Nigel Parry/Lowertown.info—The Minnesota Department of Health’s changes to counting positive COVID-19 cases by date of sample—not by date of reporting—revealed that between March and May 2020 we had a profound blind spot in case awareness, sometimes resulting in a daily discrepancy of thousands between the number of reported and actual cases—all as the state was working around the clock to formulate policy informed by the best possible scientific advice.

The Minnesota Department of Health first began posting its COVID-19 data online in its dedicated Situation Update for Coronavirus Disease 2019 (COVID-19) section, around the time of Minnesota’s first positive case was reported, on March 6th.

As the pandemic has unfolded, the Minnesota Department of Health has steadily improved its COVID-19 data collection and online data presentation as new systems have been put in place and new data has become available. Prior to March 29th, private testing data was not included in Minnesota’s ‘total tests taken’ metric. As time has progressed, the department has added new visualizations of data and new access to chart data in table format, increasing both useability and transparency of the state’s datasets.

On May 13th, just two months after the Minnesota Department of Health first began offering daily-updated COVID-19 statistics on its website, the agency changed the way it counts positive cases. Previously, it logged the date of a new positive case as the day it was first reported. The new text on their website said: “As of 5/13/2020, total case data is represented by specimen collection date. Cases over time were previously based on report date.”

Previously available positive case data had been removed, existing tables and charts retroactively updated back to early March to use only data from the new “date of sample” system.

Of course, information ‘gathered closer to the virus’ provides a better overall picture of viral spread and provides a medical—not administrative—date stamp for contact tracers. The change is obviously a sensible one.

The chart we had for Positive Cases prior to the change looked like this:

In its May 12th, 2020 media update, the Minnesota Department of Health explained the coming change in depth:

MDH will also update information they provide on specimen collection date instead of reported date. So what that means is, for their graph that shows the total positive cases over time, the MDH has been displaying this graph based on the date that the case was confirmed at MDH. They’re going to make a change, and that graph will now show cases by the date their specimen was collected. That’s the specimen that was tested and determined that they were positive for COVID-19. Using the date that a specimen was collected is a measure that’s more relevant to each case, and is in line with how other disease data is reported. It gives the MDH a better picture of what’s happening in Minnesota over time. Reporting cases this way removes the factor of reporting delays. In other words, for some out of state laboratories, it can take up to a week to get results. So it removes the factor of reporting delays and investigation timing, and more accurately reports when people in Minnesota have COVID. This does not change the MDH’s total case count, just the day that the case shows up on the graph.

I wondered how the two differing methods of counting positive cases had impacted the overall picture we had all had during those first two months. How did the plumbline we were all working from differ from the actual reality of the virus’ spread at the time?

The data previously displayed on the MN Department of Health’s website was gone with no obvious archive available. Fortunately, Lowertown.info has been archiving the raw data from the daily MN Department of Health’s COVID-19 Situation Update into its own visualizations, so both new and old data sets were available for comparison. Here are both visualized together—the red line shows cases with the new sample date method, the blue line cases logged under the former reporting date method:

“Just the day”

The MN Dept. of Health’s May 12th briefing reassured us that the new change in the way positive cases are counted “does not change the MDH’s total case count, just the day that the case shows up on the graph.”

Maybe that’s true in retrospect, if we are looking back at the whole period as a past event, but what effect did that reporting delay have on the number of positive cases that we were aware of at any one point in time?

A lot gets lost with the camera zoomed this far out.

Zooming in on March, April, and May

Here are the discrepancies between the old and new positive cases counts, visualized for the months of March, April, and May so far. Zoomed in, you can see that the old method was way behind the curve, vastly underestimating new daily cases .

The visualizations have tooltips if you hover over (desktop) or click on the lines (mobile). Explore the differences on different dates.

Notes:

• March 15th data shows 35 positive cases when counted by date of report and 125 cases when counted by date of sample, a difference of 90 cases and reporting of just 28% of the actual total at the time.
• March 30th data shows 576 positive cases when counted by date of report and 819 cases when counted by date of sample, a difference of 243 cases and reporting of just 70% of the actual total at the time.

 

Notes:

• April 7th data shows 1,069 positive cases when counted by date of report and 1,430 cases when counted by date of sample, a difference of 361 cases and reporting of just 75% of the actual total at the time.
• April 15th data shows 1,809 positive cases when counted by date of report and 2,319 cases when counted by date of sample, a difference of 510 cases and reporting of just 78% of the actual total at the time.
• April 23rd data shows 2,942 positive cases when counted by date of report and 4,264 cases when counted by date of sample, a difference of 1,322 cases and reporting of just 69% of the actual total at the time.
• April 30th data shows 5,136 positive cases when counted by date of report and 8,075 cases when counted by date of sample, a difference of 2,939 cases and reporting of just 64% of the actual total at the time.

 

Note:

• May 6th data shows 8,579 positive cases when counted by date of report and 11,389 cases when counted by date of sample, a difference of 2,810 cases and reporting of just 75% of the actual total at the time.

The two data sets for positive cases (by reporting and sample dates) are found at the bottom of this page in table form.

Daily undercounts of thousands of cases

The Minnesota Department of Health’s changes to counting positive COVID-19 cases by date of sample—not by date of reporting—revealed that between March and May 2020 we had a profound blind spot in case awareness, sometimes resulting in a daily discrepancy of thousands between the number of reported and actual cases—all as the state was working around the clock to formulate policy informed by the best possible scientific advice.

Starting March 17th, Minnesota Governor Tim Walz ordered all restaurants, bars, and cafes closed. On the day of the shutdown, the number of reported positive cases was just 60. In reality, the actual total was 242, four times the number we were all working off. Would the governor have acted sooner if he had known that the number of positive cases was four times higher than we all thought they were at the time? Even days make a difference.

This is not nitpicking. In an April 3rd article in the New York Times, Cities That Went All In on Social Distancing in 1918 Emerged Stronger for It, the authors report how Minneapolis and Saint Paul fared relatively during the 1918 Influenza Epidemic. Saint Paul implemented its shutdown one week after Minneapolis. It saw 481 deaths per 100,000 compared with Minneapolis’ 388, one-and-a-quarter times higher. In terms of employment gains, Minneapolis saw double that of Saint Paul. It’s hard to argue with such a painfully relevant lesson from history that is literally this close to home.

From early March to mid-May, those responsible for Minnesotan public health policy were making decisions based on a significant underestimate of the number of positive COVID-19 cases in the state.  Between March and May, the number of daily reported positive tests lagged 4-5 days behind the reality. Sometimes the daily discrepancy was as high as 3,000.

On April 9th, Minnesota media reported that “Gov. Tim Walz is extending a statewide stay-at-home order to May 4 to push the peak of the COVID-19 pandemic into the summer.” On April 9th, as far as we all were concerned, there were just 1,242 positive cases in the state. The correct number was 1,635—a difference of 393 or one-and-a-third times as many positive cases.

By the end of April, presumably around the time that Governor Walz was again making decisions about whether to extend the “Stay at Home” order in the face of political pressure, the undercount of positive cases was as high as 3,000, two-thirds of the actual number of cases. April 17th and May 2nd saw two “Reopen Minnesota” protests, the first outside the governor’s residence on Summit Avenue, the second around the State Capitol. Would knowing that we only were seeing a percentage of the actual cases have given pause to those gatherings?

When did it all even begin in Minnesota? March 6th had become the accepted milestone of the first Minnesotan positive case. The new “date by sample taken” data reveals that the first positive test was in fact conducted the day before, on March 5th.

As of May 13th and ongoing, Minnesota Department of Health positive case data is being both retroactively and daily presented by date of sample, not date of report. Lowertown.info will likely be reworking its various visualizations to preserve and reflect the data showing the first three months’ underreporting. From the looks of today’s positive chart on the Minnesota Department of Health’s website, the state is in the process of doing the same.

 

Will the new “date of sample” data still be compiled several days after the fact? It looks like it from the gray area displayed in the new MN Dept. of Health charts (pictured above). This short YouTube video explains the phenomenon of “data lag” and how that affects perceptions of curve peaking and flattening:

Minnesota COVID-19 Positive Results by Reporting Date & Sample Date

Date	Positive (by reporting date)	Positive (by sample date)
Mar 5	0	1
Mar 6	1	1
Mar 7	1	2
Mar 8	1	2
Mar 9	2	5
Mar 10	3	11
Mar 11	5	21
Mar 12	5	43
Mar 13	14	62
Mar 14	21	87
Mar 15	35	125
Mar 16	54	175
Mar 17	60	242
Mar 18	77	265
Mar 19	89	283
Mar 20	115	299
Mar 21	138	323
Mar 22	169	345
Mar 23	235	389
Mar 24	262	299
Mar 25	287	323
Mar 26	346	562
Mar 27	398	636
Mar 28	441	683
Mar 29	503	721
Mar 30	576	819
Mar 31	629	912
Apr 1	689	969
Apr 2	742	1053
Apr 3	789	1116
Apr 4	865	1156
Apr 5	935	1215
Apr 6	986	1328
Apr 7	1069	1430
Apr 8	1154	1532
Apr 9	1242	1635
Apr 10	1336	1730
Apr 11	1427	1804
Apr 12	1621	1862
Apr 13	1650	2010
Apr 14	1695	2163
Apr 15	1809	2319
Apr 16	1912	2484
Apr 17	2071	2666
Apr 18	2213	2784
Apr 19	2356	2910
Apr 20	2470	3188
Apr 21	2567	3449
Apr 22	2721	3753
Apr 23	2942	4264
Apr 24	3185	4776
Apr 25	3446	5115
Apr 26	3602	5499
Apr 27	3816	6161
Apr 28	4181	6751
Apr 29	4644	7429
Apr 30	5136	8075
May 1	5730	8689
May 2	6228	9016
May 3	6663	9349
May 4	7234	10158
May 5	7851	10788
May 6	8579	11389
May 7	9365	12181
May 8	10088	12809
May 9	10790	13073
May 10	11271	13244
May 11	11799	13839
May 12	12494	14122
May 13	14230	14230

Related Links

• Ramsey County’s “COVID-19 Situation Update” dashboard
• Minnesota Department of Health’s Situation Update for Coronavirus Disease 2019 (COVID-19) section
• Lowertown.info’s Minnesota COVID-19 Statistics & Charts page