Improved Infections estimate

[djay]

The current infections estimate uses the ages of the population to infer the chance of a given person dying of covid. However if this was accurate then the predicted median age of death would be ~80 instead of 65-70.

Ideas

• get better estimate of ages of those that died
  • take ages from cases - 11 days (or current avg days till death from confirmation).
  • get more data on ages of deaths - https://github.com/djay/covidthailand/issues/34
  • some combination?
  • these should raise the estimate since both should lower ages than the population?
• factor in comorbities.
  • Said to be more in thailand than global average. True?
  • if true, how to modify the IFR to factor this in?
  • if true will lower the infections estimate
• Use proper distribution for IFR age ranges
  • current interpretation of the IFR ranges is not a curve but stepped. Should use a fitted curve.
  • see https://colab.research.google.com/drive/1PKiMzdEff1ZQUMqJzfo1vwBOs2i7oUEO#scrollTo=OAh4Rfzn9UvE
  • should lower the estimate as currently risk is flat after 65 but this would increase risk up until 100+
• estimates of untested deaths? (estimated as 1-1.2x)
  • Use excess deaths corrected for expected changes in suicide and road deaths?
  • Including untested deaths estimate would raise the infections estimate

[djay]

@pmdscully

As we don't have the complete death age data or which numerical distribution type would best match Thailand's age of COVID-19 death distribution, it's still a moot point to decide which statistical distribution to use to generate (estimate) the data. In this case I mentioned Normal (Gaussian), but as we've seen from that COVID-19 IFR modelling (i.e. for UK), it's clearly not Normal..

So, we still cannot avoid this issue.

I think I can probably use the distribution predicted by the IFR applied to the thai population. The IFR risks look like this which predicts a distribution of deaths (for bangkok) except that we are seeing 67 med rather than 80 as predicted.

I think what I can do is take the median and ages ranges seen in the actual deaths and fit this distribution to it. Then use this to work out a new IFR based on this.

Not sure yet if I need to look at the commorbidities to see if greater than other countries and if they change the distribution.

[djay]

going to close this since vaccinations change the IFR so a simple model is less and less useful