search

thibautjombart / covid19_bed_occupancy

Shiny app providing estimates of future bed occupancy given recent admissions
https://cmmid-lshtm.shinyapps.io/hospital_bed_occupancy_projections/
Other
12 stars 5 forks source link

Review proof of concept report #3

Closed thibautjombart closed 4 years ago

thibautjombart commented 4 years ago

It would be great to have another pair of eyes look at the model description, check the code etc.

thibautjombart commented 4 years ago

Report is now ready as of 9349c446c3c2b815bd8fdff72901138c4457d5e8

Best compiled freshly using the factory, but attaching an html output just in case.

proof_of_concept_2020-03-22.html.zip

thibautjombart commented 4 years ago

If you are not part of the author list and your review this report, please add yourselves.

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L40

Should we make this "ICU" or "critical care" bed occupancy? It's not total hospital bed occupancy for example. Also relevant for the title? Just need to be consistent.

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L46

Perhaps rewrite as using a "given doubling time (with user-specified upper and lower bounds), feeding into a log-linear model"

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L49

Again - is this length of hospitalisation or time in ICU?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L51

How many times is many times? can we give some ballparks here e.g < 10admissions, 1000 simulations. Or is there a direct relationship to the size of the uncertainty?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L64

Are we allowing users to change this?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L139

Think this is a copy/paste inclusion from the other work - I don't think you use death data to get number of cases but actual number of cases in ICU data?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L144

For "admissions" throughout it needs to be "ICU" admissions? Perhaps to save characters there could be a statement at the top saying that here all beds are critical care ones and all admissions are to the ICU? to prevent having to constantly repeat this.

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L165

Shape and scale values missing - should be 2 and 13?

Also why is this, non critical and critical if we are just doing critical here? And don't we use these as two distributions of the same thing?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L229

Going foward, I wonder how important the reporting parameter will be. Especially for users inputting their own setting data e.g. for a hospital they are likely to know their total numbers. But I guess at a regional level it may still be low.

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L264

Does "reporting" need to be an input to the function - perhaps also with a sanity check that it is between 0 and 1 (e.g. not an inputted percentage)?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L329

Spelling mistake? should be "discrete"?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L321

Is 10 simulations enough? where can this be changed / guessing we wouldn't allow user to change?

gwenknight commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L2

Just looking at a report for UCLH - they talk about "bed days". Perhaps this is the better technical term... However, they look over a month where as we are giving daily values so I think we are OK. If we do any aggregation we should talk about "occupied bed days" - think there is a table in another version.

thibautjombart commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L40

Should we make this "ICU" or "critical care" bed occupancy? It's not total hospital bed occupancy for example. Also relevant for the title? Just need to be consistent.

This version can apply to any kind of hospitalisation really, as long as it is covid19

esnightingale commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L264

Does "reporting" need to be an input to the function - perhaps also with a sanity check that it is between 0 and 1 (e.g. not an inputted percentage)?

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L229

Going foward, I wonder how important the reporting parameter will be. Especially for users inputting their own setting data e.g. for a hospital they are likely to know their total numbers. But I guess at a regional level it may still be low.

Can we suggest any way they might roughly ascertain reporting level? Or would it just be hypothetical? Also agree with Gwen's next comment - can we define it as e.g. "% reported" so the scale is clearer?

thibautjombart commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L64

Are we allowing users to change this?

In the app, yes I think we should, eventually. There will be a need to settle for a kind of distribution, so the group may want to discuss this. Might not be needed for the first release though.

thibautjombart commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L264

Does "reporting" need to be an input to the function - perhaps also with a sanity check that it is between 0 and 1 (e.g. not an inputted percentage)?

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L229

Going foward, I wonder how important the reporting parameter will be. Especially for users inputting their own setting data e.g. for a hospital they are likely to know their total numbers. But I guess at a regional level it may still be low.

Can we suggest any way they might roughly ascertain reporting level? Or would it just be hypothetical? Also agree with Gwen's next comment - can we define it as e.g. "% reported" so the scale is clearer?

Damn yes, good catch! Will fix right now + sanity checks.

esnightingale commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L40

Should we make this "ICU" or "critical care" bed occupancy? It's not total hospital bed occupancy for example. Also relevant for the title? Just need to be consistent.

This version can apply to any kind of hospitalisation really, as long as it is covid19

Maybe specify this at the beginning then? That the approach can be applied to non-critical or critical covid19 admissions, just need to tailor the LOS distribution accordingly?

thibautjombart commented 4 years ago

https://github.com/thibautjombart/covid19_bed_occupancy/blob/9349c446c3c2b815bd8fdff72901138c4457d5e8/model/report_sources/proof_of_concept_2020-03-22.Rmd#L64

Are we allowing users to change this?

I think at some point yes, but maybe not for the initial release?