Hi! First a quick introduction. My name is Dean Katsiris. I run Anikio which is a website dedicated to the rental market in Saskatchewan. This post has nothing to do with Anikio (but we’re doing our best to help combat this virus by creating virtual tours for real estate and rental property) but it’s hosted here since I already have this platform built. Incidentally, renters, landlords, and property managers may want to look at our FAQ for answers on coronavirus and renting.
So Why Are We Providing This Information?
In my former life, I was an engineer and product manager for a large multinational company. Data and scoreboards were my way of understanding where we were, identifying problems before they became problems, and making educated decisions. I believe they’re critical for communicating progress towards a goal. This is also a way to feel like I’m doing -something- to help. My wife and youngest sister are out there everyday on the front lines, working as nurses at the hospitals. Creating and trying to update a few graphs is the least I can do!
Today, our common goal is to flatten the curve. At least, long enough for hospitals to get the PPE, ventilators, and supplies they need so that nobody will be turned away. Or, better yet, to give our researchers time to find better treatments and ideally a vaccine. Yet I haven’t found graphs showing what the curve is looking like or how we are doing at flattening from a government source. So I decided to rectify that and created these four scoreboards using official news releases from the Saskatchewan, Manitoba, and Alberta governments. I used CTV Edmonton’s historical data from March 23 for Alberta cities. This information is provided for interest only and though we strive for accuracy, we are not an official source. Please do not base any medical or critical decisions on this data, we cannot be responsible for errors.
Covid-19 Charts With a Saskatchewan Focus
All graphs are built using data directly released by provincial governments; we formerly used a table from CTV Edmonton for the Alberta data, which was reliable, but have replaced it with data directly from the Government of Alberta.
UPDATE (May 15, 2020) – Today is a great day, the last time we will be updating these graphs on a regular basis (at least for now). The northern outbreak has slowed and there are few active or new cases in the province. We’re not out of the woods, but it appears that we have slowed the first wave enough for the government to finally catch up. As of today, there is a brand new dashboard for COVID-19 stats from the government. Yes, they don’t have as much about testing as we do, and yes, our graphs are different, but in the end, the information and infrastructure finally seems to be in place. We will leave this page here for now, in case we ever need to use it again. Stay healthy, wash your hands, and be cautious. We aren’t done with Covid yet but if we can remain careful and vigilant, we will be.
There was a redrawing of boundaries by the province on March 27 (it looks like Prince Albert moved from Central to North, but no clarity as usual). We’ve seen reports that the exponential growth, at least in Canada, is 24-25% so we’ve also added the curve showing approximately what that predicts versus what we are seeing. On April 2 we noted that province wide growth has also slowed from the previous 22% trajectory. Average growth has slowed to 18% province wide and over the last three days, the average growth rate looks to be closer to 6%.
On April 13, 2020, we replaced the 3% growth curve with 2% growth rate. We wanted to err high and we did. This is effectively a doubling every 50 days. On April 19, this trajectory looks to be closer to 1%. On April 22, we added a 1% curve although what I commented that what I was seeing around the city “leads me to believe we will not be on that curve long before the growth increases.” On May 4, we climbed back up past the 2% curve as though the slowdown to 1% never happened. Over the past week, the growth rate has been closer to 4% and increasing. On the present course, I think we may hit 1000 cases before the May long weekend. Restrictions should start slowing growth in the north, but we’re one asymptomatic case away from an outbreak in even more urban areas.
While we aren’t mapping active cases directly on this graph at the moment, you can see them by looking at the difference between the shaded green ‘recoveries’ area and the height of the total cases. They are on a temporary downswing. What is happening is that the number of recoveries, which tracks the growth rate from ~ 2 weeks prior, is lagging the current growth rate. We will start to see the drop in active cases slow as we come to the beginning of the bent curve and eventually start growing again but at the new, lower growth rate assuming growth doesn’t come to a complete halt (extremely unlikely).
On May 4, we swapped the order of the graphs to put the Saskatchewan cases graph first. Initially, we were showing everything about test-positive rate and leading indicators first but at the end, most of us want to see the results. Thanks to outbreaks in Lloyd and La Loche, we’ve essentially wiped out the period of time that we were on the 1% growth curve. We raced past the 2% curve but with the restrictions in place in the north, we’re slowing back towards it again. We’d added a northern outbreak line on May 6 but it didn’t stay useful long (good news!). It appears that the reopening so far (with additional measures and distancing still in place) has been successful. That said, an outbreak of the sort we saw in the north in Saskatoon or Regina could quickly slide the scales back into dangerous territory. It was reassuring to see the recent outbreaks handled quickly and effectively.
One thing that would be interesting to see is per capita numbers for each region in the province. This would illustrate the relative risk of being in public. We still feel it would be best to map the outbreak areas more specifically.
UPDATE (May 13, 2020) – It looks like the northern outbreak is under control now. We’re seeing the growth rate return down to 1% across the province, even with some partial re-opening of businesses. It’s worth noting that while things are re-opening, they are not returning back to normal. It does seem that the measures taken, plus added awareness and caution from the public, is having a great effect on mitigating spread.
Our former leading graph, this next one, is a result of a conversation I had with Dr. Nazeem Muhajarine, a local Professor of Epidemiology at the U of S. Given that the criteria for testing has not dramatically changed, the percentage of tests that are positive is a good leading indicator (see below) of spread in the public. In other words, the trend should help predict if we are flattening the curve while we wait to see cases and test results.
We were ahead of our time on this graph. Dr. Muhajarine forwarded me an article from The Atlantic where they applied this metric in the US. Astonishingly, they are seeing a “test positivity rate” of around 20% (i.e. insufficient testing, many more cases than reported). In New York they were at 41% (!) according to the article. Certainly an indication of a much wider spread than testing and reported cases have shown, due to lack of testing. We have done very well.
UPDATE (May 4, 2020) – One issue I had with initially producing the above graph is that without public data showing which cases result from which test dates, we have to use cumulative numbers. No problem directly, but the further down the road we go, the more difficult it is to move the test positive rate. One potential solution to this dilemma is to use a rolling two-week average, and that’s something I will look at this week to see if it is a little more instructive. Dr. Muhajarine had recommended a daily case analysis, but it was too noisy to be useful on its own. I applied a 4th-order polynomial trend line to average out the noise and show the directionality.
This next graph, added April 22, also resulted from my further conversation with Dr. Muhajarine. This graph shows how growth (or lack of growth) in new cases correlates to growth (or lack thereof) in new tests. Obviously, if new tests plateau, there can only be so many new cases expected. On the other hand, we would expect to see the growth rate decrease over time due to the effect of cumulative cases and tests growing the denominator. My reason for posting this now is because I am concerned that in spite of talk about increasing testing, the last few days we have average 471 tests/day. Just 10 days ago, we were seeing an average of 971 tests/day. So perhaps (just perhaps) some of this dropping off in cases, from 3% to 1% is in part due to reduced testing. As we talk about reopening, it will be more important than ever to be watching that we don’t become complacent from a lack of testing showing a lack of cases. The virus could quickly sneak up on us like that.
UPDATE (May 6, 2020) – I haven’t come across an explanation for the sawtooth pattern in testing and cases, but there’s not doubt that the cases are growing and testing is not. We know that there have been several outbreaks in hospitals and in the north so one hopes that contact tracing is simply the reason for a higher number of cases from the same number of tests, but we should be increasing testing to ensure that we are not missing critical and possibly asymptomatic carriers which could quickly lead back to lockdown.
This fourth graph was first added March 30 for Saskatoon, thanks to some inspiration from Mark Handley’s work. If you haven’t seen Mark’s charting of global doubling rates, have a look! As Saskatoon plateaued temporarily around 160 cases, I focused on northern Saskatchewan due to outbreaks in the region. Thanks to trend comparisons, I was able to state at the end of March that “there is no doubt in my mind that we have flattened the curve, although it could all be for nothing if we don’t stay vigilant.”
On April 2, with three days of data, I was confident enough to add a 13-day doubling rate line. In fact, it didn’t last long there. On April 9, I added a 33-day trend and then a few days later changed that to a 50-day trend (2% growth). Today, April 19, Saskatoon’s growth rate is looking closer to 1%. There are now four distinct growth patterns, 3 days, 6 days, 13 days (I said when it was added that it was still early to set this final number in stone), and now 33 days (actually looking even better than that!) I said a while back that we were flattening the curve and you can see that the curve is gone. For now. I expect to see more cases coming from the Easter weekends and just from my observations of traffic out there. One contaminated flight from Calgary, which you’ll see below is struggling to keep control of their growth, could put us back into higher growth. Incidentally, going back two weeks from March 31 brings us to March 17, the day after the announcement that schools would be closed.
UPDATE (Apr 27, 2020) – Sometimes it’s great to be boring. Saskatoon’s covid cases have stabilized (temporarily) at around 150 cases for about a week. This is excellent news and something the city should be proud of. It supports the case for a light lifting of restrictions. While I expect that this week we’ll start seeing those cases creep up again, for now it’s not much to look at. The one take away is that at even a fairly low 3% growth rate, we’d be closing in on 200 cases right now. Meanwhile, Northern Saskatchewan is taking off and should be a point of focus for the province right now. Especially considering that northern Saskatchewan is sparsely populated, it is surprising to see this much growth in cases. So, finally to the update. For now, we are switching out the Saskatoon graph for one showing Northern Saskatchewan. We will bring Saskatoon’s graph back when there is more to show.
The final graph shows how the five major prairie cities have done starting from the day of their first presumptive coronavirus case. There previously was a graph tracking total cases in addition to the per capita graph below. That was removed due to the scale making detail harder to see as Calgary cases grows past 1000. The curve shapes are the same, but scaled per capita, so aside from absolute case numbers, there was no need for both. The idea of this graph was to see how different cities and provinces are doing compared to us and adapt tactics as needed. On April 3, 2020, the Sask. government finally released their very strange map that loosely defines the regions that cases are divided into. One revelation to the public is that Saskatoon and Regina include potentially significant areas outside those centres and that may skew the cases per capita up for both cities. Unfortunately without better definition from the government or SHA, it is impossible to be as accurate as we should be able to in these comparisons, although population estimates are just that even in the best of circumstances.
On April 13, I made a change from bar graphs to lines to make the data clearer. Now, as Calgary pulls away from the group with a lot of infections, I’m considering what to do to avoid missing smaller per capita changes in the other cities. I also noted that Alberta’s historical numbers all the way back to day 5 seem to change daily. Something is wrong here and in need of scrutiny.
UPDATE (May 4, 2020) – Calgary’s outbreak seems to be slowing down, let’s hope it continues. As a reminder, these are total cases per capita, not active cases per capita. Initially, it would be useful as a sign of how likely someone you encounter in public could be contagious (at least from a known cases perspective). Now with many recoveries, the graphs tend towards showing both risk and spread through the population.
We Need Leading Indicators
What we really need to be tracking are leading indicators. Leading indicators are what will predict the number of cases. If we were worried about weight loss, for example, a leading indicator might be calories consumed per day or how many minutes of exercise. A lagging indicator is something that, by the time you measure, it’s already too late. Like tracking your weight on the scale. By the time you see it’s up, you’ve already gained the weight.
If our stated goal is to reduce the spread of COVID-19, cases is a lagging indicator. Leading indicators are more tricky to define, but I’d suggest the following (and will look into trying to build these soon):
- Number of Tests – The more the population is tested, the lower the chance of unknowingly spreading the disease. I’m not sure if simply showing a growing number of tests will be helpful, though. We probably need a target for at-risk persons (again, hard to define) to be climbing towards. To date, we have conducted over 5000 tests in the province and about 1% of the tests have been positive. The percentage of positive tests is rising which indicates that testing is not keeping up with the spread. It would be great economically to have a 100% rate and not ‘waste’ any tests, but in reality that would mean that there are many, many untested people that are also probably positive. UPDATE Mar 30 – Thanks to help from a local epidemiologist, we have added some additional data to help look at trending. We have a much better handle on leading indicators in most of our graphs as of today.
- Measure of Social Distance/Isolation Compliance – The more people are isolating, the less chance they have of becoming a case and infecting others. I have no idea how we can measure this but am open to suggestions.
We Need Broader Testing
At the time of this writing, we are still being advised in Saskatchewan that “You do not need to be tested for COVID-19 if in the past 14 days you have not travelled outside Canada or had contact with someone diagnosed as having COVID-19.” On the surface, reasonable, but consider this example:
At a recent curling bonspiel in Alberta, 50-60 physicians from across western Canada were most likely exposed to someone with COVID-19. The bonspiel ran March 11-14. On March 12, one man linked to this bonspiel was admitted to an Edmonton Intensive Care Unit with an underlying medical condition known to be a risk for someone with COVID-19. This man would become Alberta’s first COVID-19 related death (March 20). It’s unclear, but from reports it appears that he tested positive on March 14.
Three days later, March 17, Dr. Alan Woo, the Sask. Medical Health Assoc. President, felt symptoms and self isolated. And on March 18, we learned that he had tested positive for the coronavirus.
If you had been in contact with Dr. Woo on March 13 and felt symptoms March 14, you would not be allowed to have a COVID-19 test until March 19. Hopefully, you would self isolate since you had symptoms even without proof of direct exposure. Even though Dr. Woo had been in contact with someone that had tested positive, you were not, and so you had to wait several days or until your symptoms were bad enough to require hospitalization. Adding a second degree of separation, if test supply allows, would help shave a lot of days of potential spread.
While we don’t have enough tests for the whole population yet, and we ARE doing better than a lot of other countries, testing is going to be a key in this. Look at the example of South Korea for an example of how testing works to flatten the curve.
The more testing we have, the less spread, and the better chance of flattening the curve and making sure we have the supplies we need as cases build. The point is to not have people die because (from COVID-19 or anything else) because we don’t have resources to treat them.
UPDATE (April 9, 2020) – The Premier of Saskatchewan has been more vocal about increasing testing volumes and has set goals for daily tests. We are encouraged by this.
We Need More Information
It is unacceptable to me that there is so little information from the government. We should have information on exactly where cases are, not just “north”. For example, it should be shared that there are at least 2 cases in Prince Albert, one case in Humboldt, one in Southend, and so on. The counterargument here is that towns or cities with no known cases may result in citizens not respecting lockdown and social distancing efforts and that is a fair point, but does not excuse the current lack of clarity in regions. We should have information on public places those cases may have been in contact with collected together. There is an argument for privacy but I believe that privacy can be maintained and still have this information available. And if not, it is up to us as a society if privacy should outweigh safety in a crisis like this.
