A Regional Peak (?), Part 2
Key takeaway: New COVID-19 cases and associated fatalities in the DC region peaked in May 2020.
Now the analysis:
Newly-confirmed COVID-19 cases and associated fatalities in the DC region have been growing at a decreasing rate for the past month or so. While I had previously reported about a regional peak in early-mid April, that clearly had proven to be a mistaken assessment on my part. With that in mind, I would like to present the data, with a focus on the daily growth of both cases and fatalities.
Percentage increase of cases and fatalities has continues to decline
Generally speaking, this means that concerns about exponential growth within our region has passed. For instance, a few weeks back I stopped calculating the doubling time, because it's pretty much lost its relevance.
You may note that I often use the notation of "<1%" growth rate on the daily report. This simply means that the true percentage growth rate is below 0.5% (my spreadsheet rounds to the nearest integer percentage point for the calculation). Even this statistic has really started to become moot, as the daily growth percentage for both new cases and fatalities is increasingly less than 1%.
For further detail on the decreasing percentage increases, I recommend visiting my recently-published June 2020 summary statistics update.
New cases and fatalities continue to decline
In theory, you could have a leveling off of the percentage rate (i.e. no more exponential growth), but remain flat-lined at a constant absolute daily increase (e.g. 500 new cases daily for the next 30 days). However, the raw count of both new cases and deaths continues to decline. For this, I use a seven-day moving average to help smooth out the irregularities associated with each day of the week (in particular, an apparent "lull" on weekends and holidays followed by "overcompensation" the next 1-2 days), and any other reporting issues.
With this, we see new cases having a very modest bimodal distribution with apparent peaks in late and early May (6 May, 1,304; 21 May, 1,309).
Notably, you can see the dip in early April (use the 9 April mark on the X-axis as a reference), when I thought the regional had initially peaked. Obviously it kept rising through May. I don't think we're seeing a repeat of that phenomenon. The lesson learned is to use a moving average, and keep waiting for more data to come in.
For fatalities, we observed a single apparent peak in early May (3 May, 64).
For both new cases and fatalities, the past few weeks have seen a sharp decline followed by a leveling-off of sorts. To put it in perspective, the average amount of daily new cases (~400) and fatalities (~10) for the past few weeks are both right around the same amount we were observing in early April. Presumably, as both figures continue to decline, we would expect to soon see number similar to those from late March, etc.
What About Testing?
I haven't really tracked testing capacity because I don't have the time or desire to do so. However, it appears that the decrease in newly-confirmed cases is not a function of decreased testing - while testing hasn't been increasing, it also hasn't fallen much since mid-May (note on this chart - it only includes DC and Northern VA testing, because I can't get longitudinal testing data for Maryland. I think the trend is sufficient, however).
Obviously fatalities are not contingent on testing, which makes them a useful statistic (albeit a bit lagging). However, the presence of a decline in new fatalities also is a positive sign.
What's With the Occasional Falling Fatality Counts?
Let's start with a discussion on methodology on calculating fatalities (one which I've been meaning to have for a while).
The three jurisdictions report fatalities different. Maryland lists "confirmed" and "probable" COVID-19 fatalities by country, Virginia only lists total fatalities by county on its master spreadsheet (the one I use) whereas some one the secondary data sets (which I don't use) notes "confirmed" and "probable" fatalities on a state-wide level, while DC simply lists total fatalities. Presumably DC's total fatality count includes both the confirmed and probable ones.
The way I reconcile these three reporting methods is to simply use the top-level data sets from each jurisdiction - I have neither the time nor desire to play games with the "probable" cases, and if the departments of health have reason to believe that a death is "probably" due to COVID-19, I'm not inclined to challenge that.
However, you may notice the occasional case where a fatality count actually falls from one day to another. For instance, Arlington had 132 recorded fatalities on 2 July, then falling to 131 on 3 July, and then once again 132 on 4 July, and yet once more falling to 131 on 5 July. In the absence of people being resurrected from the dead, what I think is happening is that fatalities are being reclassified, causing in some cases numbers to fall. Even in cases where fatalities rise, there might be downward revisions which are "masked" in a larger general rise.
Take this example for "masking": 100 COVID-19 fatalities on Monday, 90 confirmed and 10 probable. On Tuesday, there are 10 new fatalities, 9 confirmed and 1 probable. However, two of the original 10 probables turned out not to be COVID, hence you now have 108: (90+10-2)+(9+1).
When do cases and deaths hit zero?
No clue - I try to stay out of the prediction business here.
When Can our Region Open Up?
This gets far beyond my level of expertise or competency. All I can say is that the statistics suggest that:
1) The worst is past us
2) The regional measures may have worked
3) Nothing precludes additional waves/outbreaks
Speaking only for myself, I take this as a sign that the regional measures (masks and lockdowns) have been working, and that sustained efforts can help snuff out the regional outbreaks. You are welcome to disagree with this assessment (tastefully, of course).
Now the analysis:
Newly-confirmed COVID-19 cases and associated fatalities in the DC region have been growing at a decreasing rate for the past month or so. While I had previously reported about a regional peak in early-mid April, that clearly had proven to be a mistaken assessment on my part. With that in mind, I would like to present the data, with a focus on the daily growth of both cases and fatalities.
Percentage increase of cases and fatalities has continues to decline
Generally speaking, this means that concerns about exponential growth within our region has passed. For instance, a few weeks back I stopped calculating the doubling time, because it's pretty much lost its relevance.
You may note that I often use the notation of "<1%" growth rate on the daily report. This simply means that the true percentage growth rate is below 0.5% (my spreadsheet rounds to the nearest integer percentage point for the calculation). Even this statistic has really started to become moot, as the daily growth percentage for both new cases and fatalities is increasingly less than 1%.
For further detail on the decreasing percentage increases, I recommend visiting my recently-published June 2020 summary statistics update.
New cases and fatalities continue to decline
In theory, you could have a leveling off of the percentage rate (i.e. no more exponential growth), but remain flat-lined at a constant absolute daily increase (e.g. 500 new cases daily for the next 30 days). However, the raw count of both new cases and deaths continues to decline. For this, I use a seven-day moving average to help smooth out the irregularities associated with each day of the week (in particular, an apparent "lull" on weekends and holidays followed by "overcompensation" the next 1-2 days), and any other reporting issues.
With this, we see new cases having a very modest bimodal distribution with apparent peaks in late and early May (6 May, 1,304; 21 May, 1,309).
For fatalities, we observed a single apparent peak in early May (3 May, 64).
For both new cases and fatalities, the past few weeks have seen a sharp decline followed by a leveling-off of sorts. To put it in perspective, the average amount of daily new cases (~400) and fatalities (~10) for the past few weeks are both right around the same amount we were observing in early April. Presumably, as both figures continue to decline, we would expect to soon see number similar to those from late March, etc.
What About Testing?
I haven't really tracked testing capacity because I don't have the time or desire to do so. However, it appears that the decrease in newly-confirmed cases is not a function of decreased testing - while testing hasn't been increasing, it also hasn't fallen much since mid-May (note on this chart - it only includes DC and Northern VA testing, because I can't get longitudinal testing data for Maryland. I think the trend is sufficient, however).
Obviously fatalities are not contingent on testing, which makes them a useful statistic (albeit a bit lagging). However, the presence of a decline in new fatalities also is a positive sign.
What's With the Occasional Falling Fatality Counts?
Let's start with a discussion on methodology on calculating fatalities (one which I've been meaning to have for a while).
The three jurisdictions report fatalities different. Maryland lists "confirmed" and "probable" COVID-19 fatalities by country, Virginia only lists total fatalities by county on its master spreadsheet (the one I use) whereas some one the secondary data sets (which I don't use) notes "confirmed" and "probable" fatalities on a state-wide level, while DC simply lists total fatalities. Presumably DC's total fatality count includes both the confirmed and probable ones.
The way I reconcile these three reporting methods is to simply use the top-level data sets from each jurisdiction - I have neither the time nor desire to play games with the "probable" cases, and if the departments of health have reason to believe that a death is "probably" due to COVID-19, I'm not inclined to challenge that.
However, you may notice the occasional case where a fatality count actually falls from one day to another. For instance, Arlington had 132 recorded fatalities on 2 July, then falling to 131 on 3 July, and then once again 132 on 4 July, and yet once more falling to 131 on 5 July. In the absence of people being resurrected from the dead, what I think is happening is that fatalities are being reclassified, causing in some cases numbers to fall. Even in cases where fatalities rise, there might be downward revisions which are "masked" in a larger general rise.
Take this example for "masking": 100 COVID-19 fatalities on Monday, 90 confirmed and 10 probable. On Tuesday, there are 10 new fatalities, 9 confirmed and 1 probable. However, two of the original 10 probables turned out not to be COVID, hence you now have 108: (90+10-2)+(9+1).
When do cases and deaths hit zero?
No clue - I try to stay out of the prediction business here.
When Can our Region Open Up?
This gets far beyond my level of expertise or competency. All I can say is that the statistics suggest that:
1) The worst is past us
2) The regional measures may have worked
3) Nothing precludes additional waves/outbreaks
Speaking only for myself, I take this as a sign that the regional measures (masks and lockdowns) have been working, and that sustained efforts can help snuff out the regional outbreaks. You are welcome to disagree with this assessment (tastefully, of course).
Comments
Post a Comment