This is a short cheat sheet of some of the data analysis questions I keep in the back of my head.
COVID has serious lags at all stages of the information flow.
- Multi-day lag from exposure to testing
- Zero to multi-day lag from testing to result
- Negative and positive lag from testing to symptoms
- Zero to multi-day lag from new infection to start of contact tracing and testing
- Week lag from symptoms starting and potential hospitalization
- Week lag from hospitalization to ICU
- Zero to long lag from ICU to death
- Multi-week lag from ICU to discharge
- Significant lags between death and reports of death
- Widely variable death certificate reporting practices
2) Differential impacts
- COVID hits older people far harder
- Non-linear effects
- Massive current racial, ethnic, and economic disparities as a reflection of historical patterns of discrimination and power.
3) Hospital system status
- Overwhelmed hospitals have far higher age adjusted case fatality rates than not overwhelmed hospitals
- We’ve learned a lot over the first four months of the US pandemic and hospital surge
- Dex, proning, remisdivir all seem to help and reduce mortality rates and hospital duration
- We don’t know what long term looks like as we are still in the short term
- We don’t know a lot including how differential factors influence infections and outcomes (esp. with kids)
We should expect a fairly young population to have a far less severe disease course with fewer hospitalizations and deaths than a fairly old population that is identical in all aspects other than age. We should expect fatality rates to be far lower in regions with significant available and reserve staffed hospital beds then regions that are getting slammed.
Talking Points Memo had a good image of Florida’s positivity rate and case count:
The cases on 6/19 are probably generating hospitalizations at the start of July. Anything done today won’t show up in death data until after the 4th of July.
These are just some thoughts on how to think through COVID data reporting.