Thread by @JPSoucy

❗ONTARIO, STOP REACTING TO THE DAILY COVID CASE COUNTS❗

(1/48) Why daily #COVID19 case counts are not the most relevant number and what #Ontario should be paying attention to instead. Mega-thread below. 👇🧵🚨

(2/48) Let’s begin by recapping what daily reported case counts are supposed to represent. When we look at them, we get a picture of what the epidemic looked like 1–2 weeks ago, so any change in the trajectory of the epidemic will take some time to show up in the daily counts.

(3/48) Why? Because it takes ~5 days from time of infection to begin showing symptoms, and then further days to seek testing, get tested, and finally process the test and report the results. This lag is affected by a number of factors, such as the size of the testing backlog.

(4/48) Following the daily case counts creates a never-ending roller-coaster of headlines—an endless series of “drops”, “spikes”, and “plateaus”. Seldom do these headlines reflect real or meaningful changes in the trajectory of the epidemic.

(5/48) I made this point a little more forcefully in the @TorontoStar yesterday in an article by @KenyonWallace and @EdTubb.

thestar.com/news/gta/2020/…

(6/48) While it is tempting to rely only on the freshest information—the latest daily case count—the reality is that one day of data isn’t going to tell us where the epidemic is going. It’s simply too noisy a metric to be reliable on its own.

(7/48) There are many reasons why one day would have a particularly high or low number that have nothing to do with the trajectory of the virus in the community. For example, clearing a backlog may result in many older cases being “dumped” on a single day.

(8/48) When this error is replicated by decision-makers, it makes public health messaging and policy feel capricious and disorganized. We must avoid the appearance that policy is driven by headlines, rather than clearly communicated metrics.

https://twitter.com/CBCQueensPark/status/1314625259128197122

(9/48) What numbers should we pay attention to instead? To start, the 7-day rolling average of daily cases. This short-term trend “smooths out” a lot of the noise present in the daily numbers. However, these trends cannot always be taken at face value.

(10/48) The trends must be interpreted in context of what’s going on with testing in the province. In Ontario, we have had a lot of upheaval lately. There are a lot of factors to consider here.

(11/48) First, events that change the fraction of cases detected by testing. For example, the suspension of contact tracing in #Toronto, which means fewer downstream infections will be picked up.

theglobeandmail.com/canada/toronto…

(12/48) Suspending contact tracing makes it more difficult to break chains of transmission before they happen. Asymptomatic cases are particularly likely to be missed, since symptoms give people a reason to get tested regardless of whether they get a call from a contact tracer.

(13/48) Other health units have also struggled to keep up with contact tracing. Remember, contact tracing is most likely to be overwhelmed when cases are growing at an unsustainable rate—i.e., where you have the potential to miss the most cases.

orilliamatters.com/coronavirus-co…

(14/48) The province is also doing fewer tests than a week ago, as can be seen on this chart from @jkwan_md. Today, the 7-day average sits around 36k tests/day, significantly less than the stated goal of 50k tests/day by mid-October.

https://twitter.com/jkwan_md/status/1318923458584215552

(15/48) It’s hard to image that demand for testing has decreased. After all, we are entering cold/flu season, which means more respiratory symptoms, and awareness/concern about COVID-19 is as high as ever. Some stories that may shed light on the situation...

(16/48) By @meganogilvie, @katecallen, @jyangstar:

thestar.com/news/gta/2020/…

(17/48) By @MonteiroRecord:

thestar.com/news/canada/20…

(18/48) All else being equal, fewer tests mean a smaller fraction of cases being captured than before. But all else isn’t equal: events that change the population being captured by testing make it difficult to compare daily case numbers before and after.

(19/48) A little over two weeks ago, we transitioned from a walk-in based system to an appointment-based system. For some, this system makes it easier to get tested, for others, harder. The effect of this isn’t easy to predict, but it does represent a shake-up to testing.

(20/48) We also tightened up the criteria for getting tested. Other factors at work? We only got on top of the huge backlog of tests to process a little over a week ago. Backlogs increase the lag between when infections happen and when they get publicly reported.

(21/48) All along the way, there have been bottlenecks in accessing testing—whether it’s people walking away from long lines or people having difficulty getting a testing appointment that works for them.

https://twitter.com/HannanehF/status/1314045444864069635

(22/48) Keeping all of this in mind...how is Ontario doing in terms of daily cases? Are we plateauing yet? Declining? I think it’s too early to tell.

(23/48) The greatest evidence for a plateau is in Toronto. However, because of the suspension in contact tracing and pronounced issues with testing, Toronto is exactly where we would expect to be missing the most new cases compared to before.

(24/48) Removing Toronto leaves sparser evidence of a true plateau, given all the questions about testing in the province.

(25/48) What are some other metrics we should be following? Those that are less sensitive to changes in testing... 🔍🧪

Let’s talk about two: active long-term care outbreaks and hospitalizations.

(26/48) Long-term care outbreaks have been rising more or less continuously since the beginning of September, as seen in this chart from @NathanStall.

https://twitter.com/NathanStall/status/1318924503985065984

(27/48) This is not surprising, as exponential growth in COVID-19 cases began around the same time. Preventing uncontrolled spread in the community is the best way to keep the virus out of long-term care (and workplaces, schools, etc.).

(28/48) Long-term care homes are highly surveilled environments and are very high priority for testing. Residents are especially likely to have severe symptoms, which will be noticed and trigger the testing of everyone in the home.

(29/48) Thus, we are likely to detect a much higher fraction of cases in the long-term care setting than in the community. Long-term care outbreaks are a signal of uncontrolled spread in the community, whether this is picked up by community testing or not.

(30/48) Hospitalizations are another number to watch. Again, if someone has a severe COVID infection and ends up in the hospital, it is very likely that they will be tested and identified as a case.

(31/48) Hospitalizations have been rising since mid-September. Hospitalizations lag infections even more than cases because it takes time for the infection to progress from initial symptoms to severe symptoms requiring hospitalization.

https://twitter.com/jkwan_md/status/1318921850555142151

(32/48) For the same reason, ICU admissions and deaths lag infections even more than hospitalizations. Hospitalization numbers serve as a reminder of the consequences of the several weeks of uncontrolled spread in Ontario: they will keep going up and result in deaths.

(33/48) The term “casedemic” is fashionable to describe the second wave of COVID-19, noting the lack of concomitant deaths to accompany huge surges in cases, particularly in Europe. But deaths are lagged, and now they are beginning to mount.

(34/48) Don’t get me wrong, there are reasons for optimism in Ontario. We’ve been at modified stage 2 for over a week and a half in Toronto, Peel, and Ottawa, so we should be starting to see the effects of this in the data.

https://twitter.com/JPSoucy/status/1314724249039499265

(35/48) Of course, parts of Quebec have been “red zones” since late September and it’s still not clear if that province is on the downward part of the curve yet. (Their testing system is also more opaque than Ontario’s)

(36/48) An alternative data source also gives a reason for hope: SARS-CoV-2 signal in the wastewater is declining in #Ottawa.

https://twitter.com/BogochIsaac/status/1318931548989775874

(37/48) In summary: has growth slowed down in Ontario? Yes, probably. But has growth truly stopped? I don’t think we can confidently say that we’re on the other side of the hill yet.

(38/48) When we do manage to suppress cases, we need a sustainable strategy to keep them suppressed. No one wants to yo-yo from partial lockdown to partial lockdown. This means a robust #TestTraceIsolate strategy, the likes of which has worked so well in many countries.

(39/48) Let’s close with some policy actions and personal actions that can be done to combat #COVID19 in #Canada.

(40/48) Policy actions to combat #COVID19:
- Expand testing 🧪
- Expand contact tracing 🔍
- Approve rapid test kits 🏃
- Full data transparency for the data used in public health decision-making 📊

(41/48) Personal actions to combat #COVID19:
- Move social gatherings online 💻
- Masks 😷
- Physical distancing 🛑
- Ventilation 💨
- Time outdoors 🌲
- Answer the phone if public health calls 📞
- Download the COVID Alert App 📲
- Get your flu shot 💉

(42/48) Everything you need to know about aerosols, ventilation, and COVID-19 from @jljcolorado:

docs.google.com/document/d/e/2…

(43/48) Some great follows for #COVID19 #Ontario not already mentioned: @imgrund, @DFisman, @larissamatukas, @Billius27, @howsmyflattenON, @PatriciaTreble. @kellygrant1, @IrfanDhalla, @ASPphysician, @BogochIsaac.

(44/48) That’s it for today. You can find more from me and my team, the COVID-19 Canada Open Data Working Group, at our website:

opencovid.ca

(45/48) We provide daily #COVID19 #Canada #OpenData here:

github.com/ishaberry/Covi…
opencovid.ca/api/

And #COVID19 #DataViz on our dashboard:
art-bd.shinyapps.io/covid19canada/

(46/48) I also provide a daily, automated backup of nearly 100 government COVID-19 data files, including data on schools. 🏫

github.com/jeanpaulrsoucy…

(47/48) A shout-out to our awesome team: @ishaberry2, @shelbysturrock, @JWright159, @_MeghanONeill, @GBrankston, @DrTanyaRossi, @Kamalraj_ach, @VinyasHarish, @XiaotingXie, @alisonesimmons, & others not on Twitter.

(48/48) If you’re still here, thanks for reading! Have a great evening.