Read on Twitter
What is causing MINNESOTA& #39;S COVID MODEL to predict such high death totals??
A deep look into the formulas explains why ….. CLICK HERE for the thread that explains it
#minnesotamodel
A deep look into the formulas explains why ….. CLICK HERE for the thread that explains it
#minnesotamodel
There have been many questions about the new MN covid model and reactions to its comparatively high death projections. After reviewing the technical underpinnings of the #minnesotamodel, I will explain what is causing the higher projections and provide perspective for clarity.
To be transparent, I will provide supporting links & references. Long thread, so stick with it. As always, I welcome comments. And given that I just started this account to help in this fog of covid, your help getting the word out to promote smart decisions is appreciated.
In looking deeply into the #minnesotamodel function and description, there is one input used in multiple formulas that stands out, ᵟ , which is the detection rate for infections (see below)
First, what is this input? It is the degree to which we can detect actual infections within the community. It stands out because this input has a great degree of UNCERTAINTY and POWER to massively shift the projections.
From a mathematical standpoint, this detection rate input plays a significant role. It works like a water valve- a higher detection rate opens up the flow and allows more patients into the Hospitalized, ICU, and Death “states” in this SEIR model. Thus, higher numbers come out.
It is not that detection is bad, or that one would want a low detection rate going forward. It has to do with how the model is setup and how it calculates death and ICU projections. The severity inputs in the model rely on detection.
From a practical standpoint, the question is “how many people have and will go undetected?” It is clear in the technical document that the modelers have struggled with this question.
The modelers say they needed to use 1% for the input in order to “calibrate” and make the model “fit” the observed deaths up to Mar 22 in MN. But for estimating going forward, the modelers increased the rate to 50% - 90%, and noted that clustering may be complicating things
So while the R0 in the last version was raised, the detection rate was set at 75%. The Minnesota officials stated on Apr 10 that they believe the actual number infected in Minnesota is 100x the reported cases. This would mean a 1% detection rate may be more realistic.
But, if the detection rate is really around 1%, then it follows that over 2% of MN has already been infected and the max eventual deaths in MN would be less than 2,500- and that’s if everyone got infected. See the simple math below
However, as stated, the modelers increased the detection rate to 75% going forward, which will increase the model’s projected hospital beds, ICUs… and deaths. That’s a big reason why the base projection is 22,000 deaths right now and why there is such a big range.
In using 75%, the modelers footnote a CDC statement that claims up to 25% of covid cases may be asymptomatic. However, this input in the model may also need to factor in other mild cases that do not get detected. The CDC statement from the modeler’s footnote is below
It& #39;s conceivable that there are many people that have & will go undetected with mild or no symptoms. Dr. Gottlieb, fmr FDA Commissioner, said that covid exposure in the US is “somewhere between maybe 2 or 5%” That suggests a detection rate under 10%. https://www.cbsnews.com/news/transcript-scott-gottlieb-discusses-coronavirus-on-face-the-nation-april-12-2020/">https://www.cbsnews.com/news/tran...
Another reference point, the 2017-18 flu season was widespread and severe. While not covid, the CDC-reported detection rate for that epidemic was an incredibly low 0.1%. Here& #39;s a link directly to the CDC for these numbers
https://gis.cdc.gov/grasp/fluview/flu_by_age_virus.html
https://gis.cdc.gov/grasp/flu... href=" https://www.cdc.gov/flu/about/burden/2017-2018.htm">https://www.cdc.gov/flu/about...
https://gis.cdc.gov/grasp/fluview/flu_by_age_virus.html
Furthermore, Minnesota has had constrained testing capacity and has intentionally limited its testing, thus making it more likely that a significant number of infections have gone undetected. See current MDH testing limitations below
https://www.health.state.mn.us/diseases/coronavirus/hcp/eval.html">https://www.health.state.mn.us/diseases/...
https://www.health.state.mn.us/diseases/coronavirus/hcp/eval.html">https://www.health.state.mn.us/diseases/...
Finally, as reported Friday in the Economist, evidence continues to emerge that the virus has already spread more than anticipated, which again suggests a lower confirmed detection rate and a lower infection fatality rate. See the report below https://www.economist.com/graphic-detail/2020/04/11/why-a-study-showing-that-covid-19-is-everywhere-is-good-news">https://www.economist.com/graphic-d...
In summary, this little symbol ᵟ, the detection rate input in the #minnesotamodel carries with it a lot of weight and a lot of uncertainty. At its current setting and functionality in the model, with its interaction with the severity input rates (table S3), the detection rate...
... is the key driver of high ICU needs and deaths rates in the #minnesotamodel. Hopefully, we will learn more, and this input will be scrutinized and adjusted for design and level of accuracy as we move forward.
What is causing MINNESOTA& #39;S COVID MODEL to predict such high death totals??
A deep look into the formulas explains why ….. CLICK HERE for the thread that explains it
#minnesotamodel
A deep look into the formulas explains why ….. CLICK HERE for the thread that explains it
#minnesotamodel
