Post #871: Age-adjusted trend in COVID-19 mortality rate in Virginia

Posted on October 17, 2020

After mistakenly focusing on short-term trends in Virginia COVID-19 cases (Post #867), I decided to do a proper analysis of changes in COVID-19 mortality rates in Virginia.

The question is:  Has there been any reduction in COVID-19 case mortality since the start of the pandemic, in Virginia, once you account for changes in the age of the newly-infected population over time?

I got a lot of surprises.  The biggest is that there appears to have been no material improvement in hospital care over the pandemic period, in Virginia.  The fraction of individuals dying, compared to the fraction hospitalized, is no different now than it was back in May.

And yet, the fraction of cases hospitalized, and fraction who died, both fell roughly in half during the pandemic period.  And that occurred despite testing levels remaining roughly constant since early July.  My conclusion is that the average severity of illness of diagnosed cases has fallen quite a bit over time.  It’s not clear what’s causing that.

Surprise #1:  Age standardization hardly matters.

The first thing to realize is that for hospitalizations, and particularly for mortality, the younger age groups hard matter.  You can’t infer it from this graph, but as of 10/15/2020, the entire under-60 population accounted for 4% of COVID-19 deaths, and about one-quarter of COVID-19 hospitalizations in Virginia.

With that in mind, the shift in age mix hasn’t been very large, in terms of likely impact on estimated mortality and hospitalization rates.  Looking above, the green stripe and above (50+ population) accounts for just under 40% of the COVID-19 at both the start and end of the dataset.  This means that any estimated end-to-end trend line for deaths (and to a degree, hospitalizations) will scarcely be affected by adjustments for the age mix.

Surprise #2:  Ratio of deaths to hospitalizations remains unchanged.

In this section, I’m going to estimate some simple straight-line trends on “case rates”, that is, hospitalizations per diagnosed case, and deaths per diagnosed case.  And then, I’ll look at the ratio of the two (deaths per hospitalization).

As the graph above shows, hospitalization rates for diagnosed COVID-19 cases have fallen roughly in half, in Virginia, since the start of the pandemic.  I don’t think the criteria for hospitalization have changed materially over this time, so I’m pretty sure that this suggests that, even after accounting for the change in age mix, the average newly-diagnosed case is less severely ill now than it was back in May.

Source:  Virginia Department of Health COVID-19 dashboard.

Whether that drop in the hospitalization rate is due to greater testing (and so, an expanded pool of low-severity cases), or to a true downward shift in severity across all cases, is not possible to say with precision.  But I will note that there has been no material change in the rate of testing in Virginia since early July.  And yet, by eye, that trend line fits both the pre- and post-July segments of the data fairly well.

The COVID-19 case mortality rate has shown a near-identical downward trend.  Although you cannot see it by eye due to the larger scale, the “elasticity” (slope of the line, relative to the average value) is virtually identical in the hospital graph and the mortality graph.  The linear trend for the mortality rate goes from about 3.5% of cases at the start of the pandemic to about 1.75% of cases currently.

Unsurprisingly, then, is that there’s no trends in the ratio of COVID-19 deaths to COVID-19 hospitalizations.

At this point, I’m going to play fast-and-loose with the data, and say that probably means that Virginia hospitals haven’t gotten any better at saving people, since the start of the pandemic.  For all you may have read about how this drug or that drug vastly reduces in-hospital deaths — well, if so, that doesn’t appear to have shown up in Virginia’s data yet.  If you are are infected with COVID-19, and are ill enough to be hospitalized, your odds of being discharged alive do not appear to have shifted over time.

That’s fast-and-loose because not all COVID-19 deaths occur in the hospital.  We can get a view of where deaths occur in Virginia from death certificate data compiled by the US CDC, through 10/10/2020.  Only about two-thirds of COVID-19 deaths occur in the hospital, with most of the remainder occurring in the nursing home.

Because about a third of deaths occur outside the hospital, we can’t be guaranteed that the lack of trend in deaths/hospital discharges means what I asserted above.  It’s possible — but in my experience, unlikely — that there might be exactly-offsetting trends.  Hospitals might have been better at discharging these patients alive.  And, just by coincidence, more patients might have died in the non-inpatient setting. I judge that to be unlikely.

At the minimum, I think it’s clear that if there has been any improvement (in likelihood of discharge alive from the hospital, for COVID-19, in Virginia), it has been small.