def get_treatment_efficacy(
demography: pd.DataFrame,
treatment_type: str) -> Tuple[pd.DataFrame, pd.DataFrame]:
baseline_efficacy = {
data_keys.WASTING.CAT1:
get_random_variable_draws(
ARTIFACT_COLUMNS, *data_values.WASTING.BASELINE_SAM_TX_EFFICACY),
data_keys.WASTING.CAT2:
get_random_variable_draws(
ARTIFACT_COLUMNS, *data_values.WASTING.BASELINE_MAM_TX_EFFICACY)
}
alternative_efficacy = {
data_keys.WASTING.CAT1:
data_values.WASTING.SAM_TX_ALTERNATIVE_EFFICACY,
data_keys.WASTING.CAT2: data_values.WASTING.MAM_TX_ALTERNATIVE_EFFICACY
}
idx_as_frame = demography.merge(pd.DataFrame(
{'parameter': [f'cat{i}' for i in range(1, 4)]}),
how='cross')
index = idx_as_frame.set_index(list(idx_as_frame.columns)).index
efficacy = pd.DataFrame({f'draw_{i}': 1.0
for i in range(0, 1000)},
index=index)
efficacy[index.get_level_values('parameter') == 'cat1'] *= 0.0
efficacy[index.get_level_values('parameter') ==
'cat2'] *= baseline_efficacy[treatment_type]
efficacy[index.get_level_values('parameter') ==
'cat3'] *= alternative_efficacy[treatment_type]
tmrel_efficacy = (
efficacy[efficacy.index.get_level_values('parameter') == data_keys.
MAM_TREATMENT.TMREL_CATEGORY].droplevel('parameter'))
return efficacy, tmrel_efficacy
def load_duration(key: str, location: str) -> pd.DataFrame:
try:
distribution = {
data_keys.DIARRHEA.DURATION: data_values.DIARRHEA_DURATION,
data_keys.LRI.DURATION: data_values.LRI_DURATION,
}[key]
except KeyError:
raise ValueError(f'Unrecognized key {key}')
demography = get_data(data_keys.POPULATION.DEMOGRAPHY, location)
duration_draws = (
get_random_variable_draws(metadata.ARTIFACT_COLUMNS, *distribution)
/ metadata.YEAR_DURATION
)
enn_duration = pd.DataFrame(
data_values.EARLY_NEONATAL_CAUSE_DURATION / metadata.YEAR_DURATION,
columns=metadata.ARTIFACT_COLUMNS,
index=demography.query('age_start == 0.0').index
)
all_other_duration = pd.DataFrame(
[duration_draws], index=demography.query('age_start != 0.0').index
)
duration = pd.concat([enn_duration, all_other_duration]).sort_index()
return duration
def load_therapeutic_zinc_rr(key: str, location: str) -> pd.Series:
if key != data_keys.THERAPEUTIC_ZINC.RELATIVE_RISK:
raise ValueError(f'Unrecognized key {key}')
idx = get_data(data_keys.POPULATION.DEMOGRAPHY, location).index
diarrhea_duration_shift_years = pd.DataFrame(
[get_random_variable_draws(
metadata.ARTIFACT_COLUMNS, *data_values.THERAPEUTIC_ZINC.DIARRHEA_DURATION_SHIFT_HOURS
) / (metadata.DAY_DURATION * metadata.YEAR_DURATION)],
index=idx
)
diarrhea_duration_years = pd.DataFrame(
[get_random_variable_draws(
metadata.ARTIFACT_COLUMNS, *data_values.DIARRHEA_DURATION
) / metadata.YEAR_DURATION],
index=idx
)
baseline_coverage = (
get_data(data_keys.THERAPEUTIC_ZINC.EXPOSURE, location)
.query(f'parameter == "cat2"')
.droplevel('parameter')
)
duration_uncovered = diarrhea_duration_years - (
diarrhea_duration_shift_years * baseline_coverage
)
duration_covered = duration_uncovered + diarrhea_duration_shift_years
remission_rate_uncovered = 1 / duration_uncovered
remission_rate_covered = 1 / duration_covered
exposed_rr = remission_rate_covered / remission_rate_uncovered
return convert_to_dichotomous_rr(
exposed_rr,
data_keys.THERAPEUTIC_ZINC.AFFECTED_ENTITY,
data_keys.THERAPEUTIC_ZINC.AFFECTED_MEASURE
)
def load_preventative_zinc_rr(key: str, location: str) -> pd.DataFrame:
if key != data_keys.PREVENTATIVE_ZINC.RELATIVE_RISK:
raise ValueError(f'Unrecognized key {key}')
index = get_data(data_keys.POPULATION.DEMOGRAPHY, location).index
distribution = data_values.PREVENTATIVE_ZINC.PREVENTATIVE_TX_EFFICACY
exposed_rr = pd.DataFrame(
[get_random_variable_draws(metadata.ARTIFACT_COLUMNS, *distribution)], index=index
)
return convert_to_dichotomous_rr(
exposed_rr,
data_keys.PREVENTATIVE_ZINC.AFFECTED_ENTITY,
data_keys.PREVENTATIVE_ZINC.AFFECTED_MEASURE
)
def load_dichotomous_excess_shift(
location: str, distribution_data: Tuple, is_risk: bool
) -> pd.DataFrame:
index = get_data(data_keys.POPULATION.DEMOGRAPHY, location).index
shift = get_random_variable_draws(metadata.ARTIFACT_COLUMNS, *distribution_data)
exposed = pd.DataFrame([shift], index=index)
exposed['parameter'] = 'cat1' if is_risk else 'cat2'
unexposed = pd.DataFrame([pd.Series(0.0, index=metadata.ARTIFACT_COLUMNS)], index=index)
unexposed['parameter'] = 'cat2' if is_risk else 'cat1'
excess_shift = pd.concat([exposed, unexposed])
excess_shift['affected_entity'] = data_keys.LBWSG.BIRTH_WEIGHT_EXPOSURE.name
excess_shift['affected_measure'] = data_keys.LBWSG.BIRTH_WEIGHT_EXPOSURE.measure
excess_shift = (
excess_shift
.set_index(['affected_entity', 'affected_measure', 'parameter'], append=True)
.sort_index()
)
return excess_shift
def load_dichotomous_exposure(
location: str,
distribution_data: Union[float, Tuple],
is_risk: bool,
coverage: float = 1.0,
has_under_6mo_exposure: bool = True,
) -> pd.DataFrame:
index = get_data(data_keys.POPULATION.DEMOGRAPHY, location).index
if type(distribution_data) == float:
base_exposure = pd.Series(distribution_data, index=metadata.ARTIFACT_COLUMNS)
else:
base_exposure = get_random_variable_draws(metadata.ARTIFACT_COLUMNS, *distribution_data)
exposed = pd.DataFrame([base_exposure * coverage], index=index)
if not has_under_6mo_exposure:
exposed.loc[exposed.index.get_level_values('age_end') <= 0.5] = 0.0
unexposed = 1 - exposed
exposed['parameter'] = 'cat1' if is_risk else 'cat2'
unexposed['parameter'] = 'cat2' if is_risk else 'cat1'
exposure = pd.concat([exposed, unexposed]).set_index('parameter', append=True).sort_index()
return exposure
def load_wasting_treatment_exposure(key: str, location: str) -> pd.DataFrame:
if key == data_keys.SAM_TREATMENT.EXPOSURE:
coverage_distribution = data_values.WASTING.BASELINE_SAM_TX_COVERAGE
elif key == data_keys.MAM_TREATMENT.EXPOSURE:
coverage_distribution = data_values.WASTING.BASELINE_MAM_TX_COVERAGE
else:
raise ValueError(f'Unrecognized key {key}')
treatment_coverage = get_random_variable_draws(
metadata.ARTIFACT_COLUMNS, *coverage_distribution
)
idx = get_data(data_keys.POPULATION.DEMOGRAPHY, location).index
cat3 = pd.DataFrame({f'draw_{i}': 0.0 for i in range(0, 1000)}, index=idx)
cat2 = pd.DataFrame({f'draw_{i}': 1.0 for i in range(0, 1000)}, index=idx) * treatment_coverage
cat1 = 1 - cat2
cat1['parameter'] = 'cat1'
cat2['parameter'] = 'cat2'
cat3['parameter'] = 'cat3'
exposure = pd.concat([cat1, cat2, cat3]).set_index('parameter', append=True).sort_index()
return exposure
def load_mam_treatment_rr(key: str, location: str) -> pd.DataFrame:
# tmrel is defined as baseline treatment (cat_2)
if key != data_keys.MAM_TREATMENT.RELATIVE_RISK:
raise ValueError(f'Unrecognized key {key}')
demography = get_data(data_keys.POPULATION.DEMOGRAPHY, location).reset_index()
mam_tx_efficacy, mam_tx_efficacy_tmrel = utilities.get_treatment_efficacy(demography, data_keys.WASTING.CAT2)
index = mam_tx_efficacy.index
mam_ux_duration = data_values.WASTING.MAM_UX_RECOVERY_TIME_OVER_6MO
mam_tx_duration = pd.Series(index=index)
mam_tx_duration[index.get_level_values('age_start') < 0.5] = data_values.WASTING.MAM_TX_RECOVERY_TIME_UNDER_6MO
mam_tx_duration[0.5 <= index.get_level_values('age_start')] = (
get_random_variable_draws(
mam_tx_duration[0.5 <= index.get_level_values('age_start')].index,
*data_values.WASTING.MAM_TX_RECOVERY_TIME_OVER_6MO)
)
mam_tx_duration = (
pd.DataFrame({f'draw_{i}': 1 for i in range(0, 1000)}, index=index)
.multiply(mam_tx_duration, axis='index')
)
# rr_r3 = r3 / r3_tmrel
# = (mam_tx_efficacy / mam_tx_duration) + (1 - mam_tx_efficacy / mam_ux_duration)
# / (mam_tx_efficacy_tmrel / mam_tx_duration) + (1 - mam_tx_efficacy_tmrel / mam_ux_duration)
# = (mam_tx_efficacy * mam_ux_duration + (1 - mam_tx_efficacy) * mam_tx_duration)
# / (mam_tx_efficacy_tmrel * mam_ux_duration + (1 - mam_tx_efficacy_tmrel) * mam_tx_duration)
rr = ((mam_tx_efficacy * mam_ux_duration + (1 - mam_tx_efficacy) * mam_tx_duration)
/ (mam_tx_efficacy_tmrel * mam_ux_duration + (1 - mam_tx_efficacy_tmrel) * mam_tx_duration))
rr['affected_entity'] = 'moderate_acute_malnutrition_to_mild_child_wasting'
rr['affected_measure'] = 'transition_rate'
rr = rr.set_index(['affected_entity', 'affected_measure'], append=True)
rr.index = rr.index.reorder_levels([col for col in rr.index.names if col != 'parameter'] + ['parameter'])
rr.sort_index()
return rr