def test_feature_conditions(self):
cohort = pd.DataFrame({
'person_id': [1, 2, 3, 4, 5],
'index_date': [
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01')
]
})
features = {
'non_time_bound': {
'age_group': 0,
'age_at_index': 0,
'gender': 0
},
'time_bound': {
'comorbid_condition': [0, 0, 0, 0],
'drug': [0, 0, 0, 0],
'condition': [1, 1, 1, 1],
'procedure': [0, 0, 0, 0],
'measurement': [0, 0, 0, 0],
'measurement_value': [0, 0, 0, 0],
'measurement_range_group': [0, 0, 0, 0],
'visit_count': [0, 0, 0, 0]
},
'time_windows': {
'inf': 5000,
'long': 365,
'med': 180,
'short': 30,
'minimum': 0.05
}
}
expected = pd.DataFrame({
'person_id': [1, 2, 3, 4, 5],
'index_date': [pd.to_datetime('2018-01-01')] * 5,
'44831230_inf': [1, 1, 0, 0, 0],
'2_inf': [1, 1, 0, 0, 0],
'3_inf': [1, 0, 0, 0, 0],
'4_inf': [1, 0, 0, 0, 0],
'44831230_long': [1, 1, 0, 0, 0],
'2_long': [1, 1, 0, 0, 0],
'3_long': [1, 0, 0, 0, 0],
'4_long': [1, 0, 0, 0, 0],
'44831230_med': [1, 1, 0, 0, 0],
'2_med': [1, 1, 0, 0, 0],
'3_med': [1, 0, 0, 0, 0],
'4_med': [1, 0, 0, 0, 0],
'44831230_short': [1, 1, 0, 0, 0],
'2_short': [1, 1, 0, 0, 0],
'3_short': [1, 0, 0, 0, 0],
'4_short': [1, 0, 0, 0, 0],
})
output = FeaturesSelection(cohort, features, omop_tables)()
pd.testing.assert_frame_equal(output, expected)
def test_time_at_risk():
cohort = pd.DataFrame({
'cohort_definition_id': [1, 1, 1, 1, 1, 2, 2, 2],
'person_id': [1, 2, 3, 4, 5, 1, 2, 3],
'index_date': [
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2019-01-01'),
pd.to_datetime('2018-05-01'),
pd.to_datetime('2018-03-01'),
]
})
features = {
'non_time_bound': {
'age_group': 0,
'age_at_index': 1,
'gender': 1
},
'time_bound': {
'comorbid_condition': [0, 0, 0, 0],
'drug': [0, 0, 0, 0],
'condition': [0, 0, 0, 0],
'procedure': [0, 0, 0, 0],
'measurement': [0, 0, 0, 0],
'measurement_value': [0, 0, 0, 0],
'measurement_range_group': [0, 0, 0, 0],
'visit_count': [0, 0, 0, 0]
},
'time_windows': {
'inf': 5000,
'long': 365,
'med': 180,
'short': 30,
'minimum': 0.05
}
}
cohort_at_risk = cohort[cohort['cohort_definition_id'] == 1]
del cohort_at_risk['cohort_definition_id']
cohort_target = cohort[cohort['cohort_definition_id'] == 2]
del cohort_target['cohort_definition_id']
X = FeaturesSelection(cohort_at_risk, features, omop_tables)()
output = time_at_risk(X, cohort_at_risk, cohort_target, time_at_risk=200)
expected = pd.DataFrame({
'age_at_index': [28, 18, 8, 48, 58],
'gender = female': [1, 0, 1, 0, 0],
'target': [0, 1, 1, 0, 0]
})
pd.testing.assert_frame_equal(output, expected)
def test_feature_comorbidities(self):
cohort = pd.DataFrame({
'person_id': [1, 2, 3, 4, 5],
'index_date': [
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01'),
pd.to_datetime('2018-01-01')
]
})
features = {
'non_time_bound': {
'age_group': 0,
'age_at_index': 0,
'gender': 0
},
'time_bound': {
'comorbid_condition': [0, 0, 1, 0],
'drug': [0, 0, 0, 0],
'condition': [0, 0, 0, 0],
'procedure': [0, 0, 0, 0],
'measurement': [0, 0, 0, 0],
'measurement_value': [0, 0, 0, 0],
'measurement_range_group': [0, 0, 0, 0],
'visit_count': [0, 0, 0, 0]
},
'time_windows': {
'inf': 5000,
'long': 365,
'med': 180,
'short': 30,
'minimum': 0.05
}
}
expected = pd.DataFrame({
'person_id': [1, 2, 3, 4, 5],
'index_date': [pd.to_datetime('2018-01-01')] * 5,
'congestive_heart_failure_med': [0, 0, 1, 0, 0],
'valvular_disease_med': [0, 0, 0, 1, 0],
'hypertension,complicated_med': [0, 0, 1, 0, 0]
})
condition_occurrence = pd.DataFrame({
'person_id': [1, 1, 1, 1, 2, 2, 3, 4],
'condition_concept_id': [1, 2, 3, 4, 1, 2, 44831230, 44836801],
'condition_start_datetime': [
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10'),
pd.to_datetime('2017-12-10')
]
})
condition_occurrence = dd.from_pandas(condition_occurrence,
npartitions=1)
condition_occurrence = condition_occurrence.set_index('person_id')
omop_tables = {
'person': person,
'condition_occurrence': condition_occurrence,
'procedure_occurrence': procedure,
'drug_exposure': drug_exposure,
'visit_occurrence': visit_occurrence,
'observation_period': observation_period,
'measurement': measurement
}
output = FeaturesSelection(cohort, features, omop_tables)()
pd.testing.assert_frame_equal(output, expected)