-
Notifications
You must be signed in to change notification settings - Fork 0
/
T2&T3_dataset.py
195 lines (111 loc) · 6.05 KB
/
T2&T3_dataset.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
#!/usr/bin/env python
# coding: utf-8
# In[1]:
import pandas as pd
from utils import convert_icd_group, clean_files
from itertools import chain
import numpy as np
# In[40]:
ICD = pd.read_csv('DIAGNOSES_ICD.csv')
ICD.columns = map(str.lower, ICD.columns)
ICD = ICD[ICD.seq_num == 1.0] #kept only first diagnosis
ICD = ICD.drop(["seq_num","subject_id", "row_id"],axis=1)
#if you want to discover the third diagnosis level (3 digit category), uncomment these lines
# ICD["Length"]= ICD["icd9_code"].str.len()
# ICD.loc[(ICD.Length == 4), 'icd9_code']=ICD['icd9_code'].astype(str).str[:-1]
# ICD.loc[(ICD.Length == 5), 'icd9_code']=ICD['icd9_code'].astype(str).str[:-2]
# ICD =ICD.drop(["Length"],axis=1)
#uncomment this line if you want to discover the second diagnosis level (block)
#ICD['icd9_code'] = ICD['icd9_code'].apply(convert_icd_group)
#run the function clean_files() to obtain clean csv file
clean_files()
diag_df = pd.read_csv("clean_diagnosis.csv")
diag_df = diag_df.drop(["Unnamed: 0", "icd9_code"], axis=1)
diag_df = diag_df.merge(ICD, on='hadm_id')
# we need to drop columns with disease happening only once to establish the relation
counts_col1 = diag_df.groupby("icd9_code")["icd9_code"].transform(len)
mask = (counts_col1 > 1)
diag_df =diag_df[mask]
#uncomment this command to see the value count for each ICD code
#diag_df['icd9_code'].value_counts() #925
# when we dropped icd_codes that appeared only once, the number of admission for that patient drops by one
# recaculate the admission time for each patient
# keep the patient with at least two admissions
diag_df['admit_times'] = diag_df.groupby(['subject_id'])['subject_id'].transform('size')
diag_df["admit_times"].astype(int)
diag_df = diag_df[diag_df['admit_times'] >= 2]
print('patient with more than one adm', diag_df['subject_id'].nunique())
#For the 2nd setting, we include a pairs of STAY 1 followed by stay 2 and STAY 2 followed by STAY 3 and STAY 1 followed by STAY 3 (as stays can happen in between)
#For the 3rd setting, we include pairs of STAY 1 followed by stay 2 and STAY 2 followed by STAY 3(no stay can happen in between)
#Here we are forming pairs of STAY 1 followed by STAY 2 (S1_S2)
df_data = diag_df[["subject_id", "hadm_id", "icd9_code", "admittime", "dischtime"]]
df_data['next_stay'] = df_data.groupby('subject_id')['hadm_id'].shift(-1)
df_data['next_stay'] = df_data['next_stay'].astype("Int32")
df_data = df_data[[ "hadm_id", "next_stay","subject_id", "icd9_code"]]
df_data = df_data[df_data['next_stay'].notna()] #remove rows where next stay is NA
df_data.rename(columns = {'hadm_id':'STAY A'}, inplace = True)
df_data.rename(columns = {'next_stay':'STAY B'}, inplace = True)
# dropping again will influence the number of disease happening, therefore we drop those happening only once
counts_col1 = df_data.groupby("icd9_code")["icd9_code"].transform(len)
mask = (counts_col1 > 1)
df_data =df_data[mask]
#df_data['icd9_code'].value_counts() #727 diagnosis shared btwn at least two patients
df_data.to_csv("S1_S2.csv")
# STAY 2 followed by STAY 3 (S2_S3)
df_data1 = diag_df[["subject_id", "hadm_id", "icd9_code", "admittime", "dischtime"]]
df_data1['next_stay'] = df_data1.groupby('subject_id')['hadm_id'].shift(-1)
df_data1['next_stay2'] = df_data1.groupby('subject_id')['hadm_id'].shift(-2)
df_data1['next_stay'] = df_data1['next_stay'].astype("Int32")
df_data1['next_stay2'] = df_data1['next_stay2'].astype("Int32")
df_data1 = df_data1[df_data1['next_stay'].notna()]
df_data1 = df_data1[df_data1['next_stay2'].notna()]
#print('patient', df_data1['subject_id'].nunique())
df_data1 = df_data1[[ "next_stay", "next_stay2","subject_id"]]
df_data1.rename(columns = {'next_stay':'hadm_id'}, inplace = True)
df_data1 = df_data1.merge(ICD, on='hadm_id')
#df_data1['icd9_code'].value_counts() #608 diagnosis shared btwn at least two patients
# we need to drop columns with disease happening only once to establish the relation
counts_col1 = df_data1.groupby("icd9_code")["icd9_code"].transform(len)
mask = (counts_col1 > 1)
df_data1 =df_data1[mask]
df_data1.rename(columns = {'hadm_id':'STAY A'}, inplace = True)
df_data1.rename(columns = {'next_stay2':'STAY B'}, inplace = True)
df_data1.to_csv("S2_S3.csv")
# STAY 1 followed by STAY 3 (S1_S3)
df_data2 = diag_df[["subject_id", "hadm_id", "icd9_code", "admittime", "dischtime"]]
df_data2['next_stay2'] = df_data2.groupby('subject_id')['hadm_id'].shift(-2)
df_data2['next_stay2'] = df_data2['next_stay2'].astype("Int32")
df_data2 = df_data2[["hadm_id", "next_stay2", "subject_id", "icd9_code"]]
df_data2.rename(columns = {'hadm_id':'STAY A'}, inplace = True)
df_data2.rename(columns = {'next_stay2':'STAY B'}, inplace = True)
df_data2 = df_data2[df_data2['STAY B'].notna()]
#df_data2['icd9_code'].value_counts() #628 diagnosis shared btwn at least two patients
# we need to drop columns with disease happening only once to establish the relation
counts_col1 = df_data2.groupby("icd9_code")["icd9_code"].transform(len)
mask = (counts_col1 > 1)
df_data2 =df_data2[mask]
df_data2.to_csv("datasets/S1_S3.csv")
# To form the dataset for the third setting - T3
#merge 2 csv files
df1 = pd.read_csv("S1_S2.csv")
df3 = pd.read_csv("S2_S3.csv")
final_df = pd.concat([df1,df3])
final_df = final_df.drop(["Unnamed: 0"], axis =1)
print(len(final_df))
diag_once = final_df.drop_duplicates(subset=["STAY A", "STAY B", "subject_id", "icd9_code"], keep='first')
diag_once.drop(["Unnamed: 0"], axis=1)
#change the acronym ICD based on what level of the diagnosis constraint you are using
diag_once.to_csv("T3_ID+DSEQ_ICD.csv")
print(len(diag_once))
# To form the dataset for the second setting - T2
#merge all 3 csv files
df1 = pd.read_csv("S1_S2.csv")
df2 = pd.read_csv("S1_S3.csv")
df3 = pd.read_csv("S2_S3.csv")
final_df = pd.concat([df1,df2,df3])
final_df = final_df.drop(["Unnamed: 0"], axis =1)
print(len(final_df))
diag_once = final_df.drop_duplicates(subset=["STAY A", "STAY B", "subject_id", "icd9_code"], keep='first')
print(len(diag_once))
#change the acronym ICD based on what level of the diagnosis constraint you are using
diag_once.to_csv("T2_ID+SEQ_ICD.csv")