Partial vaccinated data from CDPH going back to late 2020 have been revised upward significantly.

klenwell / covid-19

Python command-line application to collect and analyze COVID-19 data.

0 stars 0 forks source link

Partial vaccinated data from CDPH going back to late 2020 have been revised upward significantly. #55

Closed klenwell closed 2 years ago

klenwell commented 2 years ago

I noticed it with this week's update. It can be seen in the data/oc/oc-immunity.csv file.

Compare Partial Vax column (#6) numbers beginning at 12/21/2020 here:

https://github.com/klenwell/covid-19/commit/628a79e079775cfb3fa1e7036311d5a238ba87cd?diff=split#diff-a7ab2d6d9795098fb37370ce91b481a695d83264c19ee0d244664e0e8c42a943L454-L459

For example, look at how counts changed for 12/15 - 12/21/2020:

3/25 commit 628a79e: 24, 464, 4036, 5475, 2324, 2053, 4468
3/20 commit 5015411: 0, 6, 46, 69, 36, 32, 58

More info on how this data is collected and used to estimate county-wide immunity here:

https://github.com/klenwell/covid-19/pull/54

klenwell commented 2 years ago

I had some questions about how these values were defined when I updated my immunity model. Did CDPH change the way they are tracked this week? In PR, I noted:

Model assumes, when second shot is reported, CDPH decrements PV and increments FV counts on the date of the first shot.

Maybe that has changed?

Here's the impact on the graph:

Before this week's update Screenshot from 2022-03-25 16-41-15

After this week's update Screenshot from 2022-03-25 16-40-54

I suspect that this is not accurate (or more ridiculously inaccurate than the before version) and the model needs to be modified.

klenwell commented 2 years ago

I restored the second shot adjustments to immunity model: https://github.com/klenwell/covid-19/commit/a3c2da68473dd3fb6562f0bcc346f6759c5d6fd8

The new graph looks much more like the original before CDPH changed its reporting data:

Before Screenshot from 2022-04-02 14-17-33

After Screenshot from 2022-04-02 14-18-03

klenwell commented 2 years ago

Solved

Problem

CDPH changed reporting methodology so that partial vaxxed and fully vaxxed were counted in full on dates that a person received first and second shot respectively. This led to double-counting of vaccinated individuals and an over-estimates of population-wide immunty.

Solution

Immunity model was updated to offset fully vaccinated counts by decrementing partial vaccination counts on earlier dates. See changes in this PR: https://github.com/klenwell/covid-19/pull/56