-
Notifications
You must be signed in to change notification settings - Fork 0
/
05b-patient-level-design.R
264 lines (198 loc) · 10.2 KB
/
05b-patient-level-design.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
## Header ------------------------------------------------------------
##
## M19 PHS 5254 Using Administrative Data for Health Services Research
## Washington University School of Medicine in St. Louis
##
## Demonstrate typical data management tasks for a patient-level
## analysis.
# Sample problem:
# 1. Identify admissions with a diagnosis of fall.
# 2. Select the first such admission per patient.
# 3. Obtain all hospitalizations for these patients.
# 4. Recreate the data set of index admissions to use for index
# admission-specific inclusion/exclusion criteria and covariates.
# 5. Create a data set consisting of index admissions and any
# admissions in the year prior, which will be used for
# inclusion/exclusion criteria and to define baseline covariates.
# 6. Create a data set of admissions in the three months following
# index admissions to determine outcomes.
## Setup -------------------------------------------------------------
library(tidyverse)
# setwd("//storage1.ris.wustl.edu/colditzg/Active/admin_course_jsahrmann")
source("../admin_course_data/code2023/coder.R")
## Constant definitions ----------------------------------------------
## Diagnoses ----------------------
# Slipping, tripping, stumbling and falls
# W00--W19
dx10_fall <- c(
"W000XXA", "W000XXD", "W000XXS", "W001XXA", "W001XXD", "W001XXS",
"W002XXA", "W002XXD", "W002XXS", "W009XXA", "W009XXD", "W009XXS",
"W010XXA", "W010XXD", "W010XXS", "W0110XA", "W0110XD", "W0110XS",
"W01110A", "W01110D", "W01110S", "W01111A", "W01111D", "W01111S",
"W01118A", "W01118D", "W01118S", "W01119A", "W01119D", "W01119S",
"W01190A", "W01190D", "W01190S", "W01198A", "W01198D", "W01198S",
"W03XXXA", "W03XXXD", "W03XXXS", "W04XXXA", "W04XXXD", "W04XXXS",
"W050XXA", "W050XXD", "W050XXS", "W051XXA", "W051XXD", "W051XXS",
"W052XXA", "W052XXD", "W052XXS", "W06XXXA", "W06XXXD", "W06XXXS",
"W07XXXA", "W07XXXD", "W07XXXS", "W08XXXA", "W08XXXD", "W08XXXS",
"W090XXA", "W090XXD", "W090XXS", "W091XXA", "W091XXD", "W091XXS",
"W092XXA", "W092XXD", "W092XXS", "W098XXA", "W098XXD", "W098XXS",
"W100XXA", "W100XXD", "W100XXS", "W101XXA", "W101XXD", "W101XXS",
"W102XXA", "W102XXD", "W102XXS", "W108XXA", "W108XXD", "W108XXS",
"W109XXA", "W109XXD", "W109XXS", "W11XXXA", "W11XXXD", "W11XXXS",
"W12XXXA", "W12XXXD", "W12XXXS", "W130XXA", "W130XXD", "W130XXS",
"W131XXA", "W131XXD", "W131XXS", "W132XXA", "W132XXD", "W132XXS",
"W133XXA", "W133XXD", "W133XXS", "W134XXA", "W134XXD", "W134XXS",
"W138XXA", "W138XXD", "W138XXS", "W139XXA", "W139XXD", "W139XXS",
"W14XXXA", "W14XXXD", "W14XXXS", "W15XXXA", "W15XXXD", "W15XXXS",
"W16011A", "W16011D", "W16011S", "W16012A", "W16012D", "W16012S",
"W16021A", "W16021D", "W16021S", "W16022A", "W16022D", "W16022S",
"W16031A", "W16031D", "W16031S", "W16032A", "W16032D", "W16032S",
"W16111A", "W16111D", "W16111S", "W16112A", "W16112D", "W16112S",
"W16121A", "W16121D", "W16121S", "W16122A", "W16122D", "W16122S",
"W16131A", "W16131D", "W16131S", "W16132A", "W16132D", "W16132S",
"W16211A", "W16211D", "W16211S", "W16212A", "W16212D", "W16212S",
"W16221A", "W16221D", "W16221S", "W16222A", "W16222D", "W16222S",
"W16311A", "W16311D", "W16311S", "W16312A", "W16312D", "W16312S",
"W16321A", "W16321D", "W16321S", "W16322A", "W16322D", "W16322S",
"W16331A", "W16331D", "W16331S", "W16332A", "W16332D", "W16332S",
"W1641XA", "W1641XD", "W1641XS", "W1642XA", "W1642XD", "W1642XS",
"W16511A", "W16511D", "W16511S", "W16512A", "W16512D", "W16512S",
"W16521A", "W16521D", "W16521S", "W16522A", "W16522D", "W16522S",
"W16531A", "W16531D", "W16531S", "W16532A", "W16532D", "W16532S",
"W16611A", "W16611D", "W16611S", "W16612A", "W16612D", "W16612S",
"W16621A", "W16621D", "W16621S", "W16622A", "W16622D", "W16622S",
"W16711A", "W16711D", "W16711S", "W16712A", "W16712D", "W16712S",
"W16721A", "W16721D", "W16721S", "W16722A", "W16722D", "W16722S",
"W16811A", "W16811D", "W16811S", "W16812A", "W16812D", "W16812S",
"W16821A", "W16821D", "W16821S", "W16822A", "W16822D", "W16822S",
"W16831A", "W16831D", "W16831S", "W16832A", "W16832D", "W16832S",
"W1691XA", "W1691XD", "W1691XS", "W1692XA", "W1692XD", "W1692XS",
"W170XXA", "W170XXD", "W170XXS", "W171XXA", "W171XXD", "W171XXS",
"W172XXA", "W172XXD", "W172XXS", "W173XXA", "W173XXD", "W173XXS",
"W174XXA", "W174XXD", "W174XXS", "W1781XA", "W1781XD", "W1781XS",
"W1782XA", "W1782XD", "W1782XS", "W1789XA", "W1789XD", "W1789XS",
"W1800XA", "W1800XD", "W1800XS", "W1801XA", "W1801XD", "W1801XS",
"W1802XA", "W1802XD", "W1802XS", "W1809XA", "W1809XD", "W1809XS",
"W1811XA", "W1811XD", "W1811XS", "W1812XA", "W1812XD", "W1812XS",
"W182XXA", "W182XXD", "W182XXS", "W1830XA", "W1830XD", "W1830XS",
"W1831XA", "W1831XD", "W1831XS", "W1839XA", "W1839XD", "W1839XS",
"W1840XA", "W1840XD", "W1840XS", "W1841XA", "W1841XD", "W1841XS",
"W1842XA", "W1842XD", "W1842XS", "W1843XA", "W1843XD", "W1843XS",
"W1849XA", "W1849XD", "W1849XS", "W19XXXA", "W19XXXD", "W19XXXS"
)
## Function definitions ----------------------------------------------
# Print `TRUE` if the given data set is patient-level (i.e., one in
# which the number of `VisitLink`s is equal to the number of records).
is_patient_level <- function(.data) {
n_distinct(.data$VisitLink) == nrow(.data)
}
## Input data --------------------------------------------------------
# Read the full SID data. (The full SID contains a moderate number of
# records where `VisitLink` and `DaysToEvent` are missing. These are
# key to a patient-level analysis, so we can discard records where
# they're missing at the outset.)
core <- bind_rows(
read_rds("../admin_course_data/fl_sidc_2015q4_core.rds"),
read_rds("../admin_course_data/fl_sidc_2016_core.rds"),
read_rds("../admin_course_data/fl_sidc_2017_core.rds"),
read_rds("../admin_course_data/fl_sidc_2018_core.rds"),
read_rds("../admin_course_data/fl_sidc_2019_core.rds")
) %>%
filter(!is.na(VisitLink))
# 646 s
## 1. Primary inclusion criterion ------------------------------------
# Identify admissions with a diagnosis of fall.
core_fall <- core %>%
mutate_flag_dx(codes = dx10_fall, name = dx_fall) %>%
filter(dx_fall == 1)
is_patient_level(core_fall)
# (Note that we cannot remove `core` at this point because we'll need
# it for the 'all records' data in step 3.)
# Record the initial sample size (in number of admissions).
nrow(core_fall) # 572477
## 2. Earliest admissions per patient --------------------------------
# Use `group_by`/`summarise` to create a data set containing each
# patient and the 'date' of his/her earliest admission with a
# fall. (Date is in quotes because we're using `DaysToEvent` rather
# than a true date.)
pat_earliest_fall <- core_fall %>%
group_by(VisitLink) %>%
summarise(earliest_fall = min(DaysToEvent))
pat_earliest_fall
nrow(pat_earliest_fall) # 458647
is_patient_level(pat_earliest_fall)
## 3. 'All records' data set -----------------------------------------
# Merge our patient-level data set to the full SID data to obtain all
# hospitalizations for these patients.
all_records <- pat_earliest_fall %>%
inner_join(core, by = "VisitLink")
nrow(all_records) # 1729605
is_patient_level(all_records)
rm(core)
## 4. Index records data set -----------------------------------------
# (Attempt to) recreate the patient-level data set of earliest
# falls. In creating the all records data set, we kept the variable
# `earliest_fall`, so we can compare that to `DaysToEvent` to get
# these records.
not_quite_pat_level_earliest_fall <- all_records %>%
filter(DaysToEvent == earliest_fall)
nrow(not_quite_pat_level_earliest_fall) # 459693
is_patient_level(not_quite_pat_level_earliest_fall)
nrow(not_quite_pat_level_earliest_fall) -
n_distinct(not_quite_pat_level_earliest_fall$VisitLink)
# This data set isn't patient-level! That means that there are a small
# number of records with identical `VisitLink`s and
# `DaysToEvent`s. These could be mistakes, but we can't be
# sure. Dealing with this would require further investigation and is
# beyond the scope of this demo. These decisions often rely on
# discussion between programmers and clinical investigators.
## 5. Baseline records data set --------------------------------------
# Create a data set of index admissions and any admissions in the year
# prior.
dsch_time_to_index <- all_records %>%
mutate(time_to_index = DaysToEvent - earliest_fall)
summary(dsch_time_to_index$time_to_index)
# `time_to_index` < 0 => before index admission date
# `time_to_index` >= 0 => after index admission date
dsch_baseline <- dsch_time_to_index %>%
filter(between(time_to_index, -365, 0))
nrow(dsch_baseline) # 753742
is_patient_level(dsch_baseline)
## 6. Follow-up records data set -------------------------------------
# Create a data set of admissions in the three months following index
# admissions.
dsch_follow_up <- dsch_time_to_index %>%
filter(between(time_to_index, 1, 90))
nrow(dsch_follow_up)
is_patient_level(dsch_follow_up)
## Bonus: Producing an analytic data set -----------------------------
# Define dummy indicator variables for each of the data sets produced
# in parts 4--6.
## Index records data set ---------
not_quite_pat_level_earliest_fall2 <- not_quite_pat_level_earliest_fall %>%
mutate(index_record = ifelse(DaysToEvent == earliest_fall, 1, 0))
pat_index <- not_quite_pat_level_earliest_fall2 %>%
group_by(VisitLink) %>%
summarise(has_index_record = max(index_record))
## Baseline records data set ------
dsch_baseline2 <- dsch_baseline %>%
mutate(pre_index_record = ifelse(DaysToEvent < earliest_fall, 1, 0))
pat_baseline <- dsch_baseline2 %>%
group_by(VisitLink) %>%
summarise(has_pre_index_record = max(pre_index_record))
## Follow-up records data set -----
dsch_follow_up2 <- dsch_follow_up %>%
mutate(post_index_record = ifelse(DaysToEvent > earliest_fall, 1, 0))
pat_follow_up <- dsch_follow_up2 %>%
group_by(VisitLink) %>%
summarise(has_post_index_record = max(post_index_record))
## Analytic data set --------------
# Merge the patient-level data sets to produce a single data set with
# variables from index, baseline, and follow-up periods.
pat_analytic <- pat_index %>%
left_join(pat_baseline, by = "VisitLink") %>%
left_join(pat_follow_up, by = "VisitLink")
pat_analytic
nrow(pat_analytic)
is_patient_level(pat_analytic)