Thread by @rubiconcapital_

1/ Grab a cup of coffee (or tea)

A comprehensive, neighborhood-by-neighborhood review of #SARSCov2 prevalence/trends in the City of Toronto.

% positivity & cases, with weekly trends since Aug, AND:

*cross referenced with neighborhood census data*

The findings are incredible.

2/ Note: even if you are not in Toronto/Canada, I think you will find this data/analysis compelling, and universally applicable re #SARSCov2/#COVID19 learning and public policy implications.

Toronto’s diversity (>51% visible minority) makes it an interesting case study.

3a/ In this thread, I show/illustrate:

1. for Toronto’s 140 neighbourhoods (and groups of hoods, e.g. DT Core, Northwest), which have increasing/decreasing % pos & cases per 100k.

(Some peaked long before the Oct 10th restrictions. Others still increasing despite restrictions.)

3b/

2. Using detailed census data from 2016, what neighbourhood characteristics/factors correlate (or do not correlate) with % pos / cases & their recent weekly trends (since Aug 30).

(Characteristics include industry employment, % low-income prevalence, and % visible minority)

4a/ Data note/sources:

- Toronto began publishing weekly neighbourhood test % positivity and tests per 1,000 (weekly), and from population data, we can infer cases per 100,000, and their trends. toronto.ca/home/covid-19/…

4b/ Data note/sources:

- for completeness, below are the neighbourhood dataset limitations/caveats from the Toronto #COVID19 tracker website. Note % pos & cases understated, but *directionally correct*.

4c/ Data note/sources:

Link to the neighb'hood census data used: open.toronto.ca/dataset/neighb…

5/ Okay, let’s get started.

6/ Here is the overall picture in Toronto since August 30th. Overall rising cases and positivity.

(Toronto population ~2.73m people (per 2016 census))

7/ But what proportion of the city is experiencing an increase vs. a decrease in those metrics?

(defining a decreasing neighbourhood, somewhat arbitrarily, as one that experienced a two week in a row decline in % positivity AND cases per 100,000)

8/ 97 neighbourhoods, representing 2.0m people (73% of the pop.), are experiencing rising positivity and cases per 100,000, and account for 84% of new cases in the latest three weeks of this dataset (ending Oct 31st).

9/ Conversely, the other 43 neighbourhoods, representing 750k people (27% of the pop.), are experiencing flat/declining positivity and cases per 100,000, and account for just 16% of new cases in the latest three weeks.

10/ What if we cherry-picked a couple of example areas? Like the Downtown Core (“DTCore”)? How is it doing?

Here, %pos & cases/100k peaked at the end of *September*, well in advance of the Oct 10th restrictions. These 10 hoods have a total population of 300,000 (11% of the pop.)

11/ Northwest Toronto (“NWT”), on the other hand, is fairing worse.

Some of the highest positivity rates and continued increases in % positivity (now >9%), despite the Oct 10th restrictions. 282k people (10% of pop., and 22% of recent new cases).

12/ What’s driving the differences? Detail later, but for now just note the following:

% of workforce in service industry (defined later):
NWT-72%
DTCore-41%

% of workforce in knowledge/work-from-home industries:
NWT-19%
DTCore-50%

Average Income
NWT-$28k
DTCore-$49k

13/ So obviously we need to explore some of the neighbourhood specific socioeconomic/demographic factors and see how/if they correlate (not necessarily “causate”) to %positivity and case trends. I will focus mostly on workforce composition.

14a/ Workforce composition. The Toronto census data recorded # of people in 20 industry groups (see picture below).

I categorize the 20 into four “groups”:
1.Services
2.Knowledge/Work-From-Home (“WFH”)
3.Education
4.Healthcare

14b/ Note that I subdivide Service into further “subgroups” into (i) Retail & Other, (ii) Foodservice & Accommodation, (iii) All Other Services. See picture again.

Industry legend source: www23.statcan.gc.ca/imdb/p3VD.pl?F…

15/ How does the % of a neighbourhood’s workforce in a given industry “group” (i.e. Service or WFH) correlate with total positivity, maximum positivity, and cases per 100,000?

Well, here’s how (shown in table form here, and graphic form to follow):

16/ Some of the data form #15 shown graphically:

17/ We see

Service (i.e. community) type jobs strongly positively correlate to % positivity and total cases and knowledge/work-from-home jobs strongly negatively correlate to positivity and cases.

Perhaps it is no surprise, but the correlation levels/consistency is incredible.

18/ Here is where it gets crazy…

19/ Let’s visually examine *trends* in positivity for neighbourhoods with the top 25 & 50 highest % and 25 & 50 lowest % of their workforces in either Service or Work-From-Home Industries, in this recent current “wave” (e.g. since August 30).

20/ This chart is incredible. The average positivity trend for the top 25 & 50 neighborhoods by % of workforce in the Services industries is a straight line up. And neighbourhoods with the lowest concentration of Services workers…positivity barely budges.

21/ Looking at it the opposite way...

i.e. Total % positivity of the neighbourhoods with the 25 & 50 highest vs. 25 & 50 lowest concentrations of knowledge/WFH workers. Neighbourhood % positivity barely budges for the highest concentrations of these workers.

22/ So what about those pesky bars and restaurants? With positivity correlating with Services workforces, doesn’t that mean bars and restaurants are the big culprit?

23/ Well, it appears somewhat.

BUT, of the three Services industry subgroups (foodservice/accom, retail & other, all other services), neighborhood concentration of foodservice/accom industry workers is the *LEAST* positively related to increasing % positivity:

24/ Quickly touching on % visible minorities and non-visible minorities.

Clearly visible minorities are struggling with #COVID19. The 50 neighbourhoods with the highest visible minority concentration account for 41% of the pop., but have 53% of cases since Aug 30, w/rising % pos

25/ % visible minorities in a neighbourhood correlates with concentration of service workers in a neighbourhood, so as we have seen above, occupation could be the driving factor. I will leave it to the immunologists/virologists to comment on immune susceptibility differences.

26/ Moving on.. (and almost done)

27/ So it appears the task & health cost of containing #COVID19 thru lockdowns/societal restrictions *FALLS HEAVIEST* on lower-income neighbourhoods, with minorities and service workers, who *must* go out into society/community to earn a living, and who may be more susceptible.

28/ ...while we protect young, healthy, knowledge/”White-Collar/Work-From-Home workers (of which I am one) who, in the younger ages, are at far less risk of a bad #COVID19 outcome. I say shame on us (although I am trying not to express explicit opinion, rather give information)!

29/ Interestingly, this is *precisely* the dynamic going on that the nice folks behind the @gbdeclaration (@SunetraGupta, @MartinKulldorff, and Jay Bhattacharya) are trying to convey. Here is Dr. Kulldorf’s very eloquent quote from a recent interview:

30/ A topic not explored herein, which I may add later, is how / why peak positivity differ so widely across neighbourhoods. The Downtown Core peaked in September at 3.3%, while others are still rising. (but maybe I leave that one to the Ontario #COVID19 Science Table to answer.)

31/ And on that note, and lastly, in Toronto, 52 neighbourhoods with a total of 935k people, or 34% of the population, *peaked in prevalence in September* at an average positivity of *2.5%*, two-weeks *prior* to the Oct 10th restrictions taking effect.

32/ End thread.

If you made it to the end, thank you so much for reading and considering.