Covid-19 and an Unexplained Rise in "Unclassified" Deaths

September 5, 2020

Updates Sept. 15

Sept. 15 note: This post takes a valuable look at likely covid deaths in different categories, but the title subject ("unclassified" deaths) is less clear than it seemed. It is new - the CDC's excess deaths page doesn't include this.  "... causes of death where the underlying cause was unknown or ill-specified (i.e. R-codes) were excluded (except for R09.2, which is included under the Respiratory diseases category). Counts of deaths with unknown cause are typically substantially higher in provisional data, as many records are initially submitted without a specific cause of death and are then updated when more information becomes available (4)." Indeed - I looked into that after I noticed that category rises so sharply over time in all places, and upon a weekly update I noted the numbers had shrunk slightly, while some other classes had grown a bit. But it seems some numbers a ways back are settled, only getting an occasional 1 or 2 removed and mostly staying the same, and still remain well over average in an interesting pattern. I think the later weeks' data will do the same, but I'll have to try for a clearer view of the scale after the dust settled a bit better. The next update should be soon, and I'll check again a week later. (end note).

---

A study published by JAMA (the Journal of the American Medical Association) a while back looked at increased death rates for specific ailments - aside from Covid-19 - during the pandemic, in five hard-hit US states: Massachusetts, Michigan, New Jersey, New York (city, state aside from city), and Pennsylvania. This found "Between March 1, 2020, and April 25, 2020, a total of 505,059 deaths were reported in the US; 87,001 (95% CI, 86,578-87,423) were excess deaths, of which 56,246 (65%) were attributed to COVID-19" while about 1/3 of this huge toll were seemingly caused by something else worth some attention. 

The 5 states with the most COVID-19 deaths experienced large proportional increases in deaths due to nonrespiratory underlying causes, including diabetes (96%), heart diseases (89%), Alzheimer disease (64%), and cerebrovascular diseases (35%) (Figure)." ... "New York City experienced the largest increases in nonrespiratory deaths, notably those due to heart disease (398%) and diabetes (356%)." Of course they also suffered the sharpest and deadliest outbreak of the virus. Coincidence?

https://jamanetwork.com/journals/jama/fullarticle/2768086

"... Not Elsewhere Classified"

I believe they use this data source, or similar:

https://data.cdc.gov/NCHS/Weekly-Counts-of-Deaths-by-State-and-Select-Causes/muzy-jte6/data

I had my own look at this, seeing several things appear to rise, but noting what I label "unclassified" is the big rise. This is the column just to the left of diseases of the heart, headed "Symptoms, signs and abnormal clinical and laboratory findings, not elsewhere classified (R00-R99)." Another source gives the R00-R-99 code as signifying "symptoms, signs, abnormal results of clinical or other investigative procedures, and ill-defined condition/s regarding which no diagnosis classifiable elsewhere is recorded." This also breaks down the definition so:

a) cases for which no more specific diagnosis can be made even after all the facts bearing on the case have been investigated;

(b) signs or symptoms existing at the time of initial encounter that proved to be transient and whose causes could not be determined;

(c) provisional diagnosis in a patient who failed to return for further investigation or care;

(d) cases referred elsewhere for investigation or treatment before the diagnosis was made;

(e) cases in which a more precise diagnosis was not available for any other reason;

(f) certain symptoms, for which supplementary information is provided, that represent important problems in medical care in their own right.

In context, what is it? Maybe just semantics? From the patterns I see, it may be other deaths were wrongly classed early on as heart disease, Alzheimer's, etc. and later were increasingly shifted to this unclear category? See below how the others were higher and then decline, just as the line for this (usually green) rises dramatically. If so, that's not mainly from more of it happening, just more of the same that's been happening getting lumped under this heading - which might become a new proper heading linking it to likely unknown Covid-19 infections. I suspect the vast bulk of these are just that, causing damage that proves fatal anywhere from a couple of days to a few months (so far) after infection. 

In most states, unclassified R00-R99 was a small number on a weekly basis in 2019. In smaller states, it's only noted in sporadic batches - frequently 12 added every few weeks to couple months (see below, Iowa). I didn't look very deeply into the patterns like seasonality, aside from US total, which showed the number holding pretty steady throughout 2019. So I guess it should start about normal in each of these places in 2019 and when we check in here at the end of January in Massachusetts, Michigan, New Jersey, New York City, New York state (aside from the city), and in Pennsylvania. It should hold steady, or if it changes, more likely it would fall a tiny bit. But here's what happens during the same SARS 2 / Covid-19 pandemic:  

Notes: with these and others, the final data point seems incomplete - things are often dropping then, but probably not to the degree shown. Somewhat crude plotting here, as usual - I think I missed a week for NYC so it end at the right level but too soon - so from some point, the data shifts right one week. The numbers are also included right here to compare. Curves vs. angles: curves just used mainly at sharp changes or to help visibility of other lines, because too big a pain to do at all spots.

This and the plottings below aren't authoritative, just to show the basic trend. 

Pennsylvania's excess deaths analysis bears repeating here: I've found the state was hit harder with Covid-19and/or likely-related "excess deaths" than most realized, even for a while half-listing a death toll close to what I suspect - nearly twice the level otherwise reported. As noted at lower right of this graphic I did recently, some other class of deaths - non-Covid as far as we know - must be well on the rise. The CDC data and JAMA study help explain why these other deaths rise - most likely they're NOT other.  Those shown in green above come to some 150-200/week for some some of these weeks that are about 500-900 deaths above average - that's around 25% of the above-average, and more than the confirmed covid deaths, which were continuing at between 95 and 125/week.

And this same mysterious syndrome or class of might be what's killing so many confirmed covid victims in England past their new 28-day cutoff. Some have been killed by flying debris or anything, before or after the cut-off, and get counter anyway or not, depending on the date. But by the numbers, it's far more who die pretty swiftly from the virus' damage, but fail to be counted because it was just a bit too slow. See: https://libyancivilwar.blogspot.com/2020/08/on-calculating-true-uk-covid-19-death.html

A Wider, National View

After finding this data source and seeing what JAMA already covered, I considered what other states might be worth a look.

Connecticut (pop. 3.6m): I had noticed Connecticut was especially hard-hit per capita, close to on par with New York and New Jersey (just confirmed Covid deaths there add up to 0.124% of the populace killed so far, most of those before the end of May). A look at trends there  shows a similar pattern - as I plotted it:

Here, as with the rest, looking at more, including:

* Diseases of the heart - moderate correlation in rising along with Covid deaths

* Diabetes - moderate to little-no correlation

* Alzheimer's - strong correlation 

* Cerebrovascular disease - JAMA study notes it, but it didn't seem clear in some cases I checked

* Influenza - complicated - actual transmission rates may be down, deaths should be, but often rise - possibly including dual infections (where flu follows covid and then kills) or undiagnosed / mis-categorized covid like many of these others might be.

The rises in those are clear enough, but the odd pattern again is with unclassified - whatever this means, it rises so dramatically to hold a solid presence above 40/week for 6 weeks, hitting 50 3 times, so that by week 28 (July 5-11) it accounts for more than 10% of all deaths statewide. 

US totals: clear rise in heart disease (2019 compared), Alzheimer's, diabetes, and also influenza (2019 trend compared).  That's quite a few likely covid deaths mischaracterized. And for the green line - even with less hard-hit states pulling the average down, the unclassified kind of death at least rises massively, from a starting baseline around 800/week to well over 3,000/week nationwide. I checked this category for 2019 and found it was steady all the same 31 months of that year: a stray 523/week at the low end, 648/650 at the high end, and otherwise all tallies were close to 600/week (range otherwise: 563-635), as shown here, increasingly divergent. Note how influenza deaths rose far above normal during the peak, then fell quickly to normal seasonal levels, just as the unclassified deaths were increasing far above their norm (which should be on par with flu).

For reference, unclassified deaths so far since week 4 in 2020: 785  - 771 - 788 - 805 - 841 - 932 - 917 - 900 - 876 - 905 - 994 - 994 (repeated) - 1016 - 1091 - 1065 - 1230 - 1229 - 1378 - 1525 - 1624 - 1790 - 1953 - 2222  -2441 - 2764 - 2928 - 3079 - 3263 - 3381 - 3224 - 2551. Total 46,045 deaths. Some of these will be the same kind as last year, and let's say it's higher than in 2019, as seen at week 4-6, 770-790/week. Average 780 x 31 weeks = 24,180. Above that line is another roughly 22,000 deaths most likely caused by unknown Covid-19 infections. This will add to the confirmed deaths (now over 191,000), and the other likely unconfirmed that were listed under heart disease, Alzheimer's, etc. This comes in the last 6 months, with the bulk of it in the last third. This is likely to continue at similar rates if not higher, depending just what the hell it is and how much more of it happens. 

Comparing to historic averages: I didn't do much of that here, except for good measure with heart disease in most cases, and with unclassified, influenza, and all-causes deaths for the US total (all just compared to provided 2019 tallies - no broader averaging). Take any of these basic 2019 trends with the basic idea of seasonal death rates; they stay about the same day-to-day, only rising with extra illness deaths in the winter, peaking around January-February about 10% higher than usual levels in the summer. As shown with US, in 2019 it was around 5,800 all-causes deaths per week in winter, and falls to about 51,000 in summer. 

I was a bit surprised to see nearly all these specific causes follow the same pattern, some more clearly than others - a bit more deadly in the winter, and less so in the summer, absent a bad heat wave. So most of these will start near peak in late January, and should fall slightly all the way across on average (there are a lot of week-to week ups and downs, likely caused more by reporting variance than actual death rates. True differences we can see will be trends playing out over at least 2-3 weeks).

Three more sample states going into that national average:

Florida (pop. 21.7m) seemed worth a look for several reasons. Only a slight increase may be notable for Alzheimer's and diabetes, but heart disease is clear in pushing a bit below, then well above the expected norm, at just the right time to provide a likely answer to that mystery. It's just now dipping back to seasonal norms. Green unclassified holds the usual pattern, seeming to increase over time and/or pick up the slack of other decreasing categories, apparently dropping a bit now from a high around 400/week. (note marked drops in last data had applied to different weeks than when I started - these are updated - week 34 will soon say something higher and plausible at such points, as week 33 now does).

Washington (my state, pop. 7.5m) - early deaths, then fairly well controlled. Early lockdown measures: schools closed March 12, general stay-at-home order 3-23, etc. It doesn't seem it ever took off enough to make these patterns very clear. Heart disease seems to have its yearly decline interrupted by upward trends, shortly before the peak of covid deaths. It's hard to tell if Alzheimer's rises in February and in March reflect the same. Diabetes is even less clear. Unclassified, however, clearly rises over the year, if to a lesser degree than in the other places - it starts about 15/week and ends around 50 - a rise of some 333%.

Iowa (pop. 3.16m) - sparse enough they felt (along with Nebraska, Arkansas, and the Dakotas) no initial lockdown was needed - it worked for a while, but Iowa's seen increased cases and deaths recently. Details: Unclassified is sporadic here (low population, at least), with two 12 weeks in a row the only continuous points, and that's right at the first peak - likely significant. Because of that lack, and a mix-up, I also plotted cerebrovascular here. It shows a notable rise in its high spikes shortly before peak, if not a clear trend.  Heart disease dips well below average as the vulnerable protect themselves amid a mild early outbreak. Then it rises to bit above average or above right before and during the first peak, holds average then rises again just before the second near-peak in late July. Alzheimer's seems to rise the same at both peaks, a bit earlier than heart on the first one, and may stay a bit higher than usual. Diabetes shows little pattern - a slight rise at the first peak, perhaps faintly at the second one.

Conclusion 

It's still not clear to me just what this green lines show, but it seems to be a very interesting subject worthy of some consideration. 

When the issue of other deaths has come up, it tends to come with a note on disruptions to the medical system caused by "lockdown" and fear, often suggesting this is the main cause of increased "other" deaths. I should at least mention this aspect. There are likely some very few deaths of these kinds, including:

* people dying of a heart attack, etc. who are afraid to getting the covid at the ER and die at home. I don't expect very many of those, and I've seen no evidence that emergency rooms were closed for such urgent cases - it seems in general just optional and routine procedures were postponed to make room.

* There might also be deaths caused by delayed medical visits, like the ones where they find that lump in time. These will mostly take a longer and varying time to become fatal, not coming in a massive surge just after and before the peak of covid deaths, and mostly won't be in the surging classes. Who suddenly dies from Alzheimer's because some routine check-up was cancelled?   

So in short, the main issue is just what's causing these deaths, not scraping for reasons to blame "lockdown" for everything possible. The patterns are pretty clear - the main problem here with "lockdown" is it failed to prevent a further tens of thousands of deaths caused by this virus so far. That's not to deny the reality of the real costs of these policies, just to say an increased death rate is not one of them - 99% of all the deaths we know of and more are caused by SARS 2/Covid-19, not by the measures imposed to keep its killing from being even worse.

More Analysis

Add Sept. 9: Something said look at Louisiana, and the numbers seemed interesting. But once plotted, it's not so clear for the work, and some discontinuous data. There are big spikes in heart disease and cerebrovascular deaths near covid peaks, but not much for multi-week trends aside from: diabetes and Alzheimer's may be a bit elevated - supposed influenza deaths rise when they fell in 2019, as it happens, just as covid was killing the most  - heart disease deaths drop well below 2019 levels in the end.

California: I figured scale might help clarify the patterns, but California didn't seem to show much aside from a strong rise in unclassified deaths. Maybe they were ahead of the curve there, and avoided mis-classification, using this mystery classification more readily. I didn't see anything notable for Diabetes or flu, but Alzheimer's has some suspiciously-timed peaks, and supposed heart disease deaths drop well below 2019's levels, and then the numbers are pushed back up as Covid-19 and control measures battle it out. 

Texas: You can thank Alex Jones for a big part in this. They had things fairly under control considering it's Texas, but it's gotten stupid and deadly, with 3,357 coronavirus deaths in just 2 weeks in July. 

Diabetes and Alzheimer's should peak around that February rise we'll ignore, then hold fairly steady while slowly declining. Instead, they rise up and down without a decline, and Alzheimer's blue line rises solidly at the covid mega-peak of deaths in July. And out of obscurity, that green line of unclassified deaths has risen to match Alzheimer's at some 250 deaths per week, when it had been around 30/week at year's start.

I'll pause on Texas to compare peak weeks 29 and 30 (I didn't want to pick just one) in a bar graph with same weeks in 2019:

* weeks 29 and 30 2019 = 2,247 deaths in the categories heart (804+828=1632),  

diabetes (104+103=207), alzheimer's (192+159=351), and unclassified (29+28=57).  

* Weeks 29 and 30 2020 = 3,304 in those same categories (heart 995+959=1954, diabetes 154+162=316, Alzheimer's 291+255)=546, unclassified 255+233=488 - all are up, the last with a 756% increase!), plus Covid-19 (1,642+1,715=3,357 = 6,661 total. I took 2019 as a fair predictor of expected deaths in the 4 classes, took each one's overage and added it up for another 1,000-1,100 likely Covid-19 deaths.

Arizona: clear rises in Alzheimer's then heart just before small May peak for covid-19 - heart 

rises again (from a lower baseline) at the July peak ...

unclassified rises to 150 and even 175 per week, when it had been just 25.

Alabama: I had to look and see what effect the Tuscaloosa coronavirus parties of June might have had. 

https://libyancivilwar.blogspot.com/2020/07/to-tuscaloosa-corona-party-runners-up.html

not a lot in my current area - fatalities take off a bit, then far worse, but this happens everywhere. 

possible rises in Alzheimer's, repeated upward tumbling of heart disease deaths, little to no increased use of R00-R99.

These mixed findings and lasting curiosity have me doing a national review for just this category. I'll have at least one more graphic and some text notes to come.