Sunday, August 02, 2020

Deaths lag cases

Perhaps you've heard that, with respect to the coronavirus epidemic, deaths are a "lagging indicator." That means some fraction of today's new cases will inevitably result in deaths at some point in the future. When viewed graphically, the curve for deaths roughly tracks the curve for cases in basic trajectory, but with a time delay.

To illustrate this, I've juxtaposed the graphs of daily cases and daily deaths for the U.S., but offset them to align the curves for the most recent surge in cases and, now, deaths. (Click on the image for a larger view.)

Adapted from The Washington Post

The current surge in cases began around June 11 (or a little earlier; the curves are smoothed with a 7-day moving average). The surge in deaths began around July 2—a delay of about three weeks.

It's intuitively obvious that the delay is due to the fact that the disease typically takes some time to fully run its course, and that time is measured in weeks. In a lot of people who eventually develop symptoms, disease progression starts slowly and builds in severity over time. Symptoms are often mild at first, followed by more severe symptoms, followed by hospitalization, followed by ventilation, and ultimately (in some fraction) by death.

The delay between diagnosis and death is therefore expected and unsurprising. The main reason the lag is noteworthy is that it underscores the dismal importance of today's high daily new cases numbers, which have recently been in the 60,000 to 70,000 range. A certain number of future deaths are, unfortunately, "baked in" to the current new cases statistics. Some fraction of current new cases will inevitably result in death, which is one reason the stunning daily new case numbers are so discouraging.

Thankfully, the percentage of "baked in" deaths has come down over time. There are three principle reasons. First, a larger fraction of new cases now involve young persons, and they are less likely to die of the disease (although it's thought that they can sustain permanent lung damage, even in the absence of symptoms). Second, our treatment capabilities have improved compared to several months ago. And third, more widespread testing is likely to turn up more cases where the individual is less severely ill than in the early going.

Daily new cases peaked at 76,491—a jaw-dropping statistic—on July 17 in the current surge. A three-week lag suggests we might see a peak in deaths toward the end of this week. It will be interesting to observe.

Incidentally, the astute observer will notice from the graphs that the lag time between new cases and deaths was a good bit shorter in the early part of the pandemic. I am not an epidemiologist and so I can only speculate (indeed, that disclaimer holds for everything I've said here), but the likely reason is that in the early going so little testing was available that those who were tested were already quite sick.

Other observations are worth noting. During the initial surge daily new cases peaked at 34,203 on April 25. That was the period when New York hospitals were being overrun. In April daily deaths were routinely more than 2,000 and sometimes close to 3,000. But in the most recent surge we've doubled the daily new cases, while daily new deaths have recently averaged around 1,300. So again, a smaller fraction of cases are resulting in deaths. Even so, more than 1,000 deaths per day occurring six months into the pandemic is itself a sobering result. We've clearly done a very poor job managing this thing. (By contrast, Germany is reporting less than 4 deaths per day.)

Another observation: For a long time, total deaths divided by total cases—a crude calculation of the case fatality rate based on headline numbers—was running around 6 percent. Now it's about half that. So, again, apparent progress. On the other hand, this calculation doesn't account for the time lag, which further limits its usefulness. In any event, the true case fatality rate is far lower, because infections are thought to exceed confirmed cases by a factor of at least 10. None of this is cause for complacency; this disease is killing an extraordinary number of people regardless of the true case fatality rate. And infected persons who are asymptomatic and who have not been tested can still spread the disease, helping to ensure it will eventually reach someone who will die of it.

Finally,  you've probably heard it said that the current surge, which began in mid-June, was not the anticipated "second wave" but just a continuation (and indeed, an expansion) of the first wave. A true second wave could still occur in the fall, and could occur simultaneously with seasonal influenza, which could be a serious complicating factor. Get your flu shot!

An oft-cited model from the University of Washington projects 230,822 total deaths by November 1. That's over a quarter million. Don't be one of them.

 Copyright (C) 2020 James Michael Brennan, All Rights Reserved

The latest from Does It Hurt To Think? is here.

0 Comments:

Post a Comment

<< Home