Hospitalization/ICU/Vent utilization question

[cls3415]

This is a great tool!! Kudos!!
This more like a question not an issue.
I was looking for the build_admissions_df sub function and have a question.
So, it looks like, e.g. hospitalization number, was based on "disposition".
And the "disposition" was calculated from "i_hospitalized_v + r_hospitalized_v" in parameter.py.
May I know why the number of hospitalization = (ihosp_ratemarket_share) + (rhosp_ratemarket_share)?
Finally, thanks for hard works and efforts on this great tool!
Thanks!!

[mishmosh]

Summarized from #chime-analysis on Slack:

Answer: At that point, it’s the number of people who have ever been hospitalized (i.e. the hosp*share fraction of the people who have ever been infected).

Followup question: How do we think about patients who spend some days in hospital not ICU and some days in ICU?

Answer: Since ever infected is cumulative, the difference is equivalent to new daily admissions. The LOS parameters are used to calculate the daily census in combination with the daily admissions.

Todos:

• update end user docs (@Hanisha13)
• update code comments to clarify (up for grabs)

[cls3415]

Thank-You @mishmosh !
I mean why do we want to add r * hosp_rate * market_share where r (recovered) is from SIR model since they are recovered, which suggests they might no longer need hospital bed resource.
Thanks!

[mishmosh]

@cjbayesian can you answer this question?

[jpettit2]

I was wondering the same thing. The change from the original formulation and the rationale for the change as I have been able to find are referenced here in issue 189 I find myself of a similar mind to @cls3415 and would also be interested to see further clarification on this.

[jlubken]

@jpettit2 @cls3415 The cumulative sums of the three dispositions are shifted by their lengths of stay/days parameters and then subtracted to remove patients. Note that the infectious_days in the SIR model is independent of the days spent in each disposition for census. Likely, patients were infectious before their condition deteriorated enough to go to the hospital.

I am not a doctor, but presumably a patient could still need serious care even if they are on the recovered/recovery side (no longer infectious). The defaults of infectious_days: 14, average days on ventilator of 10, and the variable number of days before admit, certainly suggest that this is the case.

[misken]

@jpettit2 @cls3415 I put together a notebook yesterday to try to clarify for myself (and hopefully others) exactly how the resource related computations are done. It has the math that shows why the patients = raw_df.infected + raw_df.recovered line is needed and a bunch of explanation and annotated code that you might find helpful.