def getfeaturesFromEventsTables(config):
conn = SQLConnection(config)
# getting features from labevents
query_labmsmts = '''
select * from LABMEASURMENTS;
'''
df_labmsmts = conn.executeQuery(query_labmsmts)
#observations from labevents at discharge time
query_labslastmsmts = '''
select * from labslastmsmts;
'''
df_labslastmsmts = conn.executeQuery(query_labslastmsmts)
#all extacted features from labevents
df_labs_features = get_features_from_labevents(df_labmsmts, df_labslastmsmts)
# getting features from chartevents
query_chartsmsmts = '''
select * from CHARTSMEASURMENTS;
'''
df_chartsmsmts = conn.executeQuery(query_chartsmsmts)
# getting features from chartevents
query_chartslastmsmts = '''
select * from CHARTSLASTMSMTS;
'''
df_chartslastmsmts = conn.executeQuery(query_chartslastmsmts)
df_chart_features = get_features_from_chartevents(df_chartsmsmts, df_chartslastmsmts)
df_events_tables = pd.merge(df_labs_features, df_chart_features, how='left', on='icustay_id')
return df_events_tables
def getfeaturesFromSeverityScoreConcepts(config):
"""
The scripts generate features from MIMIC-III concepts tables for severity scores
:param config:
:return:
"""
print(
"\n Extracting features from MIMIC-III concepts tables for severity scores"
)
conn = SQLConnection(config)
sapsii = conn.executeQuery('''select * from sapsii;''')
sapsii = sapsii[sapsii.icustay_id.isnull() == False]
sapsii['icustay_id'] = sapsii['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
sapsii.drop(cols_to_remove, axis=1, inplace=True)
sofa = conn.executeQuery('''select * from sofa;''')
sofa = sofa[sofa.icustay_id.isnull() == False]
sofa['icustay_id'] = sofa['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
sofa.drop(cols_to_remove, axis=1, inplace=True)
sirs = conn.executeQuery('''select * from sirs;''')
sirs = sirs[sirs.icustay_id.isnull() == False]
sirs['icustay_id'] = sirs['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
sirs.drop(cols_to_remove, axis=1, inplace=True)
lods = conn.executeQuery('''select * from lods;''')
lods = lods[lods.icustay_id.isnull() == False]
lods['icustay_id'] = lods['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
lods.drop(cols_to_remove, axis=1, inplace=True)
apsiii = conn.executeQuery('''select * from apsiii;''')
apsiii = apsiii[apsiii.icustay_id.isnull() == False]
apsiii['icustay_id'] = apsiii['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
apsiii.drop(cols_to_remove, axis=1, inplace=True)
oasis = conn.executeQuery('''select * from oasis;''')
oasis = oasis[oasis.icustay_id.isnull() == False]
oasis['icustay_id'] = oasis['icustay_id'].astype('int')
cols_to_remove = ['subject_id', 'hadm_id']
oasis.drop(cols_to_remove, axis=1, inplace=True)
sapsii_sofa = pd.merge(sapsii, sofa, on='icustay_id')
# removing repeated column
sapsii_sofa.drop(['temp_score'], axis=1, inplace=True)
sapsii_sofa_sirs = pd.merge(sapsii_sofa, sirs, on='icustay_id')
# removing repeated column
sapsii_sofa_sirs.drop(['cardiovascular'], axis=1, inplace=True)
sapsii_sofa_sirs_lods = pd.merge(sapsii_sofa_sirs, lods, on='icustay_id')
sapsii_sofa_sirs_lods_apsiii = pd.merge(sapsii_sofa_sirs_lods,
apsiii,
on='icustay_id')
df_severity_scores = pd.merge(sapsii_sofa_sirs_lods_apsiii,
oasis,
on='icustay_id')
# filling score columns with 0 since in mimic db mean scores is taken as 0
score_columns = df_severity_scores.filter(regex='score').columns
df_severity_scores[score_columns] = df_severity_scores[
score_columns].fillna(0)
# removing columns with > 50% NA value
df_severity_scores = clean_nan_columns(df_severity_scores, thres=60)
# cleaning up non-numeric columns from severity scores
df_severity_scores.drop(['icustay_age_group', 'preiculos'],
axis=1,
inplace=True)
return df_severity_scores