15. Weekly Update: Recovery at Hand but Slower than Expected

(5/6/20) Recovery is everywhere but stubbornly slow and gives pause to suggested safe easing dates. CA and NJ are significantly revised out in time due to persistently high daily deaths and hospitalization counts. We also highlight an interesting “Sunday Effect”. This week we implemented our asymmetric Gaussian model to account for the slow downslope and compare further to the UW IHME model.

You know the drill so we will launch into it.

The plots below show the familiar death rate curves for hotbed countries and U.S. states. We dropped China and Korea last week as now being “uninteresting.” Next week we will drop Iran and add Sweden to highlight a country that is paying the price for a lackadaisical approach to social containment.

Internationally there are new upgrades (U.S. marginally, France), but also two down-grades (Italy, Iran) on our 3-color ranking. Domestically MI got an upgrade.

We continue to plot a symmetric Gaussian but for visualization only. Our analyses now use asymmetric functional fits that we will detail in a separate post in the near future.

We make the following observations:

  • All of our tracked hotbed countries and U.S. states are on the downside of the death peak and therefore the prevalence (case) peak. However, it just doesn’t feel solid with several sudden surges most likely due to reporting fluctuations, but we worry about hidden deaths and small outbreaks that can grow quickly into big ones.
  • Do Fewer People Die on Sunday? We have noticed fluctuations in the death rates, but now we see that it is repeatable and is found in many countries and states. We have adjusted our plot grid lines to lie on Sundays and you can readily see the strong dip in reported deaths. These are particularly noticeable in the U.S. (NJ, LA, CA), but also in Europe (U.K. and very much Sweden not shown this week). We attribute this to a data recording quirk.
  • The symmetric Gaussian model is breaking down on the downside of the death rate curve as we expected and we have implemented an asymmetric function that has different sigma values for the rise and fall sides of the rate curve. We will discuss the details in an imminent post.

Next is our familiar table for forecasted total deaths, prevalence (current cases), and incidence (new cases) along with their values per capita (per million people) as well as dates we consider to be the earliest to begin a graduate easing of social distancing. We will continue to call this an easing date and not a safe date to dampen excessive hopefulness.

We have lowered the mortality factors for Italy and Spain from 2% to 1.5% and for NY and WA from 1.5% to 1.0% as the healthcare system in these populations are becoming less overwhelmed in treating patients.

Last week we implemented an asymmetry factor to adjust our values until we could come up with a rigorous functional form. This has now been implemented but still needs to be “burned in” and tested more but we felt it was important to apply it here. Most of our forecasts have gone up only moderately, but some rather significantly (US, NJ, CA). This has also pushed out the so-called easing date for most populations and significantly, a month or greater, for the latter ones cited. As these dates are untenable in the current social, economic, and political climate very careful limitations need to be placed on any phased easing of social restrictions that should occur before these dates. We will also give further consideration to whether our threshold of 100 active cases per million people for safe easing is too stringent and whether we could recommend more modest easing at earlier dates.

But we mustn’t lose sight that we are at great risk of prematurely easing, which as Dr. Fauci has said “could backfire.” We will need to observe outcomes in Europe and U.S. states where social easing is already being implemented. I’d be interested in your thoughts on my posting: “11. Recommended Guidelines for Easing of Social Distancing,” which proposes a three-phased easing with check-gates at each step.

We now wrap up by comparing our results to that of the Institute for Health Metrics and Evaluation (IHME) at the University of Washington (UW), which is now the most highly cited and quoted model for informing our nation on the state of COVID-19 (http://www.healthdata.org/covid/).

The IHME model seems to have dropped the statistic for ‘days from peak’ that we found to be an interesting comparison.

It is clear that we have similar components to our models with an apparent heavy emphasis on death rate. This is a very nice, visual, and professionally developed model (funded by Bill Gates) that we can’t compete with in its full glory. The IHME model has also moderately to significantly increased their total death forecasts, particularly for the U.S. to the point where it feels like they are putting a little “tilt” into it in response to latest administration and media hype. We shall see. The two models differ on Sweden in which IHME is projecting more than twice as many deaths as us. They may be factoring in a social distancing component to their model that amplifies this number. We believe that the death rate curve embodies all of these effects.

Although the two comparative models give generally similar results, by some measures we may be performing better in terms of week-to-week volatility and quickness to detect new trends (see post 14. Benchmarking COVID-19 Forecasting Models). Our model also provides calculations of current and forecasts of future prevalence (active cases) and incidences (new cases) that are notoriously difficult to measure because of lack of adequate COVID-19 testing making the reported confirmed values almost meaningless.

Leave a Reply