4998 cases reported in Tbilisi this week – higher than last week, but also lower than my projection. My projection last week was for 6629 cases, and 4998 is 24.6% less than that – a bit outside my stated 22% margin of error (which produced a range of 5171 – 8088). I’ll talk about the reasons for this error in my analysis section, but for now, just keep in mind that the numbers are still growing, just maybe not as quickly as they did last week.
Nationwide we saw 8393 cases reported this week (up from 6962 last week), with a positive test rate of 4.57% (up from 3.41% last week)
Testing saw a decline from 204386 tests last week to 183523 tests this week, a difference of 20,863.
Projections and Risk Assessment
Last week I adjusted my model to account for increased transmission due to the prevalence of the UK strain. That may have been a mistake, which I’ll discuss below, so this week I’m going back to the prior, unadjusted model. This model predicts 6079 cases reported in Tbilisi next week (with a 22% margin of error, that’s 4742 – 7416 cases). However, due to holidays, I expect actual case reports to be on the low end again.
Nationwide, using an exponential growth model based on 7-day growth rate of 7-day total cases, I’m predicting 10,118 cases (with margin of error: 7,892 – 12,344). Same caveat applies as above.
Worst case, about one in 111 people you encounter in Tbilisi will have coronavirus. In a gathering of 12 people, there’s about a one in ten chance that one or more people will be infected.
This week we have a slew of holidays – including Palm Sunday, today – and in particular I don’t expect much testing, if any, to occur on Good Friday or the whole four-day Easter weekend. So don’t be surprised if the numbers are low again next week. Holidays play complete havoc with my margins of error.
These predictions seem much more optimistic than last week’s predictions – they suggest a much lower peak than I’d feared based on the prior infection rates. I want to be completely transparent but I also don’t want to give out the impression that people can now let their guard down, because on the off chance that anyone actually acts on what I say here, I don’t want to be responsible for someone taking risks that they otherwise wouldn’t have taken and getting sick as a result.
So just to be totally clear: we are definitely still heading towards a peak. It may or may not be worse than the second peak – I happen to think it will, but Gamkrelidze this week said he thinks it might not be – but I think there’s broad general consensus that we’re looking at a large wave of sickness, which is why officials have gone from denying that we’re in a wave to reassuring the public that there will be enough hospital beds and enough diagnostic tests to get us through it.
That said, why are the numbers so much better this week than I thought they’d be last week? It’s a mix of factors, some of which are favorable, and some of which are less favorable.
One: reduced testing. Testing declined by about 10% – in fact, if it had remained steady, we’d be looking at another 400 – 600 cases in Tbilisi alone, which would have placed this week’s case report within the margin of error for last week’s prediction. However, my prediction was about 6600 cases, and adjusting for the testing deficit only gets us to 5600, at most – explaining about a third of the error.
Two, prevalence of the UK strain. Two weeks ago, Georgian health officials were telling us that the UK strain was 20-30% prevalent. Then they revised upwards to 50-70%. As of this week they’re saying 70-90%. But the prevalence of the UK strain didn’t go from 20% to 90% in two weeks. The UK strain spreads quickly, but not that quickly – it needs time to propagate through the population and become the dominant strain. Even given the highest estimates of its contagiousness, for the UK strain to go from 20% to 90% should take about seven weeks, given the infection rates in Georgia right now.
So I never really had good information about the prevalence of the UK strain, and it was a mistake to trust Gamkrelidze’s number – “the best available” doesn’t always translate to “good enough”. At 50% prevalence I would need to adjust the weekly R upwards by a factor of 1.07. At 90% prevalence I’d only adjust the weekly R upwards by a factor of 1.035. In the latter case we get a prediction range (6255; 4879 – 7630) which places this week’s numbers within the margin of error.
Another issue with the UK strain adjustment I made is that I applied the weekly adjustment rate to the 4-day ratio, which I think was a mistake – applying it to the (derived) 7-day ratio gets us down to 6301 (4915 – 7688), again placing this week’s numbers within the margin of error. This is a very small difference in cases, but in exponential functions these very small differences add up. I probably should have tested my adjusted model on historical data before using it for a prediction. Combined, if I’d applied the 90% prevalence adjustment to the derived 7-day ratio, the projection would have been 6095 (4754, 7436).
Still, ignoring the UK strain entirely gets me a better prediction than making any adjustments at all based on it, which seems strange. Probably adjusting a little bit for the UK strain would have been correct, and other factors were just so strong that the additional contagiousness didn’t matter. My takeaway is that the UK strain has probably reached such a high level of prevalence that whatever effects it was going to have on the infection rates have already happened, or else are so small that they won’t factor in to future projections. This is good news in the sense that we aren’t looking at an upcoming surge in infection rates, but bad news in that the UK strain may have worse health outcomes. I’ll say the adjustment for the UK strain accounts for about another third of the error.
Three, weather. Weather appeared to be a minor factor during the first two waves of the pandemic – not at all enough to prevent these waves absent good public policy. However, when the weather is warm enough for people to gather outdoors, but cool enough that they aren’t going back indoors to get air conditioning, people might have the tendency to gather in lower-risk settings. People might also ventilate indoor spaces more by opening doors and windows. I don’t want to quantify this effect because it’s speculative, but I also don’t want to ignore it.
Four, the “control system” – this is when people see rising indicators in the news and adjust their behavior accordingly. Last week I theorized that high death rates and news about hospitals filling up might contribute to people adjusting their behavior to avoid risk. While there is talk about hospitals filling up, it hasn’t happened yet. Death rates have been pretty stable, increasing this week but too recently to have impacted the numbers. Maybe seeing over 1000 cases, on April 13th, was a factor – moving from three figures to four figures is a psychological focal point.
Five, random statistical noise. It could just be a totally weird coincidence that fewer people got infected, or went in for tests. And also, last week could have been an outlier in terms of how many cases there were, making this week a correction, or return to the previous trend.
Six, perhaps public policy pushed in the right direction. For example, the country started cracking down on large gatherings around April 10th. These gatherings are known to play a large role in transmission.
Finally, I want to note – while this last factor wouldn’t produce a sudden decrease in infection rates, it will push rates downward, slowly at first, and more later – we may be approaching a point where enough people have been infected/vaccinated to start slowing the infection rates. Gamkrelidze says the country has had 600,000 to 650,000 cases so far, but the serological survey in Tbilisi found that
35.5% of adults sampled had antibodies an estimated 35.5% of the adult population of Tbilisi has either been diagnosed with covid or has covid antibodies.
Aside: Tbilisi Antibody Survey
Okay, so an aside: it turns out the way this is being reported is a bit deceptive. According to Tsertsvadze, they sampled 903 adults who said they hadn’t been diagnosed with covid, and 215 of them had covid antibodies. That suggests that 23.8% of adults without confirmed covid have had covid. They then added that to the estimated 11.7% of Tbilisi’s adult population with an official covid diagnosis to get 35.5%. 35.5% isn’t the actual result of random sampling, though.
Putting that aside, I notice a discrepancy. Gamkrelidze is saying for every diagnosed case of covid there’s another one undiagnosed. According to Tsertsvadze’s antibody study, for every diagnosed case of covid there are another two undiagnosed. If Tsertsvadze’s results generalize that means that anywhere from 920,000 people (triple the current official case count) to 1,320,000 people (35.5% of Georgia’s population) have had covid in Georgia.
Again, as with the UK strain numbers, I just don’t really know how to interpret this discrepancy. Even though the methodology behind the 35.5% number – relying on a combination of random sampling plus known past test results – is a bit sus, at least it is based on empirical, mostly-current information. But on the other hand, the results of screening tests (at least for teachers) seem to back up the 1:1 ratio rather than the 2:1 ratio. Perhaps some of the people with antibodies had enough exposure to the coronavirus to develop antibodies but not enough to have a detectible viral load on a PCR test or detectible antigens on an antigen test?
In any case, even the highest possible estimate of coronavirus antibody prevalence – that 35.5% of Georgians have antibodies – is nowhere near the 60-70% estimates of where you’d get “herd immunity”, or even near the 50% spot where you get a point of inflection in a logistic growth curve. At this pace of infections + vaccines we’re still months away from where we should expect to see significant effects from herd immunity. But maybe, in combination with a population of people self-isolating (like myself) and others taking reasonable precautions, it could be pushing the infection rate down a tiny bit.
Analysis, Continued: The Third Wave
Gamkrelidze says the peak will arrive in the first two weeks of May. As I said, he says it won’t be quite as dramatic as what we had in November and December. I hope that means he intends for the government to take mitigation measures sooner rather than later. Even with my revised, more modest projections, we’re heading into Easter with 10,000 cases a week nationwide and we know that Easter week means lots of travel, plus large family gatherings, plus crowds at churches.
I’m not even going to try to predict what this will do to actual case counts, or whether there will be a pause in testing that will obscure the issue. Things are going to get weird after Easter. I’m also not clear on exactly what the government is planning to do to ensure that the peak is in the beginning of May – in other words, that cases start going down at the end of May. There is still a lot of pressure on the government to relax restrictions. There is no obvious force that will stop cases from going up throughout May and even Gamkrelidze still seems to want to pretend the government can just do nothing and have the pandemic remain manageable. As I linked above, Tikaradze is saying the government is adding hospital beds already to prepare for the influx of new patients, and in addition to that, Gabunia says they are adding hotels and working with clinics to monitor patients at home. I appreciate the solution-oriented mindset – believe me, I do – but what’s unsaid here is that “we’re adding hospital beds, hotel rooms, and home-care options” sounds a like like the answer to the question “what are we going to do when hospitals start overflowing?” and not at all like the answer to the question “what should we do now that we’ve successfully managed the third wave?”
Tsertsvadze also says that delayed hospitalization is the most common cause of death for covid patients. I’m not sure if this is technically correct but it seems important that covid patients receive timely care and appropriate monitoring, and being forced to use hotel and home care because there isn’t room in hospitals would seem like an impediment to that.
Director Ivane Chkhaidze of the Iashvili Children’s Clinic (a very prestigious clinic where my own kids have been treated) says there have been 30,000 pediatric infections in the country and they’ve put together a team to treat kids with long covid and multisystem inflammatory syndrome. The issue of covid in children is bound to get worse since the UK strain apparently infects more children and since there are currently no vaccines approved for children. I bring this up because people seem to think that since the risk to young people is relatively low, it can be ignored or rounded off in some way. Stopping the virus with public policy until vaccines can get us to herd immunity will save lives, and not just the lives of overweight diabetic alcoholic cancer patients over the age of 70. Letting the virus run rampant through the population for an extra month or two will kill people, including some children, and will make others sick, in some cases for months or longer. All this stuff has consequences.
How Not To Vaccinate
Given the urgency I described above, one might think that the Georgian authorities would be doing everything they can to vaccinate the population. One would be mistaken. Warning: the following section employs heavy sarcasm.
I read an article saying that teachers would be involved in the vaccination process. The thing is, unless you get your news from facebook, you’re really missing out on what’s actually going on. So when the government says “teachers are involved in the vaccination process”, if you think that means that all teachers are now eligible to receive the vaccines, you are as hopelessly naive as I am. Because what it actually means is that the government has changed nothing, the vaccine is still available only to people over the age of 55, and teachers are involved in the sense that by coincidence there happen to be some teachers who are over the age of 55 and they will be allowed to access the vaccines just like everyone else over the age of 55. How generous!
In completely unrelated personal news, my son’s school has moved to online-only for a week because a teacher tested positive for covid last week. This is perfectly normal and not at all disruptive to kids’ educational experience and emotional health. Don’t worry about vaccinating teachers, it’s not important.
In fact, don’t worry about vaccinating anyone at all!
I took a look at the vaccine booking portal this week. Empty appointments, all week. If I were over 55 I could have booked an appointment and had my shot an hour later.
As of April 24th, only 36,056 people have been vaccinated. That’s less than 1000 per day, meaning there are currently more people acquiring immunity from recovering from covid than there are from getting vaccinated. That means they haven’t even used up the initial shipments of 43,000 AZ doses and 29,000 Pfizer doses. They haven’t even started vaccinations with the 100,000 Sinopharm doses. But like, even if they had – people aren’t booking appointments. There are full vials of vaccines sitting in empty offices all over Tbilisi. I thought it would be a while before demand outstripped supply but apparently I was mistaken.
They’re also going to be pausing vaccines for four days for Easter. Because that’s what Jesus would have wanted. Christ is risen! The rest of us are screwed. Not that I can even fault the government for pausing vaccinations – I mean, if people aren’t booking appointments on a random Tuesday in April, they’re definitely not going to book them on Easter weekend.
Obviously, they should open appointments to everyone who wants them, and then vaccinate people 24/7 until they run out of doses. They should also do first doses first. This would save lives, prevent suffering, and allow people to get back to some semblance of a normal life much more quickly.
Or, you know, they could continue vaccinating 6000 people per week. At that rate we’ll start approaching herd immunity – with 60% vaccinated – by sometime in the middle of 2028. Or who knows – maybe their goal is to reach herd immunity sooner by encouraging people to emigrate.
Living with Corona
Sarcasm aside, there’s good news and bad news this week. The good news is the infection rate dropped and the increase in new cases was lower than I expected. The third wave’s peak will probably be lower than the worst-case scenario I outlined last week. May will be bad but maybe not horrible – maybe more comparable to January than to November. At least there’s some reason to hope.
The bad news is, they’re still making plans for hospitals to overflow capacity, and the vaccination program is going nowhere fast. There’s more pandemic fatigue than ever, and people want to open everything up for summer, even as cases (and deaths) are rising. We’re about to experience a holiday bump of unknown severity.
Based on what I’ve learned this week, I’m less pessimistic about the medium-term, and more pessimistic about the long term. I’m pessimistic that vaccination will succeed at all in Georgia. It’s looking like covid will end up like the flu – a seasonal problem that conscientious people will get vaccinated against and everyone else will ignore. It’s looking like we’ll just get used to the added death toll, the added suffering, the added long-covid. We might even get used to wearing masks in many indoor settings, especially during covid and flu season.
And maybe that’s fine. Life is always full of risks and we can only focus so much on them before they start impacting our quality of life. I try to limit myself to one day out of seven obsessing about covid, and I spend the other six working, hanging out with my kids, playing games, reading about non-covid stuff, etc. Maybe by the summer it will be one day every two weeks. Maybe we’re all just getting used to life with coronavirus, and some of us are faster to adapt than others. I just can’t get over the feeling that we’re throwing in the towel when we’re on the cusp of victory – or that humanity as a whole could have beaten covid if we were just, like, 10% more conscientious.
On the other hand, supposedly viruses in general are supposed to evolve to become less deadly over time. Maybe someday corona will just be another cold. Maybe we’ll even cure it. Maybe the takeaway is that medical science gave us options in this pandemic that we didn’t have in the Spanish Flu of 1918 – that we should be grateful the death rate is so low and the vaccines were so quick to arrive. Sure, it could have been better – but it could have been a lot worse.
Conclusion: Confidence Levels
I’m always a bit self-conscious about making predictions about covid given how many experts have made predictions that have turned out to be total nonsense. This is especially so in a week where my prediction was off by 24.6%.
However, if I get one prediction in 20 wrong, that’s an accuracy rate of 95%. I should feel pretty good about that, except I’m a pessimist so my natural tendency is to focus on the one miss, rather than the nineteen hits. Of course, focusing on the one miss does help me stay sharp, and identify mistakes and biases in my thinking, so I can improve my prediction models and train my intuition over time. If you only focus on what you get right, it’s hard to improve.
I just want to make it clear again that when I report official numbers or official statements, I’m highly confident that they are in fact the official numbers and statements. If I have any particular reason to doubt them (e.g. “I think the official case count is low because the positive test rate is high”) I will state that. Usually I will link to sources.
I’m also highly confident in my week to week forecasts – generally, 90-95% confident, meaning I expect to miss about one in ten or one in twenty.
Medium-term forecasts – for instance, numbers after a week, or when the next wave will hit, or when the country will hit 60% vaccination – I’m less certain about. Generally, you can regard them as “informed speculation” unless I put a specific confidence number on them (e.g. I might say “I’m 75% confident we won’t hit 20% vaccination in Georgia in 2021” or “I’m 60% confident there will be a fourth wave starting between September and December”). What “60% confident” means is that if I make 10 predictions at 60% confidence, I expect 6 of them to come true and 4 to be incorrect.
Long-term forecasts – anything a year or more out – are informed speculation. Except the upcoming robot apocalypse – that I’m certain about.
As usual, numbers from 1tv.ge and stopcov.ge unless stated otherwise.