10/5/23: Great—with Covid cases rising in the state, the Mass. DPH has STOPPED REPORTING the hospital data necessary to produce these positivity estimates. Unless they resume reporting, the wastewater data below is the only way to see what's happening with Covid in the state.
9/28/23: After starting to drop last week, Covid positivity in both Massachusetts and New York has begun rising again slightly this week. Covid in wastewater (below) is back to its previous steady rise following a one-week anomalous drop. Currently, if you gather 23 people together at random, there's a 50/50 chance that at least one of them will be Covid-positive.
Covid tracking for New York available here.
Tracking Covid In the Wake of Reduced Testing
At-large Covid testing has declined so much that published cases-per-100,000 figures have become practically meaningless: they show a huge decline in Covid which is not confirmed by other methods of measuring the extent of infection, such as tracking Covid DNA levels in sewage wastewater and examining the overall rate of deaths. This essay presents another method of estimating Covid prevalence, using hospital data.
On January 10, 2022, the Massachusetts Department of Public Health began reporting the number of "Primary Covid" hospital patients (patients who have been administered dexamethasone, a treatment for moderate to severe Covid) in addition to the total number of patients who test positive for Covid. This makes it possible (in principle, at least) to distinguish between people who are in the hospital because of Covid versus people who were simply discovered to be Covid-positive when they were hospitalized for other reasons. Here's a graph of that data:
Besides obvious things like the chart above, this data can be used to produce a surprising result: an estimate of the percentage of the population who would test positive for Covid if everyone were tested. Here is that data, starting from the very high peak in January 2022 (which is trimmed off in the topmost chart):
The trick for generating this result is to look at all patients who are NOT being treated for Covid and calculate the percentage of them that are Covid-positive. (See Note [1] below.) The not-being-treated group is a somewhat random sample of the population which is subjected to 100% testing. (Or used to be; see below.) Although the hospital population is not a proper representative sampling (it probably skews old, for example), it's the broadest sample I know of with 100% testing.
A sample like this is useful because it measures Covid positivity in a way that doesn't depend on what's going on with testing in the population at large. With ordinary testing, if you do more testing you find more cases and the rate (cases per 100,000 of population) seems higher. Do less testing, and the rate seems lower. Fewer and fewer PCR-type tests are being performed these days, and with the advent of at-home testing, plenty of positive cases are not being reported to the DPH and don't appear in any statistics. Tracking hospital cases solves these problems.
At-large testing decline
The following chart shows the number of PCR (aka "molecular") Covid tests administered in Massachusetts from the start of the pandemic to present:
Large numbers of tests were administered during the Delta and Omicron surges, around 1/1/2021 and 1/1/2022, but PCR testing has declined steadily since then—which is not surprising, given the cost of testing, public weariness and broad availability of home test kits.
However, as a result of decreased testing, "official" Covid case counts have plummeted. (If you don't test, you don't find cases.) Here's the data since January 2022, showing both the hospital-estimated positivity levels (in blue) and the official two-week Covid case count (in red), expressed as a percentage of the state population:
Looking at the red line, it's no wonder people scoff at Covid cautions and think the pandemic is over. But those low numbers don't agree with other methods of measuring Covid prevalence, methods that don't depend on at-large PCR testing.
For example, here is Biobot Analytic's record of Covid DNA levels found in Massachusetts sewage wastewater since January 2022: [3]
Note that the wastewater curve looks a lot like the hospital-estimated curve, not the dwindling case-count curve. Similar wiggles, and same slight upward trend throughout 2022. (For Hampshire County data alone, see [4] below.)
Another method of tracking Covid is to look at death rates. The chart below is the weekly death rate (from all causes) in Massachusetts for 2015-2019, the five years leading up to the pandemic. It shows the normal seasonal variation in deaths (higher in the winter), and typical variations from the norm:
The gray dots are the weekly figures, the blue line is a smoothed running average, and the red line is a seasonal average for 2015-2019, repeated annually.
Here is the data for the years 2020 to present. The elevation of the blue line above the seasonal-average red line shows the effect of the pandemic:
The first peak, in May 2020, is from the original "Alpha" version of Covid. The second peak, in January 2021, is from the Delta variant, and the third peak, in January 2022, is from Omicron (BA.1). There was no peak in January 2023, but deaths ran 8.4% higher than normal from May 2022 through January 2023—which, for perspective, is people dying at 13 times the traffic fatality rate [5].
It is worth noting that the death rate has returned to seasonal normal levels three times during the pandemic: Once for three months starting 7/1/20, a second time for four months starting about 3/1/21, and a third time for two months starting about 3/1/22. If you locate these periods in the Biobot chart below, they all correspond to periods of low Covid levels in wastewater. Absent Covid, death rates return to normal. It's not something we're doing differently, or vaccines, that are causing elevated deaths, it's Covid.
We are currently in a fourth period of normal death rate, which began in February 2023.
The following table tallies the excess deaths for the three peaks and the milder wave following Omicron:
Are these hospital positivity estimates right?
I mean, really?? 20% of Massachusetts residents Covid-positive in January 2022? On Jan. 15, 2022, the official daily new-case rate was 279 per 100,000 of population (which works out to 0.279%), and the two-week total case count stood at 271,940—which is 3.9% of the population, not 20%. What gives?
I suspect the difference is due to insufficient testing, and that if more and more people had been tested, the rate would have kept climbing until it reached 20%. (And I suspect that more people weren't tested because of something that becomes apparent only when the higher rates are used in other calculations.) However, without a proper representative sample of the population being subjected to 100% testing, it's not possible to know if the hospital estimates are correct. Possible sources of error include:
- If hospitals are no longer testing every patient for Covid—for example, if they're only testing patients who show symptoms of
Covid—then all bets are off: the true statewide positivity level will be higher than the level shown in the graphs here by an unknowable amount.
(A paper published in December 2022 recommended that hospitals stop testing asymptomatic patients.)
Sudden irregularities in some of the charts below suggest that this may have started happening here in Massachusetts in June 2023. This
would explain the growing discrepancy between estimated positivity levels and Covid levels in wastewater.
- Treatment with dexamethasone may not be an accurate indication of which patients are hospitalized for Covid. Some patients may be being treated
with other drugs and incorrectly excluded from the Primary Covid group, a misclassification which inflates positivity estimates. Data from New York
suggests that Massachusetts is in fact undercounting primary Covid cases; see Note [6] below.
- Maybe the population at large is healthier than people who wind up in the hospital (perhaps because they tend to be younger). This would cause the hospital-estimated rates to be higher than the true rates by an unknown amount.
- On the other hand, people who are about to go into a hospital for elective procedures may get tested for Covid beforehand and postpone their visit if they test positive. This would cause the hospital-estimated rates to be lower than the true rates by an unknown amount.
Except for time shifts (which are curious), the hospital estimates are always below the percent-positive test rates, which is as expected. The majority of people who get tested for Covid (outside hospital) are feeling unwell in the first place, and as a result the percent-positive test rate is generally higher than the actual percentage of the population who are positive. This chart shows that the hospital figures are at least not a gross overestimate of the actual positivity rates.
[The red line above reaching 12% by September 2023 looks alarming, but what it means, in large part, is that 12% of the people who feel sick enough to wonder if they have Covid are in fact Covid-positive. When very little at-large testing is being done, it doesn't take much Covid in the population to produce an alarming-looking spike in positive tests.]
Why do these numbers matter?
For one thing, the Covid positivity rate affects how willing people are to mask up in public—still the most effective way of reducing the spread of the virus. In January 2023, with an official two-week case count of 18,741, it seemed like only 0.27% of the state—one person out of 370—was infected. Given this information, wearing a mask seems like Henny-Penny unnecessary overcaution. The hospital data, though, suggests that nearly 10% of the population was Covid positive at that time. At this level, if you gathered 7 random people together, there would be a 50-50 chance that at least one of them was Covid positive. Given this information, people might be more likely to wear a mask in public, thereby protecting both themselves and other people around them.
For another thing, knowing the actual number of Covid cases is necessary to be able to judge how dangerous the virus really is. The number of cases is the divisor in calculating the percentage of cases hospitalized for Covid and the percent who die from Covid-related causes. If that divisor is much smaller than it should be, Covid winds up calculating out as more dangerous than it really is. I think people have rough gut feelings for how many people around them are dying of Covid or suffering long-term symptoms, and if the official percentages on these things are greatly inflated, people wind up disbelieving what authorities say about Covid.
For the good of the public, it's important to get the numbers right.
How serious is Covid?
Using the hospital positivity estimates, the percent of Covid-positive individuals who are being treated in hospitals for Covid can be calculated. The numbers are surprisingly low: about 0.12% in January 2022—roughly 1 out of every 800 Covid-positives—and falling to about 0.06% by May: about 1 out of every 1600 Covid-positives. (Though see [6] below; the true rate is probably about 0.09%, 1 out of every 1100 positives.) Beyond the January 2022 Omicron surge, the numbers are remarkably stable despite substantial variations in the positivity rate (until things became irregular in June 2023, possibly from reduced hospital testing).
These surprisingly low numbers may point to why testing-at-large substantially underestimates true Covid positivity levels: Very few people get so ill from Covid that they need to be hospitalized, and I wonder if the majority of people who are Covid positive don't get sick enough even to wonder if they have Covid and go for a test.
Although your chances of being hospitalized for Covid are quite low, an estimated "10% of Covid cases" are said to wind up with Long Covid symptoms, which can be debilitating and last for years. Read here about Long Covid:
"Long Covid" is not just having a long-lasting bout of Covid; it's ongoing problems in the wake of a Covid infection and can occur even without a person becoming severely ill in the first place. Covid can damage a person's immune system, their heart, lungs, vascular system, neurological system and other organs. The consequences may be lifelong. Despite the ghastly number of deaths from Covid, Long Covid may turn out to be the most problematic result of the pandemic.
However, I suspect that the 10% figure means 10% of "official confirmed" Covid cases. In January 2022, the peak official two-week case count for Massachusetts was 271,940 cases, and 10% of that number is 27,194 people winding up with Long Covid. The problem is that I think the official confirmed rate is a substantial undercounting of actual Covid positivity in the state, and that more like 1,450,000 people were Covid positive during the Jan. 2022 peak. If there are 27,194 Long Covid cases out of 1,450,000 positives, that's a Long Covid rate of about 1.8%, not 10%—about 1 in 54 Covid positives. Which makes developing Long Covid about 33 times as likely as being hospitalized for Covid.
The bottom line is, the long-term consequences of Covid are why you should still Not Want To Get Covid. By default, put on a mask when you are indoors in public! Think before you join a group unmasked: consciously go, okay, with this group I am going to take a chance at getting infected. People are casually going without masks and acting like the pandemic is over because the official case-count numbers have dwindled to nothing—but Covid is far from done with us yet.
Notes
[1] Primary Covid patients also contribute to the statewide positivity rate, but for them the correct percentage divisor is the population of the state: about 7 million. By late August 2022, there were roughly 200 people being treated for Covid in Massachusetts hospitals; dividing 200 by 7 million gives a rate of about 0.003%—an insignificant addition to the nearly 5% positivity rate at the time. Even during the January 2022 peak, when there were almost 1700 primary Covid patients, they added only 0.02% to the roughly 20% overall positivity rate.
[3] The statewide Massachusetts wastewater chart was produced by aggregating county-level data from Biobot's wastewater_by_county data set (which can be found here, with recent updates coming from here). Individual county concentration levels were weighted by county population and averaged. (The 300,000-person cap on population size that Biobot uses in its aggregations was not employed here.)
[4] The following chart shows Biobot's wastewater Covid levels for Hampshire County (red line and gray dots), with the blue line giving the statewide average:
There are five colleges within Hampshire County, including UMass Amherst with 28,000 students. The influx of students probably causes the peaks in September and February, and their departure by June the low numbers during the summer months. (The May 2022 peak is a mystery; it doesn't seem to be caused by students returning from spring break, because that break occurs in mid-March.)
[5] If the 8.4% higher death rate had kept up for a full year, it would have resulted in 4970 excess deaths. The average traffic fatality rate in Massachusetts for the years 2015-2022 is 370 deaths per year; 4970/370 = 13.4.
[6] New York state also distinguishes between "primary" and "incidental" Covid patients, but in New York the criterion is a doctor's judgment call: "admitted due to COVID or complications of COVID" versus "admitted where COVID was not included as one of the reasons for admission." The Massachusetts criterion of "administered dexamethasone," while objective, possibly misclassifies patients who are being treated for Covid with other drugs. The following graphs show the percentage of Covid-positive patients who have been classified as primary in each state:
The two states being close together, there would appear to be no reason why they should have significantly different percentages of primary Covid patients. From the data above, it appears that Massachusetts' dexamethasone criterion is significantly undercounting primary Covid patients. Scaling the number of primary Covid patients in Massachusetts so their moving average is at least 42% of all positives (as it is in New York) reduces the statewide positivity estimates by about 18%, as shown in the following chart:
The red line above is my best estimate of the true Covid positivity level in the state (at least up to about June 2023, at which point I wonder if hospitals stopped testing 100% of patients for Covid). Scaling the primary Covid numbers also boosts the percentage of positives requiring hospital treatment from 0.06% to about 0.09%, close to the New York figure of 0.10%.
