Model outcomes for multiple variants

[giovannic]

Infection (copy from #6)

Inputs:

• importation rate (per variant, time varying)

Processing:

• calculate a distribution of variants based on the importation rate on infection update categorical variable indicating infection type

to-do:

• [x] make new categorical variable object and attach to variables list. (https://github.com/mrc-ide/safir/commit/6465cf186b8350c90dc2c790c1381c606f7c43d2)

• [x] make new simulation_loop to account for that variable or modify the existing one to check. (https://github.com/mrc-ide/safir/commit/52c780d8df8f19b40116db658e37aa82646218c3)

  Outcomes

Inputs:

• infection duration (per variant)
• infectiousness (per variant)
• severity (per variant)

Processing:

• Update disease progression functions based on infected variants
• Update onward infectiousness based on infected variants
• Update hospitalisation probabilities based on infected variants

This requires modification of code in:

• [x] R/efficacy_vaccination.R: need to adjust efficacy functions to use the right parameter set for the specific strain that individual or set of individuals are infected with. (https://github.com/mrc-ide/safir/commit/0f80a270441b61dcd9447ad07bcfec4dba3406e7)
• [ ] R/events_exposure_vaccine.R: parameters$prob_hosp and parameters$prob_asymp should be potentially VOC-specific.
• [ ] R/events_hospital.R: parameters$prob_severe, parameters$prob_severe_death_no_treatment, parameters$prob_severe_death_treatment, parameters$prob_non_severe_death_treatment, parameters$prob_non_severe_death_no_treatment need to be VOC-specific.
• [ ] R/events.R: many events take parameters that should be made VOC-specific.
• [ ] need to make new parameters functions to account for these changes.

The scope of these modifications will only affect the vaccination model (not the basic versions of safir that match squire or nimue).

[slwu89]

One thing we need to clear up about modeling variants is that, given we are doing an importation approach (from development doc "e.g. on each day there is an x% of new infections due to an importation.") it is very hard to think of a reasonable way to set up vaccine efficacy against new infection against specific variants.

If each day some fraction of infections ought to belong to each variant, as a proportion of the total number of new infections, then I don't see a way to respect that parameter at the same time as allowing vaccination status to affect the probability of being infected with a specific variant. So I think we need to drop one of the two.

Everything else (disease progression, etc.) is easy to make variant-specific.

RESOLVED: use current method of proportional mix each day.

[slwu89]

Parameters to make sure are voc-specific:

• dur_E (used in events_exposure_vaccine.R)
• prob_hosp (used in events_exposure_vaccine.R): needs to be a matrix of voc x ages
• prob_asymp: same as prob_hosp

[slwu89]

closing this issue, development for mechanistic multiple strains will go on in https://github.com/mrc-ide/safir2