Interventions

This guide describes how to implementing interventions in Starsim. Interventions represent actions that affect disease transmission, such as screening, treatment, and vaccination programs.

For learning-oriented content on interventions, see Tutorial 6 - Interventions.

Comparison of intervention classes

Base classes

Class	Purpose	Key features
Intervention	Base intervention class	Eligibility checking, product integration, step() method
RoutineDelivery	Continuous delivery over time	Interpolated probabilities, annual/timestep rates
CampaignDelivery	One-off or discrete campaigns	Specific years, optional interpolation
BaseTest	Base for screening/triage	Product administration, outcome tracking
BaseScreening	Screening programs	Population-level eligibility checking
BaseTriage	Triage/follow-up testing	Targeted eligibility (e.g., screen positives)
BaseTreatment	Treatment interventions	Queue management, eligibility validation
BaseVaccination	Vaccination programs	Dose tracking, vaccination state management

Screening interventions

Class	Delivery pattern	Use case
routine_screening	Continuous	Regular screening programs (e.g., annual STI screening)
campaign_screening	Discrete	Mass screening events (e.g., outbreak response)
routine_triage	Continuous	Follow-up testing for positives
campaign_triage	Discrete	Campaign-based confirmatory testing

Treatment interventions

Class	Capacity Model	Use Case
treat_num	Fixed capacity per timestep	Resource-constrained treatment programs

Vaccination interventions

Class	Delivery Pattern	Use Case
routine_vx	Continuous	Routine immunization programs
campaign_vx	Discrete	Mass vaccination campaigns

Implementation patterns

Eligibility functions

Interventions use eligibility functions to determine who can receive the intervention:

import starsim as ss

# Lambda functions for simple eligibility
eligible_adults = lambda sim: (sim.people.age >= 18).uids

# More complex eligibility combining multiple conditions  
def high_risk_eligibility(sim):
    adults = sim.people.age >= 18
    sexually_active = sim.networks.mfnet.participant
    return (adults & sexually_active).uids

# Using intervention outcomes for triage
screen_positives = lambda sim: sim.interventions.screening.outcomes['positive']

Product integration

Interventions work with products that define the actual medical intervention:

import sciris as sc
import starsim as ss

# Define parameters
pars = sc.objdict(
    n_agents = 5e3,
    start = 2000,
    stop = 2020,
    diseases = 'sis',
    networks = 'random',
)

# Define the product data
dx_data = sc.dataframe(
    columns =
        ['disease', 'state', 'result', 'probability'],
    data = [
        ['sis', 'susceptible', 'positive', 0.01],
        ['sis', 'susceptible', 'negative', 0.99],
        ['sis', 'infected', 'positive', 0.95],
        ['sis', 'infected', 'negative', 0.05],
    ]
)

# Using built-in products
vx_start = 2005
my_vaccine = ss.simple_vx(efficacy=0.9)
vaccination = ss.routine_vx(
    product = my_vaccine,  # Product object
    prob = 0.8,
    start_year = vx_start,
)

# Using custom products
dx_start = 2010
screening = ss.routine_screening(
    product = ss.Dx(df=dx_data),
    prob = 0.9,
    start_year = dx_start,
)

# Run the sim
sim = ss.Sim(pars, interventions=[screening, vaccination])
sim.run()
sim.plot()

Initializing sim with 5000 agents
  Running 2000.01.01 ( 0/21) (0.00 s)  ———————————————————— 5%
  Running 2010.01.01 (10/21) (0.13 s)  ••••••••••—————————— 52%
  Running 2020.01.01 (20/21) (0.18 s)  •••••••••••••••••••• 100%

Figure(768x576)

Adding multiple interventions

If you want to add multiple of the same intervention, you need to give them different names. (This also applies to diseases and other modules.) This example compares simulations with 0, 1, or 2 vaccines:

import starsim as ss

# Create the product - a vaccine with 80% efficacy
vx = ss.simple_vx(efficacy=0.8)

# Create the interventions -- low and high coverage
intv_low  = ss.routine_vx(name='intv_low', start_year=2010, prob=0.2, product=vx)
intv_high = ss.routine_vx(name='intv_high', start_year=2020, prob=0.6, product=vx)

# Create the sims
pars = dict(n_agents=5000, networks='random', diseases='sis', verbose=0)
s0 = ss.Sim(pars, label='Baseline')
s1 = ss.Sim(pars, label='Low vaccine coverage', interventions=intv_low)
s2 = ss.Sim(pars, label='Low + high vaccine coverage', interventions=[intv_low, intv_high])

# Run & plot
msim = ss.parallel(s0, s1, s2)
msim.plot('sis_new_infections')

Figure(768x576)