Interventions Reference¶

JAX-compatible intervention functions that modify transmission rates during simulation. These support both proportion-based triggering (activate when infection prevalence crosses a threshold) and date-window triggering (activate within a calendar date range).

`interventions` ¶

JAX-compatible intervention functions for modifying transmission rates.

This module provides two intervention dispatch strategies that can be applied during simulation:

Proportion-based (jax_prop_intervention): activates or deactivates interventions when the proportion of infective individuals crosses configurable thresholds.
Timestep-based (jax_timestep_intervention): activates interventions within a calendar date window specified as ordinal days.

Both functions support Dengue vector-borne interventions (physical / chemical, targeting b_V_T / s_V_T rates) and COVID-style interventions (social isolation, vaccination, mask wearing targeting beta, plus lockdown which replaces the travel matrix with an identity matrix).

`jax_prop_intervention(intervention_dict, prop_infectives, rates, intervention_statuses, travel_matrix)` ¶

Apply proportion-based interventions that trigger on infection prevalence.

Each intervention activates when prop_infectives rises above its start_threshold and deactivates when it falls below its end_threshold. While active, the targeted transmission rate is reduced by adherence_min * transmission_percentage.

Supported intervention keys (looked up in intervention_dict):

"physical" / "chemical" -- Dengue vector-borne interventions targeting the b_V_T (biting) and s_V_T (spraying) rates.
"social_isolation" / "vaccination" / "mask_wearing" -- COVID-style interventions targeting the beta rate.
"lock_down" -- replaces the travel matrix with an identity matrix to eliminate inter-zone movement.

Parameters:

Name	Type	Description	Default
`intervention_dict`		Mapping of intervention name to its config dict. Each config must contain `adherence_min` and `transmission_percentage`, and optionally `start_threshold` and `end_threshold`.	required
`prop_infectives`	`float`	Current proportion of infective individuals in the population (0.0--1.0).	required
`rates`	`dict`	Mutable dict of current transmission rate values keyed by variable name (e.g. `"beta"`, `"b_V_T"`).	required
`intervention_statuses`	`dict`	Mutable dict tracking the on/off boolean status of each intervention by name.	required
`travel_matrix`	`ndarray`	Square JAX array representing inter-zone travel. Replaced with an identity matrix when lockdown is active.	required

Returns:

Type	Description
	A 3-tuple of `(rates, intervention_statuses, travel_matrix)`
	with updated values reflecting any interventions that were
	activated or deactivated this step.

Source code in compartment/interventions.py

def jax_prop_intervention(intervention_dict,
                                   prop_infectives: float,
                                   rates: dict,
                                   intervention_statuses: dict,
                                   travel_matrix: jnp.ndarray):
    """Apply proportion-based interventions that trigger on infection prevalence.

    Each intervention activates when ``prop_infectives`` rises above its
    ``start_threshold`` and deactivates when it falls below its
    ``end_threshold``.  While active, the targeted transmission rate is
    reduced by ``adherence_min * transmission_percentage``.

    Supported intervention keys (looked up in *intervention_dict*):

    - ``"physical"`` / ``"chemical"`` -- Dengue vector-borne interventions
      targeting the ``b_V_T`` (biting) and ``s_V_T`` (spraying) rates.
    - ``"social_isolation"`` / ``"vaccination"`` / ``"mask_wearing"`` --
      COVID-style interventions targeting the ``beta`` rate.
    - ``"lock_down"`` -- replaces the travel matrix with an identity
      matrix to eliminate inter-zone movement.

    Args:
        intervention_dict: Mapping of intervention name to its config dict.
            Each config must contain ``adherence_min`` and
            ``transmission_percentage``, and optionally ``start_threshold``
            and ``end_threshold``.
        prop_infectives: Current proportion of infective individuals in
            the population (0.0--1.0).
        rates: Mutable dict of current transmission rate values keyed by
            variable name (e.g. ``"beta"``, ``"b_V_T"``).
        intervention_statuses: Mutable dict tracking the on/off boolean
            status of each intervention by name.
        travel_matrix: Square JAX array representing inter-zone travel.
            Replaced with an identity matrix when lockdown is active.

    Returns:
        A 3-tuple of ``(rates, intervention_statuses, travel_matrix)``
        with updated values reflecting any interventions that were
        activated or deactivated this step.
    """
    def _update_one(comp_rate, status, cfg):
        start_th = cfg.get("start_threshold")
        end_th   = cfg.get("end_threshold")
        adh      = cfg["adherence_min"]
        reduc    = cfg["transmission_percentage"]

        if start_th is None:
            turn_on = jnp.bool_(False)
        else:
            turn_on = jnp.logical_and(prop_infectives >= start_th,
                                      jnp.logical_not(status))

        if end_th is None:
            turn_off = jnp.bool_(False)
        else:
            turn_off = jnp.logical_and(prop_infectives <= end_th,
                                       status)

        new_status = jnp.where(turn_on,
                               True,
                               jnp.where(turn_off, False, status))

        reduced_rate = comp_rate * (1.0 - adh * reduc)
        new_rate     = jnp.where(new_status, reduced_rate, comp_rate)
        return new_rate, new_status

    # Update the physical intervention if defined
    if "physical" in intervention_dict:
        new_b, new_flag = _update_one(
            rates["b_V_T"],
            intervention_statuses["physical"],
            intervention_dict["physical"]
        )
        rates = {**rates, "b_V_T": new_b} 
        intervention_statuses = {**intervention_statuses,
                                 "physical": new_flag}

    # Update the chemical intervention if defined
    if "chemical" in intervention_dict:
        new_s, new_flag = _update_one(
            rates["s_V_T"],
            intervention_statuses["chemical"],
            intervention_dict["chemical"]
        )
        rates = {**rates, "s_V_T": new_s}
        intervention_statuses = {**intervention_statuses,
                                 "chemical": new_flag}

    # COVID interventions
    for name in ("social_isolation", "vaccination", "mask_wearing"):
        if name in intervention_dict and "beta" in rates:
            new_beta, new_flag = _update_one(
                rates["beta"],
                intervention_statuses.get(name, False),
                intervention_dict[name]
            )
            rates = {**rates, "beta": new_beta}
            intervention_statuses = {**intervention_statuses, name: new_flag}

    # COVID lock down
    if "lock_down" in intervention_dict:
        cfg = intervention_dict["lock_down"]
        # reuse _update_one to compute the on/off status - rate is irrelevant
        _, new_flag = _update_one(
            0.0, 
            intervention_statuses.get("lock_down", False),
            cfg
        )

        new_flag_scalar = jnp.any(new_flag)
        travel_matrix = jnp.where(new_flag_scalar,
                                  jnp.eye(travel_matrix.shape[0]),
                                  travel_matrix)
        intervention_statuses = {**intervention_statuses, "lock_down": new_flag_scalar}

    return rates, intervention_statuses, travel_matrix

`jax_timestep_intervention(intervention_dict, current_ordinal_day, rates, intervention_statuses, travel_matrix)` ¶

Apply date-window interventions that activate within a calendar range.

Each intervention is active when current_ordinal_day falls within its [start_date_ordinal, end_date_ordinal] window. While active the targeted transmission rate is reduced by adherence_min * transmission_percentage.

Supported intervention keys are the same as :func:jax_prop_intervention (physical, chemical, social_isolation, vaccination, mask_wearing, lock_down), but activation is driven by calendar dates instead of prevalence thresholds.