Merge pull request #185 from pph-collective/develop

Main

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "titan-model"
-version = "2.5.0"
+version = "2.5.1"
 description = "TITAN Agent Based Model"
 license = "GPL-3.0-only"
 authors = ["Sam Bessey <[email protected]>", "Mary McGrath <[email protected]>"]

tests/integration_test.py

@@ -199,12 +199,13 @@ def test_target_partners(make_model_integration, tmpdir):
 
     # change the partner distribution mean upward for creating model b
     for bond in model_a.params.classes.bond_types:
-        model_a.params.demographics.black.sex_type.MSM.drug_type["None"].num_partners[
-            bond
-        ].vars[1].value *= 10
-        model_a.params.demographics.black.sex_type.MSM.drug_type["Inj"].num_partners[
-            bond
-        ].vars[1].value *= 10
+        for race in model_a.params.classes.races:
+            model_a.params.demographics[race].sex_type.MSM.drug_type[
+                "None"
+            ].num_partners[bond].vars[1].value *= 10
+            model_a.params.demographics[race].sex_type.MSM.drug_type[
+                "Inj"
+            ].num_partners[bond].vars[1].value *= 10
     model_a.params.model.seed.run = model_a.run_seed
     model_a.params.model.seed.ppl = model_a.pop.pop_seed
 
@@ -305,8 +306,10 @@ def test_syringe_services(params_integration, tmpdir):
     """
     If we use syringe services, does the incidence of hiv decrease?
     """
-    params_integration.demographics.black.sex_type.MSM.drug_type.Inj.ppl = 1.0
-    params_integration.demographics.black.sex_type.MSM.drug_type["None"].ppl = 0.0
+    for race in params_integration.classes.races:
+        params_integration.demographics[race].sex_type.MSM.drug_type.Inj.ppl = 1.0
+        params_integration.demographics[race].sex_type.MSM.drug_type["None"].ppl = 0.0
+
     params_integration.model.num_pop = 500
     model_a = TITAN(params_integration)
     model_a.params.partnership.sex.frequency.Sex = (
@@ -435,3 +438,175 @@ def test_incar(params_integration, tmpdir):
     # agents should not hiv convert during incar
     for agent in time_1_incar_hiv_neg:
         assert not agent.hiv.active
+
+
+@pytest.mark.integration_stochastic
+def test_assort_mix(params_integration, tmpdir):
+    """
+    Do vastly different assorting rules result in different networks
+    """
+    path_a = tmpdir.mkdir("a")
+    path_a.mkdir("network")
+    path_b = tmpdir.mkdir("b")
+    path_b.mkdir("network")
+    path_c = tmpdir.mkdir("c")
+    path_c.mkdir("network")
+
+    params_integration.features.assort_mix = True
+    params_integration.assort_mix = ObjMap(
+        {
+            "same_race": {
+                "attribute": "race",
+                "partner_attribute": "__agent__",
+                "bond_types": [],
+                "agent_value": "__any__",
+                "partner_values": {"__same__": 0.9, "__other__": 0.1},
+            }
+        }
+    )
+
+    model_a = TITAN(params_integration)
+    model_a.run(path_a)
+
+    params_integration.assort_mix = ObjMap(
+        {
+            "cross_race": {
+                "attribute": "race",
+                "partner_attribute": "__agent__",
+                "bond_types": [],
+                "agent_value": "__any__",
+                "partner_values": {"__same__": 0.5, "__other__": 0.5},
+            }
+        }
+    )
+
+    model_b = TITAN(params_integration)
+    model_b.run(path_b)
+
+    params_integration.features.assort_mix = False
+    model_c = TITAN(params_integration)
+    model_c.run(path_c)
+
+    model_a_rels = model_a.pop.relationships
+    a_same_race = sum([1 for r in model_a_rels if r.agent1.race == r.agent2.race])
+    a_diff_race = sum([1 for r in model_a_rels if r.agent1.race != r.agent2.race])
+
+    model_b_rels = model_b.pop.relationships
+    b_same_race = sum([1 for r in model_b_rels if r.agent1.race == r.agent2.race])
+    b_diff_race = sum([1 for r in model_b_rels if r.agent1.race != r.agent2.race])
+
+    model_c_rels = model_c.pop.relationships
+    c_same_race = sum([1 for r in model_c_rels if r.agent1.race == r.agent2.race])
+    c_diff_race = sum([1 for r in model_c_rels if r.agent1.race != r.agent2.race])
+
+    assert a_same_race > a_diff_race
+    assert a_same_race > b_same_race
+    assert b_diff_race > a_diff_race
+    assert math.isclose(b_same_race, c_same_race, abs_tol=50)
+
+    # close to proportion expected
+    assert math.isclose(0.9, a_same_race / (a_same_race + a_diff_race), abs_tol=0.05)
+    assert math.isclose(0.5, b_same_race / (b_same_race + b_diff_race), abs_tol=0.1)
+    assert math.isclose(0.5, c_same_race / (c_same_race + c_diff_race), abs_tol=0.1)
+
+
+@pytest.mark.integration_stochastic
+def test_treatment_cascade(params_integration, tmpdir):
+    """
+    Does an increase in HAART result in less HIV, AIDS and death
+    """
+    path_a = tmpdir.mkdir("a")
+    path_a.mkdir("network")
+    path_b = tmpdir.mkdir("b")
+    path_b.mkdir("network")
+
+    params_integration.features.die_and_replace = True
+    params_integration.model.num_pop = 1000
+    params_integration.model.time.num_steps = 20
+
+    def run_get_death(model, path):
+        deaths = set()
+        hiv_start = set(a.id for a in model.pop.all_agents if a.hiv.active)
+        aids_start = set(a.id for a in model.pop.all_agents if a.hiv.aids)
+        hiv_only_start = hiv_start - aids_start
+        unique_hiv = set()
+        unique_aids = set()
+        while model.time < model.params.model.time.num_steps:
+            model.time += 1
+            model.step(path)
+            deaths |= set(a.id for a in model.deaths)
+            unique_hiv |= set(
+                a.id
+                for a in model.pop.all_agents
+                if a.hiv.active and a.hiv.time == model.time
+            )
+            unique_aids |= set(a.id for a in model.pop.all_agents if a.hiv.aids)
+            model.reset_trackers()
+
+        return (
+            len(unique_hiv - unique_aids - hiv_only_start),
+            len(unique_aids - aids_start),
+            len(deaths & (unique_hiv | hiv_start)),
+        )
+
+    model_a = TITAN(params_integration)  # low haart
+    new_hiv_a, new_aids_a, hiv_deaths_a = run_get_death(model_a, path_a)
+
+    for race in params_integration.classes.races:
+        for drug_type in params_integration.classes.drug_types:
+            haart_params = (
+                params_integration.demographics[race]
+                .sex_type.MSM.drug_type[drug_type]
+                .haart
+            )
+            haart_params.init = 0.9
+            haart_params.enroll.rule.prob = 0.9
+            haart_params.adherence.init = 0.9
+            haart_params.adherence.prob = 0.9
+
+    model_b = TITAN(params_integration)  # high haart
+    new_hiv_b, new_aids_b, hiv_deaths_b = run_get_death(model_b, path_b)
+
+    assert hiv_deaths_b <= hiv_deaths_a, "HIV deaths not down"
+    assert new_aids_b <= new_aids_a, "new AIDS  not down"
+    assert new_hiv_b < new_hiv_a, "new HIV not down"
+
+
+@pytest.mark.integration_stochastic
+def test_pca_awareness(params_integration, tmpdir):
+    """
+    Does a change in the threshhold for PCA result in more knowledge?
+    """
+    params_integration.exposures.knowledge = True
+    params_integration.knowledge.init = 0.1
+    params_integration.model.time.num_steps = 20
+
+    path_a = tmpdir.mkdir("a")
+    path_a.mkdir("network")
+    path_b = tmpdir.mkdir("b")
+    path_b.mkdir("network")
+
+    params_integration.knowledge.threshold = 3.5
+    model_a = TITAN(params_integration)  # low knowledge
+
+    init_active_a = sum([1 for a in model_a.pop.all_agents if a.knowledge.active])
+    init_prep_a = sum([1 for a in model_a.pop.all_agents if a.prep.active])
+    model_a.run(path_a)
+    end_active_a = sum([1 for a in model_a.pop.all_agents if a.knowledge.active])
+    end_prep_a = sum([1 for a in model_a.pop.all_agents if a.prep.active])
+
+    params_integration.knowledge.threshold = 1.5
+    params_integration.knowledge.prob = 0.05
+    model_b = TITAN(params_integration)  # high knowledge
+
+    init_prep_b = sum([1 for a in model_b.pop.all_agents if a.prep.active])
+    init_active_b = sum([1 for a in model_b.pop.all_agents if a.knowledge.active])
+    model_b.run(path_b)
+    end_active_b = sum([1 for a in model_b.pop.all_agents if a.knowledge.active])
+    end_prep_b = sum([1 for a in model_b.pop.all_agents if a.prep.active])
+
+    assert math.isclose(init_active_b, init_active_a, abs_tol=20)
+    assert (end_active_b - init_active_b) > (end_active_a - init_active_a)
+
+    assert math.isclose(init_prep_b, init_prep_a, abs_tol=20)
+    assert (end_prep_b - init_prep_b) > (end_prep_a - init_prep_a)

tests/model_test.py

@@ -34,14 +34,10 @@ def test_update_all_agents(make_model, make_agent):
     # make agent 0
     model = make_model()
     assert model.params.agent_zero.interaction_type == "injection"
-    a = make_agent(race="white", DU="Inj")
-    p = make_agent(race="white", DU="Inj")
-    # make sure at least 1 relationship is compatible with agent 0 type
-    Relationship(a, p, 10, bond_type="Inj")
     model.time = 1
+    model.params.features.die_and_replace = False
     # update all agents passes with agent 0
     model.update_all_agents()
-    assert a.knowledge.active is False
 
     # remove all bonds compatible with agent 0. agent 0 fails
     for rel in copy(model.pop.relationships):
@@ -58,6 +54,7 @@ def test_update_all_agents(make_model, make_agent):
 
     with pytest.raises(ValueError) as excinfo:
         model.update_all_agents()
+
     assert "No agent zero!" in str(excinfo)
 
 

tests/params/basic.yml

@@ -42,11 +42,11 @@ demographics:
             haart:
               init: 0.1
               enroll: &haart_prob
-                0:
+                begin:
                   prob: 0.1
                   start: 0
                   stop: 10
-                1:
+                end:
                   prob: 1.0
                   start: 10
                   stop: 100

tests/params/simple_integration.yml

@@ -2,6 +2,8 @@ classes:
   races:
     black:
       hispanic: false
+    white:
+      hispanic: false
   sex_types:
     MSM:
       gender: M
@@ -13,8 +15,8 @@ classes:
     - None
 
 demographics:
-  black:
-    ppl: 1.0
+  black: &race_dems
+    ppl: 0.5
     sex_type:
       MSM:
         ppl: 1.0
@@ -46,18 +48,18 @@ demographics:
         drug_type:
           Inj:
             ppl: 0.2
-            hiv:
-              init: 0.1
+            hiv: &base_hiv
+              init: 0.2
               dx:
-                init: 0.1
-                prob: 0.1
+                init: 0.9
+                prob: 0.9
               aids:
                 init: 0.1
-            haart:
+            haart: &base_haart
               init: 0.1
-              prob:
-                1:
-                  enroll: 0.1
+              enroll:
+                rule:
+                  prob: 0.1
                   start: 0
                   stop: 999
               adherence:
@@ -91,31 +93,16 @@ demographics:
                     type: int
           None:
             ppl: 0.8
-            hiv:
-              init: 0.1
-              dx:
-                init: 0.1
-                prob: 0.1
-              aids:
-                init: 0.1
-            haart:
-              init: 0.1
-              prob:
-                1:
-                  enroll: 0.1
-                  start: 0
-                  stop: 999
-              adherence:
-                init: 0.1
-                prob: 0.1
-              discontinue: 0.1
+            hiv: *base_hiv
+            haart: *base_haart
             num_partners:
               Sex:
                 dist_type: poisson
                 vars:
                   1:
                     value: 2
                     type: float
+  white: *race_dems
 
 partnership:
   duration:
@@ -144,6 +131,13 @@ partnership:
           prob: 1.0
           min: 1
           max: 10
+    Social:
+      type: bins
+      bins:
+        1:
+          prob: 1.0
+          min: 1
+          max: 10
   sex:
     frequency:
       Sex:
@@ -157,6 +151,24 @@ partnership:
             prob: 1.0
             min: 10
             max: 37
+      SexInj:
+        type: bins
+        bins:
+          1:
+            prob: 0.5
+            min: 10
+            max: 50
+          2:
+            prob: 1.0
+            min: 10
+            max: 37
+      Social:
+        type: bins
+        bins:
+          1:
+            prob: 1.0
+            min: 10
+            max: 100
     acquisition:
       MSM:
         insertive: 0.0011
@@ -191,6 +203,7 @@ features:
   static_network: false
   incar: false
   partner_tracing: false
+  die_and_replace: false
 
 model:
   num_pop: 300
@@ -203,3 +216,7 @@ model:
 calibration:
   partnership:
     break_point: 0
+
+hiv:
+  aids:
+    prob: 0.1