Different vaccination refuser shares for youths

src/create_initial_states/create_vaccination_priority.py

@@ -95,19 +95,13 @@
 import pandas as pd
 
 
-def create_vaccination_rank(vaccination_group, share_refuser, seed):
+def create_vaccination_rank(vaccination_group):
     """Create the order in which individuals get vaccinated, including refusers.
 
     Args:
         vaccination_group (pandas.Series): index is the same as that of states.
             Low values indicate individuals that have a high priority to be
             vaccinated.
-        share_refuser (float): share of individuals (irrespective of their
-            vaccination group) that refuse to be vaccinated.
-
-            .. warning::
-                This share must also be passed to the vaccination model!
-        seed (int)
 
     Returns:
         vaccination_order (pandas.Series): same index as that of
@@ -117,21 +111,52 @@ def create_vaccination_rank(vaccination_group, share_refuser, seed):
             rest.
 
     """
-    np.random.seed(seed)
-    sampled_to_refuse = np.random.choice(
-        a=[True, False],
-        size=len(vaccination_group),
-        p=[share_refuser, 1 - share_refuser],
-    )
-    refuser_value = vaccination_group.max() + 1
-    with_refusers = vaccination_group.where(~sampled_to_refuse, refuser_value)
-    vaccination_order = with_refusers.rank(method="first", pct=True)
+    vaccination_order = vaccination_group.rank(method="first", pct=True)
     min_at_zero = vaccination_order - vaccination_order.min()
     scaled = min_at_zero / min_at_zero.max()
     return scaled
 
 
-def create_vaccination_group(states, seed):
+def add_refuser_value_to_vaccination_group(
+    states, share_refuser_adult, share_refuser_youth, seed
+):
+    """Create the vaccination group giving refusers the highest value.
+
+    Args:
+        states (pandas.DataFrame): states DataFrame, must contain
+            "vaccination_priority_group" and "age".
+        share_refuser_adult (float): share of adults (irrespective of their
+            vaccination group) that refuse to be vaccinated.
+        share_refuser_youth (float): share of youths that refuse to be vaccinated.
+        seed (int): seed
+
+    """
+    np.random.seed(seed)
+    refuser_value = states["vaccination_priority_group"].max() + 1
+    is_adult = states["age"] >= 16
+
+    adults_sampled_to_refuse = np.random.choice(
+        a=[True, False],
+        size=len(states),
+        p=[share_refuser_adult, 1 - share_refuser_adult],
+    )
+    youths_sampled_to_refuse = np.random.choice(
+        a=[True, False],
+        size=len(states),
+        p=[share_refuser_youth, 1 - share_refuser_youth],
+    )
+
+    with_adult_refusers = states["vaccination_priority_group"].where(
+        ~adults_sampled_to_refuse | ~is_adult, refuser_value
+    )
+    with_all_refusers = with_adult_refusers.where(
+        ~youths_sampled_to_refuse | is_adult, refuser_value
+    )
+    with_all_refusers.name = None
+    return with_all_refusers
+
+
+def create_vaccination_group_without_refusers(states, seed):
     """Put individuals into vaccination priority groups based on age and work.
 
     Args:
@@ -147,7 +172,7 @@ def create_vaccination_group(states, seed):
 
     """
     np.random.seed(seed)
-    is_adult = states["age"] >= 18
+    is_adult = states["age"] >= 16
 
     vaccination_group = pd.Series(np.nan, index=states.index)
 
@@ -222,7 +247,7 @@ def _get_third_priority(states, vaccination_group):
         "(60 <= age < 70) | educ_worker | work_contact_priority > 0.88"
     )
     third_priority_non_random = states.eval(third_priority_non_random_str)
-    n_third_priority_random = 0.03 * (states["age"] >= 18).sum()
+    n_third_priority_random = 0.03 * (states["age"] >= 16).sum()
     third_priority_sampled = _sample_from_subgroups(
         n_to_sample=n_third_priority_random,
         states=states,
@@ -274,7 +299,7 @@ def _sample_from_subgroups(
     n_young = int((1 - share_to_sample_above_age_cutoff) * n_to_sample)
     n_old = int(share_to_sample_above_age_cutoff * n_to_sample)
 
-    pool = states[(states["age"] >= 18) & vaccination_groups_so_far.isnull()]
+    pool = states[(states["age"] >= 16) & vaccination_groups_so_far.isnull()]
     young_pool = pool[pool["age"] < age_cutoff].index
     old_pool = pool[pool["age"] >= age_cutoff].index
 
@@ -298,14 +323,14 @@ def _check_vaccination_group(vaccination_group, states):
     youth_groups = vaccination_group[(12 <= states["age"]) & (states["age"] <= 16)]
     assert (youth_groups == 5).all()
 
-    adult_groups = vaccination_group[states["age"] >= 18]
+    adult_groups = vaccination_group[states["age"] >= 16]
     share_group_1 = (adult_groups == 1).mean()
     share_group_2 = (adult_groups == 2).mean()
     share_group_3 = (adult_groups == 3).mean()
     share_group_4 = (adult_groups == 4).mean()
     assert 0.08 < share_group_1 < 0.10, share_group_1
-    assert 0.15 < share_group_2 < 0.17, share_group_2
-    assert 0.18 < share_group_3 < 0.19, share_group_3
+    assert 0.15 < share_group_2 < 0.175, share_group_2
+    assert 0.175 < share_group_3 < 0.19, share_group_3
     assert 0.55 < share_group_4 < 0.6, share_group_4
     res_shares = adult_groups.value_counts(normalize=True)
     target_shares = pd.Series([0.09, 0.15, 0.19, 0.57], index=[1, 2, 3, 4])

src/create_initial_states/task_build_full_params.py

@@ -92,7 +92,11 @@ def task_create_full_params(depends_on, produces):
     # 12% are undecided. 8% are opposed to being vaccinated.
     # We assume that 15% will refuse to be vaccinated.
     # source: https://bit.ly/3c9mTgX (publication date: 2021-03-02)
-    params.loc[("vaccinations", "share_refuser", "share_refuser"), "value"] = 0.15
+    params.loc[("vaccinations", "share_refuser", "adult"), "value"] = 0.15
+    # for 12 to 15 year olds, there are more refusers: 44 to 66% are willing to
+    # vaccinate their child (https://bit.ly/3BdHHgJ). Thus, for them the share
+    # of refusers is 34 to 56%
+    params.loc[("vaccinations", "share_refuser", "youth"), "value"] = 0.45
 
     # source: https://bit.ly/3gHlcKd (section 3.5, 2021-03-09, accessed 2021-04-28)
     # 82% say they would get a PCR test after a positive rapid test

src/create_initial_states/task_create_background_characteristics.py

@@ -14,7 +14,10 @@
     add_contact_model_group_ids,
 )
 from src.create_initial_states.create_vaccination_priority import (
-    create_vaccination_group,
+    add_refuser_value_to_vaccination_group,
+)
+from src.create_initial_states.create_vaccination_priority import (
+    create_vaccination_group_without_refusers,
 )
 from src.create_initial_states.create_vaccination_priority import (
     create_vaccination_rank,
@@ -81,8 +84,11 @@ def task_create_initial_states_microcensus(depends_on, n_hhs, produces):
     other_daily_dist = pd.read_pickle(depends_on["other_daily_dist"])
     other_weekly_dist = pd.read_pickle(depends_on["other_weekly_dist"])
     params = pd.read_pickle(depends_on["params"])
-    no_vaccination_share = params.loc[
-        ("vaccinations", "share_refuser", "share_refuser"), "value"
+    no_vaccination_share_adult = params.loc[
+        ("vaccinations", "share_refuser", "adult"), "value"
+    ]
+    no_vaccination_share_youth = params.loc[
+        ("vaccinations", "share_refuser", "youth"), "value"
     ]
 
     df = _build_initial_states(
@@ -94,7 +100,8 @@ def task_create_initial_states_microcensus(depends_on, n_hhs, produces):
         other_weekly_dist=other_weekly_dist,
         n_households=n_hhs,
         seed=3933,
-        no_vaccination_share=no_vaccination_share,
+        no_vaccination_share_adult=no_vaccination_share_adult,
+        no_vaccination_share_youth=no_vaccination_share_youth,
     )
     df.to_parquet(produces)
 
@@ -108,7 +115,8 @@ def _build_initial_states(
     other_weekly_dist,
     n_households,
     seed,
-    no_vaccination_share,
+    no_vaccination_share_adult,
+    no_vaccination_share_youth,
 ):
     mc = _prepare_microcensus(mc)
 
@@ -146,9 +154,17 @@ def _build_initial_states(
     df["adult_in_hh_at_home"] = adult_at_home.groupby(df["hh_id"]).transform(np.any)
     df["educ_contact_priority"] = _create_educ_contact_priority(df)
 
-    df["vaccination_group"] = create_vaccination_group(states=df, seed=484)
+    df["vaccination_priority_group"] = create_vaccination_group_without_refusers(
+        states=df, seed=484
+    )
+    df["vaccination_group_with_refuser_group"] = add_refuser_value_to_vaccination_group(
+        states=df,
+        share_refuser_adult=no_vaccination_share_adult,
+        share_refuser_youth=no_vaccination_share_youth,
+        seed=909,
+    )
     df["vaccination_rank"] = create_vaccination_rank(
-        df["vaccination_group"], share_refuser=no_vaccination_share, seed=909
+        vaccination_group=df["vaccination_group_with_refuser_group"]
     )
 
     # This is uncorrelated with the work contact priority.
@@ -339,7 +355,7 @@ def _only_keep_relevant_columns(df):
         "hh_model_group_id",
         "adult_in_hh_at_home",
         "educ_contact_priority",
-        "vaccination_group",
+        "vaccination_group_with_refuser_group",
         "vaccination_rank",
         "rapid_test_compliance",
         "quarantine_compliance",
@@ -370,6 +386,7 @@ def _only_keep_relevant_columns(df):
         "index",
         "stays_home_when_schools_close",  # not used at the moment
         "work_type",
+        "vaccination_priority_group",
     ]
 
     assert set(keep + to_drop) == set(df.columns)

src/policies/find_people_to_vaccinate.py

@@ -1,5 +1,6 @@
 import warnings
 
+import numpy as np
 import pandas as pd
 from sid import get_date
 
@@ -8,37 +9,46 @@ def find_people_to_vaccinate(
     receives_vaccine,  # noqa: U100
     states,
     params,
-    seed,  # noqa: U100
+    seed,
     vaccination_shares,
     init_start,
 ):
     """Find people that have to be vaccinated on a given day.
 
-    On the init_start date all individuals that should have been vaccinated until
-    that day get vaccinated. Since vaccinations take effect within three weeks
-    and the burn in period is four weeks this does not lead to jumps in the
-    simulation period.
+    On the init_start date all individuals that should have been vaccinated
+    until that day get vaccinated. Since vaccinations take effect within three
+    weeks and the burn in period is four weeks this does not lead to jumps in
+    the simulation period.
+
+    On June 7 the vaccination priorization was lifted in Germany. On May 28 the
+    EMA approved the first vaccine for youths from 12 to 15. However, until July
+    15 less than 5% of 12 to 17 year olds were vaccinated and the STIKO only
+    recommended the CoViD vaccine for healthy 12 to 17 year olds in August.
+
+    We model this by vaccinating individuals 12 and older that are not refusers
+    randomly starting on July 15.
+
+    Sources:
+        - https://bit.ly/2WpIegM
+        - https://bit.ly/3zlzIhe
 
     Args:
         states (pandas.DataFrame): States DataFrame that must contain the
-            column vaccination_rank. This column is a float with values
-            between zero and one. Low values mean that people get
-            vaccinated early.
-        params (pandas.DataFrame): not used.
-        seed (int): not used.
-        vaccination_shares (pandas.Series): Series with a date index. For each
-            day the value indicates the share of people who get vaccinated on
-            that day.
-        no_vaccination_share (float): Share of people who refuse to get
-            vaccinated.
-        init_start (pd.Timestamp): start date of the burn in period. On the
-            init_start all vaccinations that have been done until then are
-            handed out on that day.
+            column vaccination_rank. This column is a float with values between zero
+            and one. Low values mean that people get vaccinated early.
+        params (pandas.DataFrame): params DataFrame that contains a row labeled
+            ("vaccinations", "share_refuser", "adult").
+        seed (int): used after 2021-06-07.
+        vaccination_shares (pandas.Series): Series with a date index. For each day the
+            value indicates the share of people who get vaccinated on that day.
+        init_start (pd.Timestamp): start date of the burn in period. On the init_start
+            all vaccinations that have been done until then are handed out on that day.
 
     """
+    np.random.seed(seed)
     date = get_date(states)
-    no_vaccination_share = params.loc[
-        ("vaccinations", "share_refuser", "share_refuser"), "value"
+    no_vaccination_share_adult = params.loc[
+        ("vaccinations", "share_refuser", "adult"), "value"
     ]
 
     if not (vaccination_shares < 0.05).all():
@@ -47,18 +57,33 @@ def find_people_to_vaccinate(
             "If this was intended, simply ignore the warning.",
         )
 
-    cutoffs = vaccination_shares.sort_index().cumsum()
-    # set all cutoffs before the init_start to 0.
-    # that way on the init_start date everyone who should have been vaccinated
-    # until that day gets vaccinated.
-    cutoffs[: init_start - pd.Timedelta(days=1)] = 0
+    if date < pd.Timestamp("2021-07-15"):  # priorisation. only >= 16 get vaccinated.
+        cutoffs = vaccination_shares.sort_index().cumsum()
+        # set all cutoffs before the init_start to 0.
+        # that way on the init_start date everyone who should have been vaccinated
+        # until that day gets vaccinated.
+        cutoffs[: init_start - pd.Timedelta(days=1)] = 0
+
+        lower_candidate = cutoffs[date - pd.Timedelta(days=1)]
+        upper_candidate = cutoffs[date]
+        lower = min(lower_candidate, 1 - no_vaccination_share_adult)
+        upper = min(upper_candidate, 1 - no_vaccination_share_adult)
 
-    lower_candidate = cutoffs[date - pd.Timedelta(days=1)]
-    upper_candidate = cutoffs[date]
-    lower = min(lower_candidate, 1 - no_vaccination_share)
-    upper = min(upper_candidate, 1 - no_vaccination_share)
+        to_vaccinate = (lower <= states["vaccination_rank"]) & (
+            states["vaccination_rank"] < upper
+        )
+    else:  # no priorisation. >= 12 get vaccinated
+        unvaccinated = ~states["ever_vaccinated"]
+        eligible = states["age"] >= 12
+
+        # the highest value denotes refusers
+        willing = (
+            states["vaccination_group_with_refuser_group"]
+            < states["vaccination_group_with_refuser_group"].max()
+        )
+        pool = states[unvaccinated & eligible & willing].index
+        n_to_draw = int(vaccination_shares[date] * len(states))
+        sampled = np.random.choice(pool, n_to_draw, replace=False)
+        to_vaccinate = pd.Series(states.index.isin(sampled), index=states.index)
 
-    to_vaccinate = (lower <= states["vaccination_rank"]) & (
-        states["vaccination_rank"] < upper
-    )
     return to_vaccinate