Thursday, March 26, 2020

Reckless Research?

Models are just that, models. I have multiple models. Many of them are financial models for start up companies. But what we then do as things start to happen is adjust the models to reflect the facts as they occur. The Imperial College of London is a model ten days ago which predicted an apocalypse. Now a short tens days later they seem to be dramatically backtracking stating:

 The modelling showed that implementing measures early on can have a dramatic impact. If all countries were to adopt this strategy at 0.2 deaths per 100,000 population per week, 95 per cent of the deaths could be averted, saving 38.7 million lives.

 In my opinion and in my experience this is grossly reckless. We really do not know enough about the propagation of this virus to make any reliable predictions. As we have been seeing in New York, Boston and Washington we have some 30,000, 500, and 550 respectively.

New York is larger but the attack rate is substantially higher. One cannot say the cities are materially different in population risks.

When we see these facts and then when we see the calamities from the London folks we then lose any and all trust that they can make any credible statements. Reckless?

BTW, we tagged this when it came out and the NY Times spread it like a call in a theater.

In a recent NEJM report the authors rationally state:

Thus, several questions are especially critical. 
First, what is the full spectrum of disease severity (which can range from asymptomatic, to symptomatic-but-mild, to severe, to requiring hospitalization, to fatal)?
Second, how transmissible is the virus?
Third, who are the infectors — how do the infected person’s age, the severity of illness, and other characteristics of a case affect the risk of transmitting the infection to others? Of vital interest is the role that asymptomatic or presymptomatic infected persons play in transmission. When and for how long is the virus present in respiratory secretions?
And fourth, what are the risk factors for severe illness or death? And how can we identify groups most likely to have poor outcomes so that we can focus prevention and treatment efforts?

Reasonable. And making projections without even these simple steps, is well, I said what my opinion is.


NJ Updates

We start with the state total daily increase as below:
The we look at the County Daily increases:

Now finally the total numbers:

Clearly we are seeing a significant daily increase today. The per county numbers tell somewhat a geographic story.Hudson and Union are proximate to New York. Bergen is the canary in this scenario.

Politically Correct Fact Hiding: A Potentially Lethal Decision

It appears that New York City is refusing to release details of COVID-19 infections. This is akin to hiding cholera in London and allowing massive spread. ProPublica notes:


Instead New York, along with several other state and county governments around the country, has released daily data only on the county, or borough, level. That means there is just one figure for COVID-19 cases in all of Kings County — Brooklyn — which has a population larger than 15 states. The roughly 4,600 confirmed COVID-19 cases among Brooklyn’s 2.6 million residents account for 8% of the confirmed cases in the entire country. There is also just one coronavirus case figure for the 2.2 million residents of Queens, where there are just over 5,000 confirmed cases. The lack of detailed information makes it difficult for medical workers, journalists and the public to establish whether particular communities in the city are being harder hit and to get beyond anecdotal accounts of which of the city’s roughly 60 hospitals are already overwhelmed.

As we have noted again and again, testing and data are essential. Models are meaningless without validation and validation demands data.

To quote my father, "Prior planning prevents poor performance" That can only be done with data. In my opinion and based upon my experience the refusal to make public the data is making the problem worse! 

Also without this data, how, one wonders, can those wizards at the NY Times ethically present their alleged models?

Interesting Result

The NIH Director's Blog has an interesting read. Suggest looking at it. He states:

The researchers went on to analyze genomic data related to the overall molecular structure, or backbone, of the new coronavirus. Their analysis showed that the backbone of the new coronavirus’s genome most closely resembles that of a bat coronavirus discovered after the COVID-19 pandemic began. However, the region that binds ACE2 resembles a novel virus found in pangolins, a strange-looking animal sometimes called a scaly anteater. This provides additional evidence that the coronavirus that causes COVID-19 almost certainly originated in nature. If the new coronavirus had been manufactured in a lab, scientists most likely would have used the backbones of coronaviruses already known to cause serious diseases in humans. So, what is the natural origin of the novel coronavirus responsible for the COVID-19 pandemic? The researchers don’t yet have a precise answer. But they do offer two possible scenarios. In the first scenario, as the new coronavirus evolved in its natural hosts, possibly bats or pangolins, its spike proteins mutated to bind to molecules similar in structure to the human ACE2 protein, thereby enabling it to infect human cells. This scenario seems to fit other recent outbreaks of coronavirus-caused disease in humans, such as SARS, which arose from cat-like civets; and Middle East respiratory syndrome (MERS), which arose from camels. The second scenario is that the new coronavirus crossed from animals into humans before it became capable of causing human disease. Then, as a result of gradual evolutionary changes over years or perhaps decades, the virus eventually gained the ability to spread from human-to-human and cause serious, often life-threatening disease.

The concern of many is that it was a man made variant with a gain of function insert.

Models, Models, Models

OK, here is my model for a city of 6 million. The NY Times presented a model that alleges to have some validity. Frankly, in my opinion based upon the current facts and evidence I really wonder if anyone has gotten it yet. My general problem is that the presenter is so politicized that one's ability to trust what is said is highly questionable. Trust is a key factor. We have so few we can trust.

Let us summarize some key facts:

1. COVID-19 is a single stranded enveloped positive RNA virus of about 29,000 bases.

2. The virus enters the patient via nasal passages

3. The virus is aerosolized by infected individuals with or without symptoms.

4. It appears that transfer is dominated by hand contact with infected surfaces and then hand transfer to the nasophyranx.

5. The latency period is about 10 days until symptoms arise.

6. Symptoms may vary dramatically. This is a serious problem for the near asymptomatic are carriers.

7. We  are assuming that once free from symptoms that there are viral antibodies which will inhibit secondary infections.

Now models are complex and prone to error. A model for New York is not a model for Boston. In a recent piece in Science the writer notes:

Models are at their most useful when they identify something that is not obvious, Kucharski says. One valuable function, he says, was to flag that temperature screening at airports will miss most coronavirus-infected people. There’s also a lot that models don’t capture. They cannot anticipate, say, the development of a faster, easier test to identify and isolate infected people or an effective antiviral that reduces the need for hospital beds. “That’s the nature of modeling: We put in what we know,” says Ira Longini, a modeler at the University of Florida. Nor do most models factor in the anguish of social distancing, or whether the public obeys orders to stay home. Recent data from Hong Kong and Singapore suggest extreme social distancing is hard to keep up, says Gabriel Leung, a modeler at the University of Hong Kong. Both cities are seeing an uptick in cases that he thinks stem at least in part from “response fatigue.”  “We were the poster children because we started early. And we went quite heavy,” Leung says. Now, “It's 2 months already, and people are really getting very tired.” He thinks both cities may be on the brink of a “major sustained local outbreak”. Long lockdowns to slow a disease have catastrophic economic impacts and may devastate public health themselves. “It’s a three-way tussle,” Leung says, “between protecting health, protecting the economy, and protecting people’s well-being and emotional health.”

 Based upon data we have been examining, albeit limited and a major defect of the Government's process, we know that models must reflect:

1. Demographics such as age, education, employment etc

2. Psychographics such as cultural propensity for proximity. Certain cultures have a great deal of physical contact such as hugging, kissing, large family dinners and the like.

3. Population density

4. Social interaction. Such things as where is food purchased, social gatherings such as religious sites

5. The transfer mechanisms. This is the ways the virus is initially transferred via proximity, surfaces, hand washing etc

6. Openness of the community. How open is the community to new entrants.

These are just a few. Then the model must be built by census tracts at the very least if not by zip codes. The NY Times model make the absurd assumption that the nation is homogeneous. Morgantown WV is NOT Bedford Sty!

If one is to build a model one must validate it. Or better, as Popper suggests invalidate it. One must build models and then with test data validate them, modify them.

In fact what engineers do is build models and then using the techniques of system identification adjust the model parameters and even its structure as data is obtained. There should be hundreds of Zip Code models and data adjusting each of them. From these dynamically adjusted models one then should attempt to determine why some regions are doing well while others are failing. That is the way decisions are made, not by one giant, one size fits all, model. This latter approach does a massive disservice and erodes trust. Lose that trust folks and we have real problems.

As to reigniting the country and the economy we can do so then with the models and DATA! However the facts are simple:


1. assuming no NEW infected people entering the pool, this burns out in 10-14 days after quasi quarantine.

2. assuming no mutations, the above remains true. However there is always the risk that this virus has a gain of function gene

3. assuming no Typhoid Mary types, namely infected but not sick and no antibody protection individuals, thus a silent carrier

Namely we really do not know the full scope of the pathology. Reinfects will occur, and that is why border control is critical. We quarantine horses, dogs and cats, a new Ellis Island? My concern is that there is a pandemic of politics and no one believes anyone else. That may be worse than the virus.  So a suggestion:

1. Test test test

2. try it in a less critical set of locations. Kansas City or Richmond or Knoxville and test test and see what happens

Unwinding NYC is very unwise since I can almost assure people there are Typhoid Mary types somewhere. Bottom line: this will be a gigantic but necessary somewhat controlled experiment....and we must stop the politics, please.

One final thought. To understand the dynamics of this pandemic, perhaps one needs the expertise of market researchers not just epidemiologists. The human factor always dominates. Market researchers find out who, why and what people want to buy. It parallels disease progression. Really!

Wednesday, March 25, 2020

Common Sense Testing

Seeking Alpha reports:

Results from a study led by UnitedHealth Group Research & Development and OptumCare showed that COVID-19 tests performed on specimens collected by patients themselves, rather than healthcare professionals, are on par with the current approach. Specifically, the sensitivity of self-administered COVID-19 tests was more than 90%, consistent with clinician-administered results. The study,  prepared for publication, included almost 500 patients at OptumCare facilities in Washington state. Widespread adoption of self-sampling would reduce exposure for healthcare workers while improving efficiency.

This makes great sense. It is simply a nasal probe, a Q tip if you will. The problem as is done now is the logistics of setting up a site, operating it, and the gruesome sampling by site staff by jabbing the Q tip up the nose. The cost would drop. It could be set up on line, and all it would require would be a sample tube, Q tip like collector, label numbered and return envelope.  It is akin to Ancestry collections. The logistics would best be done by sending to local labs, with 24 hour delivery 24 hour turn around, secure web site report for HIPPA compliance.

Also one could pick up a "kit" at CVS, bring to patient, patient logs in and enters information and gets number for label on kit. Return kit to CVS and daily pickup. 24 hours end to and is doable. No parking lots, no staff, logistics are easy, HIPPA no problem, data entered into system from the get go.  

 I wonder if Governments even think this way!