9 April 2020
I didn’t publish an update yesterday, and in the near future my updates will be less frequently than daily given other commitments.
In general, there are so many sources doing a good job of synthesizing the latest data that my focus will be less on quantity/frequency than quality. I’ll publish fewer updates, and they will be increasingly orineted towards “insight and analysis” more than reporting the latest information.
Underreporting deaths
In the last post I discussed the underreporting of deaths and the difficulty in comparing data on deaths across countries, and asked:
I am not certain, but think that “deaths in hospitals for confirmed COVID-19 cases” may be something like a standard for at least ECDC data. (Does anyone have a source that can confirm this?)
Helpfully, several of you (thank you!) confirmed that this is indeed the case. My friend C pointed me to a transcript of the UK Chief Scientist addressing a question on this here (bold emphasis mine):
Sir Patrick: (24:16)
I’m happy to do that. The international reporting standard for death, all the other countries are based on hospitalized deaths confirmed and that’s the same as the data that you’re seeing. The ONS data, which are important look at overall deaths on death certificates where coronavirus is mentioned so that they are not confirmed deaths necessarily. It’s important to have both of those, but that’s what the difference between the two numbers is.
Tying up loose ends on masks
I discussed masks in some detail on Sunday 5 April. In the update I noted that the WHO were not yet recommended widespread use of masks. Almost as though the WHO were reading these posts (hah!), they updated their recommendations the next day. The updated advice, in typically WHO cautious and heavily qualified language, moves towards the emerging consensus view I described in that earlier post: that widespread use of (homemade, not medical) masks might reduce transmission by reducing the distance droplets travel — but carefully noting that there is a lack of evidence and that medical masks should be reserved for medical use.
Other articles (too many to link) note increasing frequency of masks being required, and increasing voluntary mask adoption. In my extremely limited sample (one trip to the grocery store per week here in France), I’ve noted week-on-week adoption going from 80% to 100% of people working in the grocery store, and from 10% to 70% among customers, even though masks are voluntary (though now gently encouraged) in France.
Ventilator best practices?
There is a growing chorus challenging the ways in which ventilators have been used in many situations so far.
StatNews reports that ventilators may be overused: that they may be of limited benefit, may in fact do harm in many cases, and that other, less-invasive approaches like nasal cannula may have better outcomes for some subsets of patients.
There has been a lot of attention in the UK on the statistics for patients who are placed on ventilators, when Prime Minister Boris Johnson was moved to an ICU to make ventilation possible if required. (Thankfully, earlier tonight we learned that he has now left the ICU.).
What statistics there are make for grim reading. An anaesthesiologist in the US I spoke to reported that 80% of the patients in his hospital who went on to a ventilator ultimately died. This NPR article summarizes several small studies that all report high fatality rates for those COVID-19 patients placed on ventilators. This article in the German press (FAZ) from a pulmonologist argues that ventilators are overused.
This is clearly a space in which the best practices are evolving rapidly. What is definitely clear is that being placed on a ventilator is a very bad sign and indicates that the chance of recovery is fairly low. What we don’t yet know is the extent to which this is because only the sickest patients are placed on ventilators, or because ventilation is the wrong treatment.
Underreporting update
I’ve previously discussed (in a “Resources” post, not yet updated with latest figures) the issues of underreporting of cases.
A study I like (though it has its limitations) has updated its estimates. See here, and the table below. The right way to think about this data is not as a highly accurate estimate of underreporting, but an answer to the question, “If fatality rates are relatively consistent, what would that imply about how many cases there must be as compared to the number of reported casts?”
The chart below shows the 95% confidence interval by country. Here, 100% means that (according to this estimate) all cases are being reported; and 10% that only 1 in 10 are being reported.
False negatives
A recent pre-print analyses the risks of false negatives in PCR tests based on throat and nasal swabs.
The bad news? As epidemiologist Adam Kucharski summarises, “The authors estimated that in a single test of some who first had symptoms 10 days ago, there’s a 33% chance of a false negative with a nasal swab, and 53% chance with a throat swab.”
The probability of a correct positive test decreases rapidly with time from the onset of symptoms. Given this, PCR tests will be less helpful in assess the overall attack rate of the disease in a population.
These false negative rates are higher than I expected and could further contribute to undererporting.
Summer may not help
On the question of whether seasonal effects may reduce the reach or severity of the epidemic in the summer, this study is cautious that we should not expect it to.
Subjective Probability 