Exploring excess deaths

I have spent the last 6 months trying to provide context for all of the data that has been dumped on us during this pandemic in order to help people understand why this is a big deal and why it needs to be taken seriously. Over this time I have responded to some of the common arguments people use to dismiss and/or downplay the pandemic, but as soon as these arguments are refuted, the doubters just move the goalposts and come up with a new excuse for dismissing and downplaying the severity of the situation.

Back in March, it was "very few people have died yet, so why are we making such a big deal of this?" It didn't take long for us all to get a lesson on the power of exponential growth. Over 180,000 deaths later, can we finally put that one to bed?

Then in April, it was "you can't believe the numbers, they are counting suspected deaths". I addressed that one as well. It turns out, that using suspected deaths in addition to confirmed deaths is what the CDC normally does when tracking epidemics because often people are not tested in time.

Then in May, it was, "Mission accomplished, we flattened the curve! Its time to get back to normal." Of course, this was premature and it was entirely predictable what happened next: cases spiked after people stopped social distancing.

In June as cases rose, the argument was "Yeah, but that is only because we are testing more." This could only explain a small fraction of the new cases, but it didn't stop people from using that excuse.

In July, people said, "Yeah, but deaths are still low." Of course, that was wishful thinking and the deaths started rising soon after, as expected.

I could go on and on, but you get the picture. A cynic might question whether continuing to present the data in an attempt to persuade is a waste of time, and maybe it is. But, I am stubborn, and I am not ready to give up on these people yet, so here goes another rant.

The latest complaint seems to be "You can't believe the death totals because most of the people who have died would have died from their comorbidities anyway." As I discussed last time, listing multiple causes of death on a death certificate is normal. However, it is fair to ask:Is it really reasonable to attribute all of those 180,000 deaths to COVID? How do we know the comorbitities weren't the real cause? As usual, let's look at the data.

The CDC publishes data on the total number of deaths over time. One way to see if these people would have died anyway is to look at how the number of deaths in 2020 compares with previous years. If the number of deaths was in line with what we usually observe, than that would be a sign that these deaths likely would have happened anyway. Here is a plot to help with this comparison:

There are a few things to note:

1. The thin curves represent the number of total deaths in each week (by all causes) in the US for each of the last 6 years. The death totals usually reach a peak of around 55k-65k per week in the winter before dropping to a low of 50k-55k in the summer. That variation is partially driven by flu season.

2. Before this year, the trends were pretty consistent across years and tend to be pretty close to the 2015-2019 average (thick red curve). The only place where you see substantial deviation from those trends is in a particularly severe flu season in late 2017 and early 2018. You can see this in the fact that the blue curve (on the right side) and orange curve (on the left side) are noticeably higher than other years.

3. The death totals from 2020 were pretty typical for the first few weeks and then something happened. For weeks 5-11 it behaved like a particularly bad flu season and then, death totals skyrocketed to almost 80000 in one week in April. This is a clear outlier. Since then the death totals have decreased from the peak, but they have remained substantially higher than normal.

Keep in mind, that this data is COVID agnostic. This simply tells us that in 2020, more people are dying than normal, but it does not tell us why. It does suggest that comorbidities (which are not unique to 2020) are not a credible explanation for the number of deaths that we are seeing.

To look at this, we can compare the number of deaths in 2020 to the number of deaths we would have expected in a normal year. The difference is called the number of excess deaths. There are various ways of defining excess deaths, some are more complicated than others (using the upper bound of a confidence interval) but I keep things simple and use excess deaths to refer to the deaths above average. (i.e. the difference between observed deaths and the average deaths from 2015-2019). Below I have displayed the number of excess deaths in each week and the number of deaths that have been attributed to COVID-19. Note: This data runs through the end of July since the CDC is still tabulating death counts for August.

A few excess deaths could be a statistical fluke, but this is a LOT of excess deaths. This suggests that ...newsflash: something unusual happened in 2020. Something caused these extra deaths and it wasn't just the usual suspects.

So what caused the excess deaths? Some of those deaths could be explained by side effects of lockdowns. When hospitals canceled elective procedures, some people who normally would have gone to the hospital stayed home and some may have died prematurely as a result. Some people who normally would have had their conditions diagnosed during routine care did not get that routine care and as a result some of their conditions went undiagnosed. There have probably been rises in suicides and given the amount of social unrest, perhaps homicides as well. (I have looked but I haven't found data on this yet; if anyone knows where to find it, let me know.)

Are these factors a satisfactory explanation for all or even most of the excess deaths that have occurred? Of course, not. There is a more reasonable explanation. Notice how the trend in excess deaths tracks closely with the number of COVID deaths. Instead of resorting to a vast and complex hoax involving fabricated data, the simplest explanation is that the most of the deaths were caused by the global pandemic that has affected every country on earth. If anything, the number of excess deaths suggests that we may be undercounting COVID deaths rather than overcounting. You can see this in the plot below where the y-axis displays the cumulative number of deaths (year to date).

As of the end of July, there had been around 150,000 total deaths attributed to COVID, but there had actually been over 225,000 more deaths than in a typical year. Maybe some of those deaths were due to normal fluctuations. Maybe some of these deaths were caused by mitigation measures in response to the pandemic. But, if you want to argue that most of the now 180,000+ COVID deaths that have been reported were actually not caused by the pandemic, then you should know that you have over 225,000 excess deaths to account for. Many of those deaths occurred before the lockdowns took effect or soon after, so if you want to blame them on the response, the data is not on your side.