« Intermission over | Main | Reading Gelman's take on election polling »


Feed You can follow this conversation by subscribing to the comment feed for this post.

Michael Droy

You miss the point.
The claim made frequently is that death by covid takes years of peoples lives. 12 years according to one study based on life expectancy without co-morbidities.
Death with co-morbidities clearly takes fewer years of lives - months in some cases.
The issue is not whether the difference is more complex than some like o simplify it to. The issue is whether covid deaths take 12 years off life of 6 months.
The all death by Covid position is far more misleading that the with of by question.

A Palaz

The question from this post is is test within 28 or other number of day a comorbidity? Or some other categories?. But also and maybe more importance is that I may treat people different, if I have not other or make correct diagnosis. So would you say this us not important too?


MD: My point is that whatever one proposes doing for Covid19, one's got to do for all causes fo death. I am not aware of the 12-year claim, and I wasn't responding to that.


AP: The core question is causal inference. If the person did not have Covid, would the person have died anyway on that day? But if we want to do it right, it would take a lot of time to investigate every death. Not worth the time.
For example, consider those in pallitive care. By definition, all of the patients are likely to die "soon". If Covid-19 swept the facility, and killed all of them at once, are these deaths with Covid or by Covid?
I'd just leave this question here to help clarify one's thinking.

A Palaz

So I would say yes maybe we can guess on this die anyway effect. So say hoe many peoples is dying in 28 day from birthday? In any year guess 28/365. Then how many people gets positive test in period? ..and so on. It is important because it feed VE as you deal with in your next post.
And VE feed policies.

At the lower level would you admit a positive test persons for observations or a negative test? It feeds into admission policies. But we must note that I have seen reports that the test status is also sometimes only checked after discharge against database so.in many case there is also maybe no clear relation of test to admission.

Michael Droy

Kaiser - surely the point of any analysis is to provide support for those making policy decisions.

For that reason understanding whether Covid deaths steal 6 months or 12 years of life is crucial and the only real question in town. It can't be ducked by saying we don't measure that for other causes of death.

There is an attempt to do an Actual QALY analysis of Covid lockdown policy here. Strangely it is the first one I have seen although there was nothing to stop it being done in May or June 2020. There might be a lot to criticise but it is THE Question.
I'm curious what you think about it.

Lives saved and lost in the first six month of the US COVID-19 pandemic: A retrospective cost-benefit analysis


MD: I agree that excess loss of life is one of the most important metrics. I think I wrote about it before as well - excess deaths don't work in the long run, too many factors are then involved. I have saved the paper but my pile is getting large so hopefully I will get to it soon. Thanks for the tip.


AP: Not many people appreciate the nuances of all measurements. In your response, I'm seeing you're grappling with it. There are so many little things that can affect the measure. No one has yet asked whether hospital admission or death certification takes into account vaccination status. It stands to reason that one would think someone died of comorbidites rather than Covid-19 if fully vaccinated, no?

Michael Droy

Kaiser - surely loss of life is the worst metric when we are comparing effects on young people and on the elderly or infirm. Quality adjusted Life Years is not just the way to go in the future, it has been the core measure for public health policy for decades.
(or did you mean remaining years of life by excess loss of life?)

Ron Kenett

The data can be analyzed using competing risk models. These are used in reliability analysis of product failures. Several failure mechanisms can lead to failure. eventually a failure is attributed to on mechanism. For an intro to reliability models see my book on Modern Industrial Statistics https://www.wiley.com/en-gb/Modern+Industrial+Statistics%3A+With+Applications+in+R%2C+MINITAB%2C+and+JMP%2C+3rd+Edition-p-9781119714903

The comments to this entry are closed.

Get new posts by email:
Kaiser Fung. Business analytics and data visualization expert. Author and Speaker.
Visit my website. Follow my Twitter. See my articles at Daily Beast, 538, HBR, Wired.

See my Youtube and Flickr.


  • only in Big Data
Numbers Rule Your World:
Amazon - Barnes&Noble

Amazon - Barnes&Noble

Junk Charts Blog

Link to junkcharts

Graphics design by Amanda Lee

Next Events

Jan: 10 NYPL Data Science Careers Talk, New York, NY

Past Events

Aug: 15 NYPL Analytics Resume Review Workshop, New York, NY

Apr: 2 Data Visualization Seminar, Pasadena, CA

Mar: 30 ASA DataFest, New York, NY

See more here

Principal Analytics Prep

Link to Principal Analytics Prep