people to avoid further contacts until definitively cleared of the virus. There are hundreds of personnel behind this regime labelled “test, trace, isolate, quarantine” or TTIQ for short. Along with the baseline social distancing measures, TTIQ can knock almost five points off the reproduction number.
This would bring the effective R value below one, meaning that total infection numbers would eventually peter out. Even before the advent of vaccines against the novel coronavirus, there was reason to have confidence that outbreaks could be stamped out with a swift application of these approaches. Michael Baker is a Professor of Public Health at the University of Otago and from early on in the pandemic he advised the government of Aotearoa-New Zealand that elimination of the virus was a realistic goal. Professor Baker’s presentation to an audience in Wellington was accompanied by slides, and while they’re not essential, I’ve embedded all the references he points to into a transcript available at our website. That’s racp.edu.au/podcast. The talk was given on April 30th, before the delta strain had reared its head on our shores, but there are still many great insights that endure.
MICHAEL BAKER: There is of course a phenomenal history of pandemic diseases. One of the challenges between pandemics is reminding people that pandemics do happen. You know, we used the opportunity back in 2018 just to remind people about the anniversary of the largest pandemic in the modern age – 1918 influenza – and in New Zealand, this swept through our country in six weeks and killed 9,000 people . And now COVID-19 is making a mark on this historical record.
So there is a systemic way of sizing up pandemics and working out which ones should you really worry about. And there’s three main things you need to know: how transmissible, because that determines how many people get infected, how severe is the impact, including inequalities, and how controllable is it? Obviously by definition, you don’t know a lot of these attributes early on in a emerging disease. So, this is why you have to use modelling to work out what a pandemic will do, you can’t say, “What did it do last year?” Which is what applies to things like diabetes and other things that emerge slowly— you can extrapolate from past experience. You can’t do that with a pandemic, you have to take some reasonable estimates of parameters, put them in a model and generate scenarios for what might happen.
So, we did this work for the Ministry of Health in New Zealand – I know the modellers all around the world are doing this and our most plausible estimate was around 13,000 deaths, about 0.3 percent of the population, which is like 25 seasonal flu epidemics stacked on top of each other, and it’s moving into the realm of the 1918 flu pandemic in terms of population impact.
So, one of the other huge concerns I think across New Zealand and Australia is about inequalities – we know pandemics magnify inequality. This is looking at a century of flu pandemics—The rate ratio for mortality for Māori compared with non-Māori—and you can see it’s markedly higher right up until the present day, until 2009; a two to three-fold higher risk of death in these events.
So, one of the other techniques we’ve got in addition to epidemiology, of course we’ve got molecular medicine and genome typing, and that has given us amazing new insights into the behaviour of pandemics. And this is showing that the diversity of organisms that arrived in New Zealand was very representative of the global diversity, so probably like Australia, because of tourism and so on, we imported virus from all around the globe. But one of the amazing insights from this is that only one in five cases actually results in much sustained transmission, so it’s like you threw a five-sided dice every time you get a case in the community, and if it comes up with one number, then it’s going to generate further cases. And obviously, that’s enough to sustain a global pandemic, but it is this very lumpy or heterogenous risk of transmission.
I know as epidemiologists, we are sort of paid to worry, and I remember in January hearing about a new coronavirus in China—we’ve been there before. And by the end of January, this was definitely going to be a global pandemic, there was no question from the way it was seeding across the world, so that was very depressing. But by the end of February, there was this really remarkable report, and this is a joint commission from WHO and China that went around China – a very credible scientist saying, “This pandemic has been stopped in Wuhan,” and that had never happened before, that a pandemic had been stopped in full flight. So, that convinced me, and I think hopefully a few others, that elimination should be the preferred approach. I assumed every country on earth would follow that guideline, and as we know, it didn’t.
So, by mid-March, you could see that the virus was transitioning into local transmission in New Zealand, so we certainly needed a lockdown at that point. So, this is going back to the most basic epi principles for infectious diseases, you either control it or you eliminate it. Now control with pandemic diseases is usually about mitigation – that’s what all the influenza plans have – that’s reducing the intensity, so it doesn’t overwhelm health systems. But elimination is very well established, there’s nothing exotic about elimination. So, we just immediately wrote an elimination strategy for New Zealand and promoted it very vigorously to the government and really, New Zealand committed to elimination on 23 March when our prime minster said, “We’ve got this alert level system and we’re moving to the highest level of containment – alert level 4 – as fast as we possibly can.” And that astounded people internationally because they said, “Well, you’ve only got 100 cases, no one’s died. Why are you overreacting?” And yes, you have to overreact, and that’s turned out to be supported.
So, I mean, how many of you know about the reproduction number now? And ee all know this is a core concept for working out the level of containment you have of a disease that’s transmitted between people. And this has three things that drive it: transmissibility, contact rate and duration of infection. You can’t alter duration of infection until you have vaccines and antivirals, so you’re left with just two things you can change, and these are almost medieval tools of cordon sanitaire, isolation, quarantine and so on, so they’re very blunt instruments, but that’s what you have. If you’d said in January, “We’re going to shut the borders for an indefinite period or largely close them except for quarantine,” people would have laughed at you. And there were some days when no one arrived in or left New Zealand.
And the third area: dampening down transmission – the traditional approach of this—“Stay at home if you’re sick, wash your hands, sneeze into your elbow”—doesn’t work for COVID-19. It works to some extent, but when most of the transmission is from people who are pre-symptomatic, and it’s respiratory droplets and aerosols, those things don’t work. Masks is the only thing that stops transmission, and we were all very slow to adopt masking because it’s not part of our tradition. For being a mask advocate, I’m now immortalised on some mask packaging. And everyone in New Zealand knows the other part of dampening down transmission – physical distancing – and that’s the alert level system, which is a brilliant piece of risk communication—I think we all understand this; it’s kind of etched in our retinas now.
So, New Zealand had the most intense – one of the most intense lockdowns in the world for five-weeks, then it reduced it to a lower level for two-weeks and we emerged into a virus free country after that. People said, “Oh we should be living like the Swedes, they enjoy so much freedom.” It’s rubbish, we had much less time under lockdown than virtually every country on earth. If we extrapolate – if we apply the mortality risk in Sweden to New Zealand—that’s 0.14 percent of the population has died from this infection—that’s around 7,000 deaths prevented in New Zealand by not having a fairly typical mortality rate for Europe. And it’s important to remember that people who die from this infection are not all extremely old and frail and living in a rest home—there’s an average of 16-years of life lost per death across the globe at the moment.
And the other thing that was not predicted, was that all-cause mortality has dropped by about five percent as a result of elimination. So that’s around 1,500 lives or people who are living this year who would have normally died last winter, and that’s really the end of circulating influenza and other respiratory viruses. But really, the idea that 500 people die every winter in New Zealand from influenza and we just accept that – worse than the road toll – when we now know that actually, that’s preventable. We’re not exactly sure, I mean, maybe mask wearing in waiting rooms and public transport over the winter months may be one of the tools we can use, but just to think about the alert level system and the fact that perhaps that should be more of a feature of our ongoing existence.
And of course, you need a social safety net; you have to support the people who are most disadvantaged by these very stringent measures, and a huge amount of effort went into this in New Zealand with various forms of wage subsidy schemes and so on. But the other thing that I think was remarkable was the way that communities responded – and we have done quite a bit of work with Kokiri Marae in Wellington, and they did remarkable things to support their communities in terms of income support and particularly food security and healthcare access.
And so, if you think about how do you institutionalise all of that? I think you want ways of trying to de-politicise, and I’m pleased that it’s flagging that we’re going to have a national public health agency. I think the elimination strategy offers huge potential to deal with other global health threats. A programme we just set up – the SYMBIOTIC programme—conceived before COVID-19, but it is quite a good fit. And one of the goals of this is to work through all the other infectious diseases that are high candidates or good candidates for elimination at the moment, and it includes things like HCV, even TB and certainly H. pylori is a big focus for us, and that causes 90 percent of stomach cancer in New Zealand.
And just when you start to feel very bleak, occasionally there is cause for optimism. What are the things that the COVID-19 response has made more thinkable? And I would say support for active government; trust in the benefits of collective action; better infection control; innovative healthcare delivery; a focus on mental wellbeing, connectedness; universal basic income; awareness about the quality and safety of workplaces; reduce commuting, road crashes, air pollution; and homeless was ended during the lockdown in New Zealand. So, all the unthinkable things suddenly became very doable during that period and hopefully we can carry forward that sense of agency we have as a result of the pandemic response. And we can shape the future, but I do think we have to be incredibly determined and focused on doing that.
MIC CAVAZZINI: You heard Michael Baker describe border closures and quarantine as “medieval tools of pandemic control.” Indeed, they go back to the 14th Century when the Great Plague wiped out half of Europe’s population. At the Venetian port , we know now as Dubrovnik, ship crews from other hotspots in the Mediterranean were forced to stay on a deserted rocky island before coming into town. It started out as a thirty day period, but it’s thought that the biblical connotations of forty days testing were too hard to resist. Hence “una quarantena di giorni.”
During the emergence of the novel coronavirus last year, Australia and Aotearoa-New Zealand were blissfully isolated from worst it delivered to the other continents. We drastically reduced flight traffic, and made incoming passengers quarantine in hotels for two weeks before they dispersed into the community. So were these quarantine measures good enough? Since the start of the pandemic New Zealand has experienced eleven quarantine failures and Australia is up to 22. Between them they led to ten lockdowns, including the ones both countries are enduring now. As a fraction of all COVID-19 cases arriving on our shores since the start of the pandemic, 1 in 204 has resulted in the virus slipping into the community. At various times, the Australian prime minister and other members of Government have tried to suggest the rate of failure is fifty times lower, but that’s only if you take as the denominator every single returning traveller, not just the infected ones, which becomes a pretty meaningless ratio.
Already at the RACP Congress in May, Professor Marylouise McLaws highlighted several limitations of our border quarantine measures. She’s a Professor of Epidemiology, Hospital Infection and Infectious Diseases Control at the University of New South Wales and Australia’s focal point for the WHO Global Outbreak Alert and Response Network. Right off the bat, there’s the question of case detection and triage of new arrivals. People with a SARS-CoV-2 infection already have a massive viral load days before symptom onset, orders of magnitude higher than what it is for the original SARS coronavirus or MERS. So it’s no good just pulling aside travellers who have a fever or seem unwell, Professor McLaws said you need to screen all people on arrival with a rapid-antigen test, as these assays have a turnaround under 30 minutes. I can’t imagine this would be more costly than the security we’ve become used to since the 2001 terror attacks, at least until flight numbers return to their pre-pandemic levels.
MaryLouise McLaws said that under a more precautionary border regime, positive cases would be transferred to dedicated quarantine facilities not to hotels. Consider that a hospital infectious diseases ward has a ventilation system that extracts and exchanges the entire air volume at least ten times every hour. In the absence of that high tech option, facilities like Howard Springs, 30km south of Darwin, allow passengers to sit out their two week stay in isolated cabins. State and Federal leaders wasted months passing the buck about who would pay to build more such camps, but you wonder why a Big4 caravan parks couldn’t have been commandeered rather than settling for inner city hotels. Professor McLaws said that hotels might be ok for arriving travellers who returned a negative COVID test, but with some extra provisions in place; not putting them in facing rooms; CCTV in corridors to minimise and trace breaches; and repeated testing of travellers and quarantine workers every day or two.
Recall, however, that the delta strain jumped host even before it got to check in to its four-star hotel. It likely arrived in the country around the 11th of June with an infected FedEx flight crew from the US, and while these staff are tested on arrival it’s assumed these tests came back false negative. The virus was then passed to the limousine driver who picked them at up at the airport, as he wasn’t required by his employer or by public health orders to wear a mask nor was he prioritised for vaccination. A study submitted to the MJA indicates that many border workers too were unvaccinated at the time despite being at the front of the queue for the vaccine rollout. This was six weeks after Australia had briefly stopped flights from India when the severity of the new variant became apparent.
The limousine driver tested positive on the 16th of June, and another four cases popped up within two days. Mask were again made compulsory on public transport, but the NSW Premier held off on declaring mobility restrictions believing that TTIQ would be enough to stop this becoming an outbreak. By the 26th June scores of cases had been detected in the Eastern Suburbs and stay-at-home orders were implemented in these areas and central Sydney. That approach had worked for the northern beaches outbreak in December, and the Premier was lauded by federal ministers and many commentators for being able to the state open for business. But that cluster was certainly constrained by the fact there are very few roads in and out of the northern beaches, and its peak was superceded within days by the delta outbreak. The virus started cropping up in Sydney’s south-west and on the 9th July stage 3 restrictions were put in place city wide, almost a month after the limo driver had tested positive.
By mid-July researchers at the Burnet Institute and at the Uni of Sydney’s Complex Systems Research Group were presented modelling data that showed Sydney was treading water at best, and that tougher restrictions were needed to actually bring case numbers down. Stay-at-home orders were implemented on the 18th of July, but still florists and hardware stores and construction sites remained open and there was call for the list of essential services to be defined more strictly.
At this time, Professor Tony Blakely from the University of Melbourne’s Centre for Epidemiology and Biostatistics Research advised that if the strictest restrictions were adhered to, the number of daily new cases should dwindle to a single hand count in 5 to 8 weeks. He and his team had earned their stripes guiding Victoria out of its second wave of the alpha strain almost exactly a year prior to Sydney’s delta outbreak. Tony Blakely’s presentation to Congress relied on some complex graphs, but I want to share with you his general approach to integrating epidemiological and economic parameters into a single forecasting simulation.
TONY BLAKELEY: My entry into this was quite atypical, I think – mind you, nothing’s particularly typical about COVID. My background is modelling NCDs. I’ve done a bit of infectious disease, but not a lot. What motivates me is this thing called the hundred manila folder problem. There’s 100 manila folders on that desk, they’re all different interventions, we need to decide between them. They’re hopelessly incomparable in terms of health gains, costs. So, that’s what’s really driven my career for the last 10 plus years; to be trying to create infrastructure that can answer any generic question like this. “What is the health gain, say, of an intervention? Could be reducing salt in bread. How much does the change of price on tobacco affect consumption?” And as part of that, we’re developing this thing called SHIE – Scalable Health Intervention Evaluation – and we had inside this black box some models, some things called proportional multi-state life table models, which I’ll get back to at the end of this talk. Well, what we did was we just dropped into there a COVID model that then feeds into our other models, so that’s how we did it.
So let’s ask some of these questions: can we quantify risk for highly uncertain events? I think so. This is an example of a piece that I published in The Conversation late in April, which just goes through how you can start with 10,000 people in Britain, and if they’ve got 0.5% chance of being infected, and you do all these things including vaccinating them on the plane, but it’s no silver bullet; you might end up with one person out of those 10,000 people who unwittingly gets out into the Australian or New Zealand community without realising they’re infected – without having hard quarantine. The point being here is that I also presented this type of stuff early in the pandemic to Federal Court as an expert witness, and there were about five of us expert witnesses, and another expert witness who I have a huge amount of respect for, was very sort of down on this idea, sort of saying, “Well look, Tony, you just can’t model all this stuff. There’s always going to be somebody who goes around the queue in an immigration sense or gets out of quarantine, and you’ve got assume the worst is going to happen.” And I can see the advantages of both sides, but I still think particularly now as we’re starting to think about opening up again, that we do need to have ways of quantifying that risk, even if it’s hard to do and it’s very uncertain.
Can we model policy responses? So, we ended up collaborating with some people who do traffic modelling using this thing called an agent-based model. Prior to the pandemic I was pretty down on agent-based models, I said they were too hard, too many inputs, just too hard, because it was a microsimulation – individuals, agents. But when we threw ourselves at it was really useful for modelling policy responses like lockdowns, wearing masks, all that sort of thing. And I won’t go through the details, I think it’s been well covered, but we ended up providing the modelling underpinned the roadmap out of the Victorian second wave. Now I’m not saying it’s best, but it did turn out to be fit for purpose, and it worked.
MIC CAVAZZINI: The agent-based models Tony Blakeley was describing have a resolution down to individual human beings, so you need to run them hundreds of times to capture some of the serendipity of each human interactions, and whether or not virus transmission takes place—it’s no accident that they’re called Monte Carlo simulations. As cases in Melbourne and the regions kept creeping up over July and August last year neither masks mandates nor work-from-home orders were enough to turn the tide. Only when the strictest lockdowns including curfews were implemented did the numbers creep back down from a peak of about 500 cases a day.
As the numbers fell, Professor Blakely and his team tried to establish when it would be safe to lift the lockdowns without causing another breakaway surge. They modelled what would happen when you reopened at an average 25 or 10 or even 5 new cases per day for two weeks. Out of these three strategies, the Chief Medical Officers and Premier chose the most conservative—and as they say in the movies, Christmas was saved, after a punishing 16 weeks.
It’s ironic that the same lure of a lockdown-free Christmas is being made to Australians right now. Though the target is no longer to eliminate the virus, an impossible task at around 18,000 infections across the country, but to get enough people vaccinated so that we can “live with the virus.” I’ll come to the uncertainty around that threshold later, but first I want to take a look at the ideological debate about the merit of lockdowns that’s been raging since the early days of the pandemic, when there was no recourse to a vaccine.
I won’t address the finger-pointing between politicians as so much of politics occurs behind closed doors. But I will use as stands in the libertarian pundits who’ve argued furiously that “The cure will be deadlier than the virus.” Adam Creighton, economics writer for the Australian newspaper, referred to the public health response as “medical tyranny” and “the Great Insanity.” Economics Professor Gigi Foster of UNSW described the lockdowns as “A Human Sacrifice” on every media platform looking for some controversy. But the Walkley award for false equivalence might go to veteran Channel Nine reporter and editor Chris Uhlman who compared our lockdowns to a penal colony, the East German police state, George Orwell’s 1984 dystopia and the plight of asylum seekers in offshore detention camps.
There are three main premises that are variously put forward against these public health interventions. First, that stay-at-home orders aren’t sufficient to reduce transmission of SARS-CoV-2 and death. Second, that lockdowns actually cause morbidity and mortality through disruption of routine healthcare and mental health impacts. Finally, is the premise that strict lockdowns weaken the economy so profoundly that they can’t possibly be better than an approach that keeps peoples’ livelihoods intact.
So how do the numbers stack up? Adam Creighton and Gigi Foster have made much of an unreviewed observational analysis from the US National Bureau of Economic Research that is an interpretation of The World Mortality Dataset published in eLife. The authors simply compared mortality before and after the implementation of Stay at Home policies during 2020. They found that across 43 countries, there were, on average, more deaths right after lockdowns. The authors took this as evidence for failure of these policies, but it’s kind of a no brainer that if there’s already virus spreading widely in the community, the lockdown won’t magic all the prevalent cases away the instant it’s put in place.
Another interpretation of the same dataset published in June’s BMJ Global Health presents a rather different picture. The lead author was Gideon Meyerowitz-Katz an epidemiologist at the University of Wollongong. He and his colleagues highlighted that New Zealand Australia, South Korea and Thailand experienced little or no excess mortality during 2020. What they had in common was that they’d implemented strict lockdowns when there were few cases in the community. The authors agreed that for other countries where outbreaks had already gotten out of control, you couldn’t say whether lockdowns had provided a mortality benefit or not.
Rather than just viewing lockdowns in a binary way, a couple of research groups have used big data to examine population mobility in a literal sense. In an article just published in the EMBO Press researchers used iPhone map metadata from Israel and four European countries. The averaged mobility of the user sample did drop measurably when significant social restrictions were implemented and the researchers found that the quicker this happened from an seeding event, the lower the mortality impact of an outbreak. However, the amplitude of the drop in mobility was not correlated to deaths so the authors inferred that the strictness of social measures isn’t as important as the speed of implementation.
Another team at the University of Hongkong instead took data from the city’s tap-on metro system as a digital footprint of the population. In a paper in Nature Communications from March they describe how case numbers were processed to give a time-calibrated lookup of the effective reproduction number of an outbreak. This was correlated closely to their proxy measure of mobility and there were notable declines when work-from-home restrictions were put in place. Novel analytic methods like those described in these papers could help refine our public health measures to be more effective, with less of the heartache that comes with blunt “medieval” interventions.
Let’s go back to the claim from sceptics that lockdowns actually increase excess morbidity and mortality from disrupted healthcare and what they call “deaths of despair”. Kim Sutherland and colleagues at the Agency for Clinical Innovation examined data of healthcare episodes in NSW. They reported that in the four months between March and June 2020 there was a 14% reduction in visits to emergency departments and a 33% decline in planned surgical activity. Even with the creation of new telehealth item numbers for general practices, the there was a 22% decline in the rate of primary care consultations, and half of breast screens were deferred too.
The longer-term impact of these delayed visits remains to be observed, but it’s not unique to communities that go into hard lockdown. As Gideon Meyerowitz-Katz writes, in the UK too there was a significant drops in general presentations to emergency departments and cancer screenings, likely because people were afraid of being exposed to the virus and because staff had been reallocated to deal with COVID cases. There’s also evidence that patients who attended ED departments in the middle of an outbreak, for reasons other than COVID-19, had poorer health outcomes. Even as I speak, the New South Wales and Victorian health systems are starting to show signs of strain under the rising COVID-19 cases; there have been high acuity retrievals for fractures forced to wait for an hour because the ambulances are backed up on ED ramps waiting to unload.
It’s intuitive to presume that lockdowns are terrible for mental health, and there are recent data from NSW Health that support such link. In the under 18 cohort, mental health presentations to emergency rooms have increased by 36% over 2019 rates, and self-harm presentations by 47%. These are tragic figures that need a considered response, but according to Meyerowitz-Katz “there is consistent and robust evidence from many countries that government interventions to control COVID-19 have not been associated with increased deaths from suicide…We do not mean for the conclusion of this paper to be that lockdowns cannot cause any harm…While it is likely that lockdowns do have negative effects, [the data] show quite convincingly that the [government] interventions themselves cannot be worse than large COVID-19 outbreaks, at least in the short term.” I don’t know how much of this has been quantified in high mortality countries, but it’s been impossible to avoid the countless anecdotes of people mourning the death of a loved one, of health workers burnt out, and along with many other essential workers, traumatised by their exposure to mortal danger.
For sake of argument, let’s agree that hard and fast lockdowns have a lesser impact on general wellbeing than an uncontrolled outbreak, because they end sooner. For weak and slow lockdowns the data aren’t as clear, though the harms are likely in the same ballpark. So are lockdowns worth the staggering economic costs, when you consider that NSW and Victoria are each forgoing around $1bn a week in lost productivity, according to the AMP bank. In her 2020 commentary for the Australian Journal of Labour Economics Gigi Foster drew attention not just to the lost employment of working adults but also the lifetime impact on earnings for students whose education had been interrupted. According to her “back-of-envelope estimates” this deficit of billions of dollars adds up to more statistical lives than would otherwise have been actually lost to COVID-19 without lockdowns in place.
We can’t be too squeamish about putting a dollar value on human life given we do this all the time in health economics. Government agencies like the Pharmaceutical Benefits Advisory Committee and Pharmac might ask whether it’s worth subsidising an expensive new anti-cancer agent that extends a life by a couple of years, or a rheumatology drug that allow people with arthritis to return to everyday activities and work. They do this using the concept of Quality Adjusted Life Years or QALYs. One year in perfect health is worth one QALY and a disease-related reduction in quality of life brings that value down. PBAC doesn’t have an explicit value for one QALY, but in practice it’s usually around $55,000 Australian dollars, and therefore the value of an average statistical life adds up to a bit over four and a half million dollars. Just as we can use this benchmark to weigh up new drugs, we can do the same for other public health interventions like lockdowns, often using a broader measure called Health Adjusted Life Years or HALYs.
In an article for the Conversation, Tony Blakely and colleagues directly acknowledged Gigi Foster’s concerns but warned that letting the virus off the leash to keep businesses open longer would incur an even greater economic penalty. This was based on modelling of GDP losses linked to different lockdown strategies now published in the July edition of JAMA Health Forum. Professor Blakely and colleagues took epidemiological data from the 2020 outbreaks and overlaid four scenarios where restrictions would kick in when new COVID case numbers were at 8 a day, 30 a day, 120 a day, or 700 new cases a day. The first two strategies are focused on complete elimination of the virus while the second two aim to suppress case numbers just enough to stop the health system getting overwhelmed.
The simulation was run a hundred times and on every one of those runs the more modest suppression strategies resulted in greater health costs of treating COVID-19 patients. That’s pretty logical, given the loss of HALYs was 8 to 40 times greater and hospital beds are not cheap. But the surprising finding was that all four strategies ultimately resulted in a similar number of days under social restrictions. In fact, for the suppression strategies, a greater proportion of these were under stage 4 or ‘hard lockdown.’ When GDP losses over 12 months were compared, there was a lot of overlap between the four strategies.
TONY BLAKELY: So, what we’ve done is we’ve now taken that agent-based modelling stuff and linked it up to our models that estimate HALYs and costs. I’m just going to show you the main results. This thing of great beauty is called a cost-effectiveness acceptability curve – it comes from health economics. What it is, is on the y axis is the probability that your policy is optimal – it’s the probability because you’ve got hundreds or thousands of runs in here, and you can see for which unfolding of reality, which policy worked out best; aggressive elimination being what Victoria and New Zealand did, moderate elimination more what New South Wales did – a bit later to go to lockdown; Tight suppression would be South Korea, say, Singapore, where they don’t aim to eliminate, but they keep the levels low; Loose suppression will be something like what Europe was doing before Christmas last year.
And you can see here you have various willingness to pay—that’s how much you pay for a HALY or a QALY and you can quite clearly see here from a health system perspective – taken into account the health expenditure – either of those elimination strategies is preferable; they are optimal more the time. And then we also did it by bringing in all the GDP losses because the GDP losses far exceed what’s happening in health expenditure. Again, a similar story, there’s a lot of uncertainty here, but all the way to $500,000 per HALY, we always see that moderate elimination is coming up preferred half the time, and then a quarter of the time aggressive elimination.
MIC CAVAZZINI: Just to make that clear, if you factor in the broader economic costs of COVID-19, one of the elimination strategies was cheaper and more effective in 3 out of 4 runs of the simulation. That’s at the $55,000 per HALY threshold we’re used to, and all values above or below it.
Armchair commentators love to disparage the boffins and their models, but when there’s so little real-life evidence to go on, and the stakes are so high, the boffins are worth a listen. And there is some observational support for their claims too. A study published in the Lancet compared economic performance of OECD countries that opted for suppression versus those that went for elimination of COVID-19. Up to March 2021, those going for elimination were performing several points better in terms of relative change in GDP. And on a measure of civil liberties known as the Oxford Stringency Index, the strict countries actually suffered half as much as those that let their hair out like Sweden.
It’s true, we don’t know yet, and haven’t modelled, all the downstream costs, social and economic, of having people stay home without work or school. It seems reasonable to imagine that these would in proportion to the short-term GDP losses shown by this research and therefore no worse in a well-executed elimination plan. It has to be noted that the Tony Blakely also didn’t factor in the costs and rewards of the Australian government’s economic stimulus package into their models, an eye-watering sum of 507 billion dollars. But like I said, in that parallel universe where we took a softer approach instead, the economy is no better off that it is for us here, and that parallel government is likely doling out an equivalent rescue package. Indeed, as a proportion of GDP, our 27 percent is on par with the US, and not all that far from Sweden’s 23% or Germany’s 20% stimuli.
Professor Foster rightly points out that the economic penalty of lockdowns is being disproportionately felt by lower socio-economic groups, service workers who don’t have the luxury of tapping away on a laptop at home as I do. But as reported by the Australian Institute of Health and Welfare, they are also the more likely to suffer the worst of the pandemic. I quote, “There were almost 4 times as many deaths due to COVID-19 for people living in the lowest socioeconomic group compared with the highest socioeconomic group, and age-standardised mortality rates were 2.6 times as high.” Rather than sacrificing the health of those communities to ensure business as usual, targeted social supports can alleviate the worst of the lockdowns.
I’ve no doubt that the sceptics will take the worsening outbreak in NSW and Victoria as validation that the past sacrifices were a waste of time and money, that we should have just let the virus rip through the community and burn out. Well, another modelling study published in June’s PLOS One presented a sobering response to this proposal. The work was led by Professor Tom Kompas, Director of the Australian Centre for Biosecurity and Environmental Economics. In an unvaccinated population, abandoning any control measures would lead to a peak of 260,000 cases needing hospitalization at one time, 80,000 of which would need intensive care management.
Australia only has about 2300 permanent ICU beds. We were able to pull together another 4000 last year in anticipation of a surge, which journalist Adam Creighton crowed was a waste money—exactly the response predicted by Winnie the Pooh at the start of this podcast. But only 60% of those new beds could be supplied with mechanical ventilators, according to an audit published in the MJA. And the real limitations would be the extra 3000 intensive care specialists and 30,000 ICU nurses needed to actually staff those beds around the clock.
Of course, using public health interventions we have avoided that worst-case disaster, and the vaccination program has made significant progress, but we are not out of the woods yet. Over the month of August, the number of COVID cases in ICU in NSW went from 54 to 150, and by the first week of September total occupancy of beds was at 80% of capacity. Normally patients stay in intensive care a few days before they go back to the wards to recover. But research from the UK found that the average length of stay in the ICU for patients who survived COVID-19 was three weeks. So letting the virus rip doesn’t just mean an increase in the number cases coming into the ICU, there’s also severe bottleneck in the turnover of those beds. Of course it wouldn’t just be COVID-19 patients who would suffer, but anyone needing intensive care after surgical complications or a traumatic injury.
Just like in Tony Blakeley’s models, Tom Kompas and colleagues observed that delaying lockdowns by two weeks or a month after the start of an outbreak provided no economic gain over a snap lockdown, it only increased fatalities and the duration of restrictions. It’s only because of our precautionary approach that we never had to turn underground carparks into hospital wards like they did in Nevada last year, or ice rinks into morgues like they did in Madrid. But the lockdown sceptics presume they have a better insight than those lived experiences, and better numbers than our epidemiologists.
Another parameter they gloss over is the health and productivity cost of the long-term morbidity associated with COVID-19. Of people who recover from COVID-19 a substantial proportion experience at least one ongoing symptom six months after the infection has cleared; reports range from 10 to 30 per cent of all patients, and 50 to 75 percent of hospitalized ones. The symptoms, caused by an overexuberant inflammatory damage include fatigue, shortness of breath, cardiac abnormalities and neurological symptoms. Based on Professor Kompas’s morbidity figures, I estimate that a “let COVID rip” scenario could have led to 2 to 4 million Australians suffering from long COVID. That’s comparable to the numbers living with chronic pain, which according to Deloitte Access Economics costs $73 billion dollars a year in lost productivity.
The last chapters of this story are still being written, and it will take longer still for analysts to pick through the data record in more detail. But for now, I think there is no obvious health benefit or economic advantage from letting COVID-19 rip through an unvaccinated population, or implementing lockdowns only as a last resort. Now that Australia’s two most populous states have given up on the elimination game, you can understand the concern that the other COVID zero jurisdictions have about flinging open their borders. They have less relative capacity in their ICUs, and without the live threat of infection in the community, vaccination rates are lagging. It’s also worth looking at Aotearoa-New Zealand, which locked down with little complaint on August 17th the instant the first case of delta variant was detected in the community. After the outbreak peaked at 83 new cases, it has been reigned in to an average in the low twenties at time of recording on September 13th, and restrictions are easing outside the Auckland epicentre.
Until we reach high population coverage of the vaccines, there is no magical third way that these teams of physicians, epidemiologists, and computational scientists have missed that would allow the economy to thrive without real penalties in lives and dollars. Listening to Gigi Foster on Sky News, perhaps I’ve been duped by public servants and politicians with their “mouths in the trough of COVID… keeping this madness going” because they’re profiting off rising Amazon shares or something. Or Chris Uhlman may well identify me as, “a lockdown fetishist” who has “learned to love Big Brother.”
He adds, “Politicians should not outsource their decision-making to doctors and there is such a thing as bad expert advice.” This reductionist claim is wrong in that even now the Premiers have hardly been beholden to the advice of their medical officers. Uhlman’s claim is also dangerous as it undermines experts when their non-partisan voices are needed to inspire public trust and civic behaviour. This is especially critical now that that hesitant people have to be convinced to get vaccinated. As Australian’s have heard repeatedly since the 3rd of August, the National Plan promises that once 70% of the adult population is vaccinated against SARS-CoV-2 then the strictest lockdowns could be lifted. This is based on modelling by the Doherty Institute, and I’ll mention some of the scrutiny of these numbers later. But to understand how vaccine-virus interactions are packed into population models, here is the Institute’s Director of Epidemiology, public health physician Professor Jodie McVernon. Her talk to the RACP Congress preceded the launch of the National Plan but clearly lays out some of the first principles and uncertainties in this work.
JODIE McVERNON: OK, so in thinking about the uses of modelling, I’m going to be talking really about modelling vaccination as a case study here, and in that – the picture that I had in the last slide was a famous cartoon from the era of smallpox, and I think the smallpox furphy is one of the things I find most tiresome about vaccines. There are many who believe – well, there seems to be a sort of an urban myth that having a vaccine means we can eliminate a disease, but in fact, smallpox is the only human disease that we’ve ever really achieved that for. And that was about some very special characteristics of smallpox the disease, and smallpox the vaccine. Basically, those characteristics were that you either died of smallpox or you were immune for life, and the vaccine was the same. So, once you had been vaccinated, that was end of story.
For most vaccines, their actions are much more subtle, and most of the vaccines that we use in our National Immunisation Program schedule have nothing like that mechanism of action and can’t eradicate on that basis. So, modellers talk about this SIR paradigm – and this is a simplified representation, really, of the state of any given individual in a population in relation to an infectious disease; they’re either susceptible to it, and at risk of catching the infection. If they meet a person who’s infected and infectious at some level of contact in the community, transition to a state of being infected and infectious themselves. And there’s a positive feedback loop here because the more infected and infectious people there are out there, the more likely you are to bump into one. So, as prevalence rises, that rate of new infections increases, so, incidence rises. And then having been infected and infectious, individuals may recover and for at least some point in time, be partially resistant to reinfection for many diseases, but not all. But then often over time, that immunity wanes, either because of loss of antibody or because of loss of the function of antibody. And that could be because of new variants of a disease, like influenza, or what we’re seeing with new variants of COVID-19 and concerns about potential vaccine escape.
And in terms of thinking about what an immunisation program achieves in the population; usually reducing death and disease is our primary objective. Vaccines often don’t reduce infection acquisition, they might just reduce severity – and this was the case for the inactivated polio vaccines. They actually allowed people to have polio, transmit polio, but simply stop the progression to neuroparalytic polio. For individuals who then do become infected – they might have a reduced viral load or bacterial load, so their infectiousness might be overall reduced if the load is down, and if their clearance time is quicker, their infectiousness is shortened. So, one or other will reduce the onward spread of disease and, you now, we need to build all of these different steps in to understand what the overall impact there will be.
So, in the way we think about models and in the way we implement the impacts of infection and immunity, we look closely for evidence from basic biology from clinical trials, from observational studies, to see if we can identify at what steps the vaccine is acting. And our challenge with COVID-19 vaccines was all we had initially was clinical endpoint trials, but over time we’re accruing more evidence – and I’ll show a snapshot of that now – but it’s continuing to build quite rapidly at this point in time. And so here were some of the data from Israel – the world’s most highly immunised country through a very useful agreement that was made with Pfizer at the beginning – showing that for those first clinical endpoint outcomes; thinking about documented infections, symptomatic infection, hospitalisation, severe disease and death; we saw a marked difference from as early as 14-days after vaccination between the unvaccinated and vaccinated cohorts that were very convincing evidence that we could certainly stop those outcomes of public health importance, and generate an impact.
But what were the steps along the way that led to that? So here we have a cohort study involving healthcare workers in the UK, showing that in the group who received vaccine as opposed to not, that over time the ability to detect any evidence of infection—so these are asymptomatic or symptomatic individuals who are swabbed repeatedly—showing a reduction in acquisition of infection in healthcare workers, who are obviously all at a reasonably high-level of exposure risk.
So, that’s that first step from S to I; if acquisition is down, that’s step one. What happens then if individuals become infected? So, here we have some data looking at individuals who had a breakthrough infection and looking at CT values – so, CT values correlate inversely with viral load. Over time, as the vaccine kicks in, even a breakthrough infection seems to have a lower viral load, and on basic biological principles, we’d assume that should be less infectious. And there are bridging studies of households, looking at household transmission, showing that for reduced viral titres of COVID-19, the likelihood of onward-spreaded infection is, in fact reduced. Similarly, we have studies for the AstraZeneca vaccine showing changes in viral titre, and we see that those individuals with a prolonged duration of shedding is reduced for those who have been vaccinated. The effects are seen for both variants of concern and non-variants, that this viral shedding is lower.
So, we take all this biological information and we fit it into models that think about impacts at different scales. We might model agent-based systems that think about close contact settings, households, quarantine settings. We also often scale these things up to the whole population level—what does vaccination then mean for the risk of transmission? And there are various stages at which vaccine may have favourable impacts: at the level of our borders, in a nursing home or across the whole community, and we’re trying to think about where those gains are achieved by the program.
Overall though, I think all of us are now aware that we will not eliminate COVID. We may be able to constrain its spread—I think the emerging evidence is very positive in that regard, albeit we’re watching variants of concern very closely. And really, over the longer term, we need to understand more about the degree and duration of protection. But basically, what these vaccines will help us to do is, at whole population level and particularly in populations like Australia, achieve the levels of immunity needed to transition to endemicity and that really is the achievable end goal of our program and really, and really making this a liveable disease. Thanks for your time.
MIC CAVAZZINI: There are now some answers to those questions Jodie McVernon was asking in May, even though the variant of concern she was talking about then was the beta strain of SARS-CoV-2. The delta variant we’re dealing with now may be twice as virulent, so people shed more virus before they become symptomatic and end up infecting more people. But it’s not all bad news. A manuscript from Singapore’s National Centre for Infectious Diseases showed that that while mRNA vaccines don’t reduce the peak viral load of delta in infected people, they do speed its decline. Therefore the amount of time that a case will remain infectious is reduced.
Another study just published in the New England Journal of Medicine show that after two doses, both the Pfizer and Astrazeneca vaccines were almost as effective against the delta variant as are against alpha. After a single dose, however, the effectiveness of either vaccine against delta was shown to be only around 30%, which must less effective than a single dose is against alpha. It's also been shown that although the effectiveness of Pfizer’s vaccine starts higher it wanes more quickly, so by six months the two vaccines are expected to have comparable effectiveness.
Professor McVernon and her colleague will no doubt be updating these ever-changing numbers as they come in, and considering the future need for vaccine boosters. I doubt they would argue with the aphorism that “all models are wrong, but some are useful". That is to say every model a gross simplification of reality, that necessarily has to make certain assumptions to make up for all the unknowns. That’s true of the National Plan vaccination targets as well. As I said, the first threshold for easing of restrictions was 70% of the population aged sixteen and over. At 80% vaccine coverage then it was predicted that most other restrictions on recreational venues and travel could be abolished. 80% vaccination coverage would knock the effective reproduction number down by a little over two points. If we maintained baseline restrictions and indoor mask use and TTIQ, the effective R number should come down to 1.2 to 1.4. Not enough to end the outbreak, but enough to keep morbidity at a level that could be tolerated by the health system. The death toll would be expected to fall between 1000 and 1500 people, as compared to the flu seasons of recent years that have ranged from 150 to 1100 deaths.
In the Doherty models, the performance of TTIQ was calibrated to a worst-case scenario comparable to Victoria’s second wave. But over the first week of September, Sydney experienced case numbers that were almost two and a half times greater. Once our army of contact tracers reaches capacity, more and more contacts go unaccounted for and the transmission rate escalates. Richard Dennis, chief economist at the Australia Institute, has written that when the model was being put together, its creators made defensible assumptions about how well Sydney would keep a lid on case numbers. But now that reality on the ground has changed, he reckons it’s disingenuous for the Premier or Prime Minister to insist the National Plan a “safe” one that allows us to “open up…with confidence.” In particular, the national averages will overestimate vaccination rates in some of the most vulnerable communities, who lag behind for reasons of access and health literacy. And if pre-existing risk factors for COVID-19 complications are also overrepresented in those communities, it makes this strategy even less safe for those cohorts. We should take heed from Israel and Singapore, who’ve shown that as well organised as their vaccination campaigns have been, there is no golden ticket to herd immunity.
The Doherty models always warned that stage 3 or 4 lockdowns would have to remain in the back pocket just in case. Tony Blakeley’s team at the Uni of Melbourne have previewed new findings that show we’ll be relying on these public health measures for 41% of the next year if we want to stick to the mortality forecasts of the National Plan. It’s a constantly changing field with many other researchers weighing in, but I can recommend an excellent synthesis from science journalist Liam Mannix for the Sydney Morning Herald and the Age. The biggest variable is how the expert advice is taken on board and communicated by our policymakers.
That’s what we’ll be talking about in the next episode with Jess Kaufman from the Vaccine Acceptance, Uptake and Policy Research Team at the Murdoch Children's Research Institute. I also interviewed Professor Allen Cheng, who as chair of the COVID-19 group at ATAGI, had the headache of making a determination on the safety of the Astrazeneca vaccine, when it’s risk of blood clots was poorly quantified.
For now, I want the thank Michael Baker, Tony Blakely and Jodie McVernon for allowing me to share their presentations here. And also Mary-Louise McLaws, Tom Kompas, Gideon Meyerowitz-Katz and Professor Blakely’s colleague Laxman Bablani, for making sure I’d represented their work accurately. As always, I’m indebted to the members of the podcast editorial group who helped me polish this podcast u, and also to Frank Beard from the College’s expert advisory group for COVID-19. I’m Mic Cavazzini, and I’ve been presenting an editorial view here not speaking on behalf of the Royal Australasian College of Physicians. You can email me at firstname.lastname@example.org. Thanks for listening.