Immunity rework #24

RHull-IDM · 2025-08-06T06:26:48Z

Updated vaccination intervention which now leverages the starsim.Intervention class.

New features:

Configurable start and end dates, which can be passed in as date strings or decimal years, e.g. "2001-01-01" or 2001.0
Vaccine explicit strains + multivalence: Can now incorporate as many strains as desired in vaccine if provided manually, or will default to most dominant single strain if none provided
Immunity waning: vaccines can now be configured with a waning rate and delay

Tests have been added for each of these features as well.

Closes #20

Note: I still need to update the baseline test values before merging.

… handles waning and multivalent vaccines, as well as flexible scheduling. While the waning does work as expected internally, it is not yet connected to the immunity model.

# Conflicts: # rotasim/rotasim_genetics.py

…o update severe chance.

…o update Rota class to always use the vaccine interventions.

…rain

…olved

…lasses because they're too inflexible

…ivery. The delivery class was too inflexible an incomplete, and this simplifies the code as well. Added tests for each of the new intervention features.

…recoveries is greater than number of infecteds

…d test_alt to match

daniel-klein

Hi @RHull-IDM, I added a few comments for your consideration. My quick review focused on the new interventions.py file rather than changes to rotasim or rotasim_genetics. Thanks!

rotasim/interventions.py

daniel-klein · 2025-08-06T16:38:05Z

rotasim/interventions.py

+        ppl = self.sim.people
+        if self.sim.ti in self.timepoints:
+            if self.sim.ti == self.timepoints[0]:
+                # if there is no strain associated with the product, select the most prevalent strain


So in different sims (e.g. change rand_seed or other parameters), the vaccine might contain different antigens, based on prevalence at the first timepoint of the vaccine? This feels error-prone, consider requiring users to specify g and p.

This maintains the behavior of the original model. If @Suvanthee or Kyra or Alicia want to chime in I'm happy to change to whatever makes sense.

The intended behavior was to

Let the model reach an equilibrium for the first 5 years

At 15 years pick the vaccine strain as the most prevalent strain in the last 10 years (skipping the first 5 years where the model is in the initialization phase)

Once the vaccine strain is picked we will not change it and use it in the future for both doses.

I think we should add the behavior where we skip the first 5 (time_to_equilibrium) years because at the start the simulation is very volatile similar to here.

For different seeds and other parameters we may indeed see different strains being picked. Our goal here was to study the effect of targeting the most prevalent strain though a vaccine in many settings. I think the current code achieves that.

An option that allows users to specify the strain would be a great addition too.

Suvanthee is correct that the original intention was to choose the most prevalent strain, even if that change from iteration to iteration. We asked Ryan to add the option to specify vx_strains so that we can maintain consistency in the vaccine strain across iterations as needed (e.g., if want to model specifically the effects of a G1P8 vaccine like Rotarix) - and expect that we will primarily use this functionality moving forward, but retain default behavior of selecting most prevalent strain if vx_strains are not specified.

daniel-klein · 2025-08-06T16:42:02Z

rotasim/rotasim.py

+                vx_intv_found = True
+                break
+
+        if not vx_intv_found:


Is it possible to run a simulation without the RotaVaxProg?

It was previously an integral part of the model and was not separate from the rest of the model logic, so I implemented it this way to maintain consistency. If there's a case for not using the vaccination program I will make it optional.

It would be nice to have this functionality. Earlier we had a vaccine_hypothesis where this was skipped. But I now realize that I got rid of it while removing the vaccine_hypothesis for simplifying the model

Suvanthee

Hi @RHull-IDM This looks great. I am slowly going through the changes. left some comments.

Suvanthee · 2025-08-26T03:07:42Z

rotasim/interventions.py

+        ppl = self.sim.people
+        if self.sim.ti in self.timepoints:
+            if self.sim.ti == self.timepoints[0]:
+                # if there is no strain associated with the product, select the most prevalent strain


The intended behavior was to

Let the model reach an equilibrium for the first 5 years

At 15 years pick the vaccine strain as the most prevalent strain in the last 10 years (skipping the first 5 years where the model is in the initialization phase)

Once the vaccine strain is picked we will not change it and use it in the future for both doses.

I think we should add the behavior where we skip the first 5 (time_to_equilibrium) years because at the start the simulation is very volatile similar to here.

For different seeds and other parameters we may indeed see different strains being picked. Our goal here was to study the effect of targeting the most prevalent strain though a vaccine in many settings. I think the current code achieves that.

An option that allows users to specify the strain would be a great addition too.

Suvanthee · 2025-08-26T03:38:30Z

rotasim/rotasim_genetics.py

-                ve_s = self.vaccine_efficacy_s_d2[pathogen_strain_type]
-            else:
-                raise NotImplementedError(f"Unsupported vaccine dose: {vaccine.dose}")
+            ve_s = vaccine.product.vaccine_efficacy_s[n_doses][pathogen_match]


I think the waning immunity should also affect the protection against severe infection. Previsously we only applied the severity reduction when the vaccine immunity has not waned (see if vaccine is not None and not vaccine.is_waned: above).

I think we need to apply the same reduction for the probability of getting a severe case as for vaccine immunity here.

Agreed - behavior should be to apply the same waning multiplicative factor to VE_i and VE_s (or wane total VE and then use the same ratio to calculate VE_i and VE_s at each time step)

Added waning to the severity calculation.

Suvanthee · 2025-08-26T03:46:55Z

rotasim/interventions.py

+
+    def breakdown_vaccine_efficacy(self, ve, x):
+        """
+        Break down the vaccine efficacy into its components


I will send a description for this tomorrow.

Suvanthee · 2025-08-26T03:49:32Z

rotasim/rotasim.py

+                vx_intv_found = True
+                break
+
+        if not vx_intv_found:


It would be nice to have this functionality. Earlier we had a vaccine_hypothesis where this was skipped. But I now realize that I got rid of it while removing the vaccine_hypothesis for simplifying the model

RHull-IDM · 2025-10-20T18:28:27Z

Ended up going a different direction with a full refactor to work better with the latest Starsim version

RHull-IDM added 15 commits July 18, 2025 17:01

Added a new vaccination class based on Starsim's routine_vx class. It…

3dbc205

… handles waning and multivalent vaccines, as well as flexible scheduling. While the waning does work as expected internally, it is not yet connected to the immunity model.

added vx_intervention test

edee11f

Merge branch 'main' into immunity_rework

198ac9e

# Conflicts: # rotasim/rotasim_genetics.py

lots of changes. close to working but still not running. still need t…

01b9c82

…o update severe chance.

everything seems to work now, but more testing is needed. Also need t…

c439b04

…o update Rota class to always use the vaccine interventions.

moved vx check to top of contact event

f3c0639

fixed infecting probability in cases where more than one infecting st…

d0c1856

…rain

save rel_beta in local par to avoid constant lookups

659e470

moved fitness check to prob_variant_infect_host

e3615ee

fixed strain ref

f1ba7f1

added start year to params

27acd09

check for no doses first when chcking vaccine immunity

af9ba2e

rolled back some of the complexity, performance issues seem to be res…

04fc264

…olved

committing current state but going to move away from the starsim vx c…

95b42e9

…lasses because they're too inflexible

RotaVaxProg now inherits only from BaseVaccination and not RoutineDel…

682e5e1

…ivery. The delivery class was too inflexible an incomplete, and this simplifies the code as well. Added tests for each of the new intervention features.

RHull-IDM requested a review from Suvanthee August 6, 2025 06:26

RHull-IDM added 6 commits August 5, 2025 23:34

pinning starsim to version 2.3 for now

a16b03e

fixed vax initialization order issue and handled edge case where num …

9557c4d

…recoveries is greater than number of infecteds

added more parameters to method signature of intervention, and update…

5be1b59

…d test_alt to match

fixed types on tests

60ed169

suppressed some of the excessive warnings if verbose not enabled

7ef6182

updated baseline events files

d74e8e6

daniel-klein reviewed Aug 6, 2025

View reviewed changes

RHull-IDM added 7 commits August 6, 2025 11:28

cleaned up parameter initialization

dba86df

more cleanup

b5bdf7f

fixed initialization

fc97331

added ability to pass Product directly into the RotaVaxProg constructor

7fed4cc

fixed bug in selection of agents eligible for vaccination

0fc99e5

renamed 'to_vx' to 'is_vx_eligible'

f990ae5

updated baselines

3c7349b

Suvanthee reviewed Aug 26, 2025

View reviewed changes

RHull-IDM added 2 commits August 27, 2025 14:02

added waned_effectiveness to severity calc

bd3473d

added description for breakdown_vaccine_efficacy

c7f987d

RHull-IDM closed this Oct 20, 2025

Immunity rework #24

Immunity rework #24

Uh oh!

Conversation

RHull-IDM commented Aug 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

daniel-klein left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Suvanthee left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

RHull-IDM commented Oct 20, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

RHull-IDM commented Aug 6, 2025 •

edited

Loading