forked from hobbes8080/risc-19-icu
-
Notifications
You must be signed in to change notification settings - Fork 0
/
R_MANAGE_TRANSFORM.R
276 lines (253 loc) · 12.4 KB
/
R_MANAGE_TRANSFORM.R
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
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
## Data transformation
## 2020-03 M. Hilty, P. Wendel Garcia
## -----------------------
## initial data processing
## date/time of snapshot
snapshot_date
## same, without spaces etc for filename stamps
snapshot_date_str <- gsub("[[:blank:]]", "", snapshot_date)
snapshot_date_str <- gsub("-", "", snapshot_date_str)
snapshot_date_str <- gsub(":", "", snapshot_date_str)
## create factors
data$centername_redcap <- factor(data$redcap_data_access_group)
data$form <- factor(data$redcap_event_name)
data$patient_ID <- factor(data$record_id)
s <- strsplit(as.character(data$patient_ID), "-")
data$center_ID <- factor(sapply(s, function(m) m[1] ))
## correct variable types
data$lactate <- as.numeric(data$lactate)
data$pct <- as.numeric(data$pct)
## all forms in redcap have all columns
## prepare admission and discharge forms. later merge them to replace the NA columns in the respective ones.
patients_adm <- data[data$form=="icu_admission__day_arm_1", ]
patients_dis <- data[data$form=="icu_discharge__out_arm_1", ]
## prepare long icu form
patients_icu <- data[data$form=="icu_admission__day_arm_1", ]
patients_icu$time <- 0
for(i in c(1, 2, 3, 5, 7)){
df.t <- data[data$form==paste("day_", i, "_arm_1", sep=""), ]
df.t$time <- i
patients_icu <- rbind(patients_icu, df.t)
}
## remove adm and dis columns from icu DF
df.t <- cbind(patients_icu$record_id, patients_icu[, c(col_icu:(col_dis-1))])
colnames(df.t)[1] <- "record_id"
df.t <- cbind(df.t, patients_icu$time)
colnames(df.t)[length(colnames(df.t))] <- "time"
## add this here since it is needed below, but remove thereafter
patients_icu <- cbind(df.t, patients_icu$patient_ID)
colnames(patients_icu)[length(colnames(patients_icu))] <- "patient_ID"
## ------------------------
## ICU calcualtions
## ------------------------
## convert Fio2 from liter o2 to fio2
for(i in 1:nrow(patients_icu)){
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 1)
patients_icu$fio2[i] <- 24
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 2)
patients_icu$fio2[i] <- 28
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 3)
patients_icu$fio2[i] <- 32
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 4)
patients_icu$fio2[i] <- 36
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 5)
patients_icu$fio2[i] <- 40
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 6)
patients_icu$fio2[i] <- 44
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 7)
patients_icu$fio2[i] <- 70
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 8)
patients_icu$fio2[i] <- 80
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 9)
patients_icu$fio2[i] <- 90
if(!is.na(patients_icu$o2_liter[i]) & patients_icu$o2_liter[i] == 10)
patients_icu$fio2[i] <- 95
}
## P/F ratio
patients_icu$pf_ratio <- patients_icu$pao2*7.5/(patients_icu$fio2/100)
## ventilatory ratio
## VR=(VEmeasured*PaCO2measured)/(VEpredicted*PaCO2perdicted)=(RR*8*IBW*PaCO2measured)/(100*IBW*5)=RR*PaCO2/62.5
## IBW=22*height^2 (not needed)
patients_icu$vent_ratio <- patients_icu$rr*patients_icu$paco2*7/500
## correct ddimer units
## logical <- (patients_icu$ddimer-round(patients_icu$ddimer))>0
## logical[is.na(logical)] <- F
## patients_icu$ddimer[logical] <- patients_icu$ddimer[logical]*1000
## remove patient_ID column from patients_icu to make it possible to merge with patients_char below
patients_icu <- patients_icu[, -which(colnames(patients_icu) %in% "patient_ID")]
colnames(patients_icu)
## ---------------------
## create a patients_char variable. it combines the admission and outcome forms
patients_char_adm <- patients_adm[, -c(col_icu:(col_admin-1))]
patients_char_dis <- cbind(patients_dis$record_id, patients_dis[, c(col_dis:(col_admin-1))])
colnames(patients_char_dis)[1] <- "record_id"
## save col nums for later merging and subcollectioning
num_adm <- ncol(patients_char_adm)
num_dis <- ncol(patients_char_dis)-1
## merge
patients_char <- merge(patients_char_adm, patients_char_dis, by=c("record_id"), all.x=T)
## ------------------------
## calculate CHAR
## ------------------------
## calculate time periods
patients_char$sympt_to_hosp <- as.numeric(patients_char$date_adm - patients_char$date_first_symptoms)
patients_char$sympt_to_dg <- as.numeric(patients_char$date_first_symptoms - patients_char$date_first_symptoms)
patients_char$hosp_to_icu <- as.numeric(patients_char$date_adm_icu - patients_char$date_adm)
patients_char$los_icu <- as.numeric(patients_char$date_discharge_icu - patients_char$date_adm_icu)
patients_char$los_hosp <- as.numeric(as_datetime(patients_char$date_discharge_hosp) - patients_char$date_adm)
## parameters
patients_char$bmi <- patients_char$weight / (patients_char$height/100)^2
## generate simple comorbidities aus dem extended dataset
for(i in 1:nrow(patients_char)){
if(!is.na(patients_char$adm_comorbid___3[i]) &
patients_char$adm_comorbid___3[i] == 1)
patients_char$adm_comorbid_simple___0[i] <- 1
if(!is.na(patients_char$adm_comorbid___0[i]) &
patients_char$adm_comorbid___0[i] == 1)
patients_char$adm_comorbid_simple___1[i] <- 1
if(!is.na(patients_char$adm_comorbid___1[i]) &
patients_char$adm_comorbid___1[i] == 1)
patients_char$adm_comorbid_simple___2[i] <- 1
if((!is.na(patients_char$adm_comorbid___4[i]) &
patients_char$adm_comorbid___4[i] == 1)|
(!is.na(patients_char$adm_comorbid___5[i]) &
patients_char$adm_comorbid___5[i] == 1))
patients_char$adm_comorbid_simple___3[i] <- 1
if((!is.na(patients_char$adm_comorbid___9[i]) &
patients_char$adm_comorbid___9[i] == 1))
patients_char$adm_comorbid_simple___4[i] <- 1
if((!is.na(patients_char$adm_comorbid___15[i]) &
patients_char$adm_comorbid___15[i]==1)|
(!is.na(patients_char$adm_comorbid___16[i]) &
patients_char$adm_comorbid___16[i]==1)|
(!is.na(patients_char$adm_comorbid___17[i]) &
patients_char$adm_comorbid___17[i]==1)|
(!is.na(patients_char$adm_comorbid___18[i]) &
patients_char$adm_comorbid___18[i]==1)|
(!is.na(patients_char$adm_comorbid___19[i]) &
patients_char$adm_comorbid___19[i]==1)|
(!is.na(patients_char$adm_comorbid___20[i]) &
patients_char$adm_comorbid___20[i]==1)|
(!is.na(patients_char$adm_comorbid___21[i]) &
patients_char$adm_comorbid___21[i]==1)|
(!is.na(patients_char$adm_comorbid___22[i]) &
patients_char$adm_comorbid___22[i]==1)|
(!is.na(patients_char$adm_comorbid___23[i]) &
patients_char$adm_comorbid___23[i]==1))
patients_char$adm_comorbid_simple___5[i] <- 1
}
## transform bool vars
patients_char$survivor_icu <- factor(patients_char$survivor_icu)
levels(patients_char$survivor_icu) <- c("ICU non survivor", "ICU survivor")
patients_char$survivor_hosp <- factor(patients_char$survivor_hosp)
levels(patients_char$survivor_hosp) <- c("Hospital non survivor", "Hospital survivor")
patients_char$adm_comorbid_simple___0 <-
factor(patients_char$adm_comorbid_simple___0)
patients_char$adm_comorbid_simple___1 <-
factor(patients_char$adm_comorbid_simple___1)
patients_char$adm_comorbid_simple___2 <-
factor(patients_char$adm_comorbid_simple___2)
patients_char$adm_comorbid_simple___3 <-
factor(patients_char$adm_comorbid_simple___3)
patients_char$adm_comorbid_simple___4 <-
factor(patients_char$adm_comorbid_simple___4)
patients_char$adm_comorbid_simple___5 <-
factor(patients_char$adm_comorbid_simple___5)
patients_char$health_work_yn <- factor(patients_char$health_work_yn)
## merge pat char with icu data
patients_icu <- merge(patients_icu, patients_char, by=c("record_id"), all.x=T)
## perform calculations that require merged data
## ------------------------
## SOFA (in patients_icu)
patients_icu$sofa <- 0
for(i in c(1:nrow(patients_icu))){
rs <- patients_icu$respsupp[i]
mv <- rs==5|is.na(rs)
patients_icu$sofa[i] <-
as.numeric(SOFA(gcs=patients_icu$gcs[i],
pao2=patients_icu$pao2[i],
fio2=patients_icu$fio2[i]/100,
tc=patients_icu$thrombos[i],
bilirubin=patients_icu$bili[i],
msap=patients_icu$map[i],
norepinephrine=patients_icu$nor[i],
weight=patients_icu$weight[i],
creatinine=patients_icu$crea[i],
urine=patients_icu$estimated_urine_output[i]*1000,
mv=mv,
NAtreatasnormal=T)[1])
}
patients_icu$sofa
## SAPS II and APACHE II (in patients_char)
get_param <- function(nm, fn, rec_ID, ds){
res <- NA
res <- patients_icu[patients_icu$record_id==rec_ID &
patients_icu$time %in% ds, nm]
case <- F
if(fn=="max"){res <- max(res, na.rm=T); case <- T}
if(fn=="min"){res <- min(res, na.rm=T); case <- T}
if(!case) res <- NA
if(!is.na(res) & abs(res)==Inf) res <- NA
return(res)
}
days <- c(0, 1)
patients_char$saps <- NA
patients_char$apache <- NA
for(i in c(1:nrow(patients_char))){
## TODO: should choose by max subscore, not by min/max parameter
rec_ID <- patients_char$record_id[i]
age <- patients_char$age[i]
msap <- get_param("crea", "min", rec_ID, days)
hr <- get_param("hrmin", "min", rec_ID, days)
ssap <- msap + 40
rr <- get_param("rr", "max", rec_ID, days)
## cave 0 may probably mean NA or anuria if data quality is bad...
urine <- get_param("estimated_urine_output", "max", rec_ID, 1)
temp <- get_param("temp", "max", rec_ID, days)
pao2 <- get_param("pao2", "min", rec_ID, days)
paco2 <- get_param("paco2", "max", rec_ID, days)
fio2 <- get_param("fio2", "max", rec_ID, days) / 100
HST <- get_param("HST", "max", rec_ID, days)
hct <- get_param("hct", "min", rec_ID, days)
leuco <- get_param("leuco", "max", rec_ID, days)
bilirubin <- get_param("bilirubin", "max", rec_ID, days)
creatinine <- get_param("creatinine", "max", rec_ID, days)
ph <- get_param("ph", "min", rec_ID, days)
rs <- get_param("respsupp", "max", rec_ID, 1)
mv <- rs==5|is.na(rs)
hco3 <- get_param("hco3", "min", rec_ID, days)
gcs <- get_param("gcs", "min", rec_ID, days)
## adm_comorbid_simple 5: all categories (cancer, hemat, aids)
hematcancer <- patients_char$adm_comorbid_simple___5[i]==1
natrium <- patients_char$sodium[i]
kalium <- patients_char$potassium[i]
medical_adm <- T
akf <- patients_char$complications_short__2[i]==1|
patients_icu$rescue_resp__3[patients_icu$record_id==rec_ID &
patients_icu$time==1]==1
if(length(akf)==0) akf <- NA
## immunocompr and chronicorganfailure are covered by the conversion of the extended dataset into adm_comorbid_simple___5
scores <- SAPS(age=age, HR=hr, ssap=ssap, temp=temp, pao2=pao2, fio2=fio2,
mv=mv, urine=urine, HST=HST, leuco=leuco, kalium=kalium,
natrium=natrium, hco3=hco3, bilirubin=bilirubin,
gcs=gcs, hematcancer=hematcancer, medical_adm=medical_adm,
msap=msap, RR=rr, paco2=paco2, pH=ph, creatinine=creatinine,
hct=hct, akf=akf,
NAtreatasnormal=T)
## scores <- data.frame(SAPS=1, APACHEphysio=1, APACHE=1)
patients_char$saps[i] <- as.numeric(scores$SAPS)
patients_char$apache[i] <- as.numeric(scores$APACHE)
patients_char$apache_physio[i] <- as.numeric(scores$APACHEphysio)
## also copy sofa day 0/1 into the CHAR data.frame
sofa <- get_param("sofa", "max", rec_ID, days)
patients_char$sofa[i] <- sofa
}
patients_char$saps
patients_char$apache_physio
patients_char$apache
## create patients_adm dataframe (only the admission criteria, useful to build prediction models) from patients_char.this step is repeated here to include all calculated variables.
patients_adm <- patients_char[, c(1:num_adm, (num_adm+num_dis+1):ncol(patients_char))]
## remove los, death
rm <- which(substr(colnames(patients_adm),1,3)=="los"|
substr(colnames(patients_adm),1,5)=="death")
patients_adm <- patients_adm[, -rm]