Two days ago, New York started including probable COVID deaths in its reports.
We’ve known all along that not all cases and deaths were being reported in the numbers we see. On the April 7th, Gothamist reported that the increase in deaths in the city was far above normal ( https://gothamist.com/news/surge-number-new-yorkers-dying-home-officials-suspect-undercount-covid-19-related-deaths ) and the New York Times followed up with it’s own story reporting that overall deaths in the city were over twice the normal rate ( https://www.nytimes.com/interactive/2020/04/10/upshot/coronavirus-deaths-new-york-city.html ) The data from those articles suggest that actual COVID related deaths may be up to twice the confirmed numbers. Even with the inclusion of “probable” cases, New York is likely missing some COVID related deaths.
We’ve also already seen this become a politicized issue. Governments generally want the reported numbers to be as low as possible so that they appear to have been effective in their response. Trump famously said that he didn’t want infected passengers to disembark from the Grand Princess cruise ship because they would add to the official number of cases in the US. He has accused NY of inflating the number of deaths and has pushed back against the need for widespread testing. Similarly, there have been suggestions that China is underreporting its cases and deaths. Governments are motivated to under-count their own country’s cases and deaths and to over-count those of other countries.
On the other hand, accurate counts are essential to effective public health policy.Without an accurate count of cases and deaths, there is no way to accurately estimate the risks of different policy approaches. Without widespread testing, the only effective policy is to act as if everyone is infected, which is more or less what we’re doing now. Of course, this has major negative impacts on social life and the economy.
So, governments are sort of in a pickle. They want low numbers but also low restrictions. Unfortunately, they can’t have both. This applies at all levels of government and different officials take different approaches.
From a data perspective, we have to make a choice about what to do when there are sudden changes in what gets reported. There isn’t a single right choice here. If we include the probable cases, then it may look like there was a sudden surge in deaths when there really wasn’t. If we don’t include them, then it looks like there are fewer deaths than there were. If only some jurisdictions are reporting probable cases, then including them can make comparisons across jurisdictions more difficult. On the other hand, there are already differences in how jurisdictions report cases and deaths, so why worry about one more?
So far, the major trackers have not settled on a single approach. Johns Hopkins tracker has simply included the probable numbers. The COVID Tracking Project does not include them. So there is a difference of more than 3,000 in the number of deaths these two trackers report for NY. 1point3acres, the tracker I have been using for my projections seems to take a middle ground. They do not include probable cases for their graphs or state level reports, but do add them in for the top level national numbers. Their graph currently shows 30,892 deaths but their top line is 34,873 deaths with a note that it includes 3,981 probable deaths.
The upshot of all this is that we are likely to see increasingly divergent numbers about deaths. Some of this is due to choices that various trackers make. More of it is about what media sources deem to be most helpful for their readers to understand. A whole lot of it will be based on politics. I expect the White House to report only confirmed cases, but Pelosi’s remarks will include the probable numbers. Fox News will stick with confirmed, but MSNBC will include probable. Etc. Choices in what to report will be influenced by the underlying political effects of the choice. I also expect that states with Democratic governors will increasingly begin to follow New York’s lead in reporting probable cases while those with Republican governors will tend to report only confirmed cases. If I’m right about this last expectation, then the numbers will be increasingly divergent as more states follow NY.
I haven’t decided how to roll this all into my projections yet. I’m relying on trackers for my numbers, and my current setup doesn’t deal well with the discrepancies between how they have approached the issue. I’ll be thinking about how to approach that over the next few days. If anyone has a view about what you would find most helpful, let me know in the comments please.
For today, I’m going to stick with what I’ve been doing and project only the confirmed numbers. Keep in mind that you may see reports of about 4000 more deaths than these when probable numbers are included in media reports.
After some lower numbers at the beginning of the week, the last few days have seen higher numbers of both cases and deaths. It’s too early to tell whether this is a change in the overall trend or just some noise in the data. The rates haven’t changed much since Wednesday, the increases have largely balanced out the decreases. New cases are still averaging about 29,000 per day and deaths are increasing at about 9% per day. Here is the projection based on the unrounded numbers.
| Cases | Deaths | |
| Current | 677,969 | 30,867 |
| 4/17 | 706,583 | 33,466 |
| 4/18 | 735,197 | 36,284 |
| 4/19 | 763,811 | 39,339 |
| 4/20 | 792,425 | 42,651 |
| 4/21 | 821,039 | 46,243 |
| 4/22 | 849,653 | 50,136 |
| 4/23 | 878,267 | 54,358 |
| 4/24 | 906,881 | 58,935 |
| 4/25 | 935,495 | 63,897 |
| 4/26 | 964,109 | 69,277 |
Again, let me know if you have a view on how to most helpfully deal with the inclusion of probable cases.