Single vaccine

[giovannic]

Todo:

• [ ] Parameterise a vaccine
• [ ] Parameterise a vaccine strategy
• [ ] Create variable to track time of last vaccination(s)
• [ ] Create vaccine distribution process
• [ ] Modify infection process xply prob_infection and prob_severe by vaccine efficacy for each individual

[pwinskill]

Nimue currently has vaccinated individuals going through 2 compartments before losing vaccine protection (if duration of vaccine derived immunity is <Inf). We assume immunity last for the longest period determined by natural expsoure (if it has occured) or vaccination.

I'm wondering if we want to model each vaccine with a very flexible set of (age-dependent) characteristics that could include any combination of the following:

• Infection blocking
• Reduction in hospitalisation rate
• Reduction in onwards transmission (independent to reductions in durations/infection etc?).
• Reduction in duration of severe illness
• Reduction in duration of mild illness
• Increase in the proportion of mild cases that are symptomatic

[pwinskill]

Some of the above, may be version 3 considerations, but it might be easier to code the at the same time with null defaults

[pwinskill]

After futher discussion with OJ, proposed intial appraoch would be to include a flexible vaccine parameterisation, first to capture reductions in the probability of infection and/or the probability of severe case.

This would be implemented by allowing the user to specify an efficacy curve (for each mode of action) as a look up table. When implementing, for vaccinated individuals we will estimate the time since fully vaccinated (vx_t) and use that value to look up the vaccine-impact from the appropriate table that will then be used to modify prob_infection or prob_severe

[giovannic]

Sweet (I've put these tasks in the header)

I still have some questions around distribution strategy. Are we gonna do like nimue?

Parameterise vaccine

We need to take:

• efficacy_curve, a vector of efficacy values for an arbitrary number of timesteps after vaccination

Parameterise vaccine strategy (I'm guessing here)

We need to take:

• prob_vaccination, the probability of each age group getting the vaccination
• n_vaccine, the number of available vaccines available at tt_vaccine
• tt_vaccine, the timesteps when n_vaccine become available

[pwinskill]

Yes, I think the distribution part may be the trickiest bit. I think some aspects we should retain in the nimue style, namely, for each vaccine implemented:

1. Allow the user to define a population level, time varying vaccination rate: doses per unit time
2. Define a prioritisation matrix - I think the nimue approach of a age X prioritisation step coverage matrix works well
3. Translate 1 and 2 to vaccination of individuals in appropriate age groups at appropriate times in the IBM

[giovannic]

Vaccine allocation:

• variable for prioritisation block (how do we assign this?) Perhaps with vectors for:

  • age ranges
  • occupation
  • time period (tbc by @pwinskill )

• Exclude people who are in hospital or sick

Subsequent doses:

• Must count towards total available doses
• Deterministic delay time for booster doses
• Strategy allows you to decide if subsequent doses are affected by sickness
• There is a dropout based on bernoulli draws

[giovannic]

Ok so. For allocation we should have:

• We should be able to parameterise eligibility over time. Eligibility can be based on:
  • age range
  • hospital worker (yes/no) - these variables have to be sampled from data that will soon arrive in nimue
  • has comorbidites (yes/no) - these variables have to be sampled from data that will soon arrive in nimue

There will be an age based coverage/hesitancy which shouldn't be exceeded.

Each timestep, eligible individuals will be uniformly allocated doses (maintaining coverage and total availablity)

[pwinskill]

We might want to include:

• Vaccine eligable and/or vaccine hesitant

as well. This would allow us to capture those who cannot or will not have the vaccine