Games with Models, 5/2/20 Data

COVID19-hospitalizationsandprojections-050220-mod-featured

As today’s chart shows, Rhode Island revised its hospitalization numbers again.  The effect on trends is minimal, but it does make one wonder why.  The number of cases isn’t being revised, at least that I’ve noticed; why are hospitalizations and death, going back weeks?

One possibility, especially on the hospitalization front, is that the state is reclassifying people when they test positive for COVID-19.  So, to return to my hypothetical teen who broke a limb skateboarding:  He enters the hospital on some date and a few days later is tested positive for COVID-19, perhaps having caught it in the hospital or from a visiting parent.  It could be he’d now be reclassified as COVID-19 upon his date of entry.  But this is speculation.

All in all, today’s data shows good news.  The rate of increase in cases continues to slow, although not as much as my model wants to predict.  It’s possible that the 14-day infection rate will dip below one tomorrow, but we’ll see.

Another interesting note is that although the reported number of deaths is up by 17 today, the state is only showing a single one yesterday.  The rest are revisions back to April 17.

This variability may be entirely justified by the actual flow of data, but it’s making the reporting misleading.  In turn, it’s making it very difficult for the public to come to conclusions about what the government is telling — and instructing — us.

COVID19-hospitalizationsandprojections-050220-mod

  • Cases:
    • Projection for 5/2: 9,132
    • Actual for 5/2: 9,289
    • Projection for 5/3: 9,451
  • Hospitalizations:
    • Projection for 5/2: 341
    • Actual for 5/2: 333
    • Projection for 5/3: 323
  • Deaths:
    • Projection for 5/2: 284
    • Actual for 5/2: 296
    • Projection for 5/3: 301


  • Mario

    Yeah, the deaths are a little odd. For one, the latest revision has more people dying in the hospital on the 29th than died that day altogether. I think some of those deaths must have been yesterday and were misattributed to the latest hospital numbers.

    For that reason, even though it’s trending down I only expect 1 fewer hospitalization tomorrow for 332 (+35, -30.5, -5.5) but that will look better once the 29th is revised back closer to 340. Also 278 more cases, but as always that’s really tough without knowing how low the testing falls. I also show 8 people dying as of tomorrow for 304 total deaths, but if I’m right about the misattribution, some of those were already announced today, so I wouldn’t be surprised if its a little lower. At the same time, I think I’ve been underestimating that number quite a bit recently (I have the long-term outlook back over 450, and I’m not sure what change made that happen).

    The new charts the governor showed were interesting. Well, one of them was, and was characteristically missing gridlines. The gap in the projection is sizeable, but at least it gives up something to look at. If I measured correctly, it shows between 162 and 296 hospitalizations on the 19th (which in my numbers is the 20th, but whatever). Right now I have 219 on that day, which is solidly in the center of the range, a little closer to the good scenario.