Examples of workflows enabled by docker
Introduction
One of the goals of the Epiconductor project is to make the analysis
of NHANES data simpler. To this end, it provides tools that can be
used to (a) download all public NHANES data and and load it into an
SQL database, and (b) make snapshots of this database available
locally through docker images. The primary goal in distributing the
pre-built docker images is to provide a ready-to-use tool for data
analysts who may not be conversant with database technologies. The
docker image thus includes R and RStudio, along with a version of the
nhanesA R package that is configured to obtain data from the SQL
database rather than the CDC website. We expect most users of the
docker image to use it through the nhanesA package. Such analyses
will run even outside the docker environment as well, albeit more
slowly as it will involve downloading data from the CDC website.
The potential uses of the docker image goes beyond this, however. As
NHANES collects and publishes data on a continuous basis, it is
natural to want to combine data across cycles. This is not
straightforward because a certain variable of interest may be present
in different tables in different cyles, or may not even be present in
all cycles. With all data locally available, it is possible to
efficiently check where variables are to be found (if at all), examine
differences over NHANES cycles, and detect a range of anomalies that
any analyst needs to be aware of, but would normally be difficult to
find. This vignette provides some illustrative examples, and describes
some relevant tools in the phonto package that could be useful to
practitioners.
Combining data across cycles
We start with a straightforward illustration of the process of
combining data across cycles, where we combine the demographic data
table from all cycles. NHANES table names typically have a suffix; the
suffixes _A, _B, _C,and so on generally correspond to NHANES
cycle years from 1999-2000, 2001-2002, 2003-2004, etc. However, it is
important to highlight that not every table strictly adheres to this
naming convention. For instance, while DEMO_B and DEMO_C are
associated with the 2001-2002 and 2003-2004 cycles, respectively, the
corresponding table for the 1999-2000 cycle is named ‘DEMO’, without
the _A suffix. While this pattern holds for most tables, certain
tables such as SSAFB_A and SSANA_A from the 1999-2000 cycle do
include the _A suffix. To assist users in navigating these
variations, the nhanesA package includes the
nhanesSearchTableNames() function, which allows users to easily
locate all table names containing a specific string, thus simplifying
the process of identifying relevant table names.
library(nhanesA)
demo_all <- nhanesSearchTableNames("DEMO")
demo_all
[1] "DEMO" "DEMO_B" "DEMO_C" "DEMO_D" "DEMO_E" "DEMO_F" "DEMO_G" "DEMO_H"
[9] "DEMO_I" "DEMO_J" "P_DEMO"
The last table in this list merits special mention. During the
2019-2020 cycle, data collection was disrupted by the COVID-19
pandemic. Therefore, the partial 2019-2020 data (herein 2019-March
2020 data) were combined with data from the previous cycle (2017-2018)
to create a nationally representative sample covering 2017-March 2020.
These data files have the same basic file name, e.g., DEMO, but add
the prefix P_. These ‘pre-pandemic’ files require special handling
and the CDC has provided substantial guidance as well as updated
survey weights.
We can now collect all these datasets using nhanes()
function. Normally, this process is likely to be somewhat slow as
several files will need to be downloaded from the CDC website. Within
the docker environment, however, it should be almost instantaneous.
all_demo_data <- sapply(demo_all, nhanes, simplify = FALSE)
object.size(all_demo_data) # ~45 MB
47137304 bytes
sapply(all_demo_data, dim)
DEMO DEMO_B DEMO_C DEMO_D DEMO_E DEMO_F DEMO_G DEMO_H DEMO_I DEMO_J P_DEMO
[1,] 9965 11039 10122 10348 10149 10537 9756 10175 9971 9254 15560
[2,] 145 37 44 43 43 43 48 47 47 46 29
The first row in the output above gives the number of participants in
each cycle, and the second row denotes the number of variables in the
corresponding DEMO table. We can see that each cycle has around
10,000 participants, who are unique across cycles. Note, however, that
the larger number of participants in the P_DEMO dataset is
misleading, because many of these participants are actually from the
previous cycle as described above. We will drop this table before
combining the remaining datasets.
The differing number of variables across cycles indicate that
variables are not measured consistently across cycles. In fact, many
variables included in the DEMO table in the first cycle were
subsequently included in other tables, and others have been dropped
altogether or added. We can make a list of the variables that are
common to all DEMO tables, and combine the corresponding data
subsets together, as follows.
all_demo_data <- head(all_demo_data, -1)
common_vars <- lapply(all_demo_data, names) |> Reduce(f = intersect)
common_vars
[1] "SEQN" "SDDSRVYR" "RIDSTATR" "RIDEXMON" "RIAGENDR" "RIDRETH1"
[7] "DMDCITZN" "DMDYRSUS" "DMDEDUC3" "DMDEDUC2" "DMDMARTL" "RIDEXPRG"
[13] "DMDHRGND" "RIDAGEYR" "RIDAGEMN" "DMDHHSIZ" "INDFMPIR" "WTINT2YR"
[19] "WTMEC2YR" "SDMVPSU" "SDMVSTRA"
demo_combined <-
lapply(all_demo_data, `[`, common_vars) |>
do.call(what = rbind) |>
transform(cycle = substring(SDDSRVYR, 8, 17))
dim(demo_combined)
[1] 101316 22
The combined dataset can be analysed further using standard tools. For
example, the following code uses the lattice package to summarize
the number of participants by recorded ethnicity and gender by cycle.
library("lattice")
demo_combined |>
xtabs(~ cycle + RIAGENDR + RIDRETH1, data = _) |>
array2DF() |>
dotplot(Value ~ cycle | RIAGENDR,
groups = RIDRETH1,
layout = c(1, 2), type = "b",
par.settings = simpleTheme(pch = 16),
auto.key = list(columns = 3))
plot of chunk demoplot
One must be cautious when combining data across
cycles, because the NHANES data are sometimes inconsistent in
unexpected ways. As a simple example, consider the DMDEDUC3
variable, which records education level of children and youth. The
following code illustrates that the values of this variable have
inconsistent capitalization in different cycles.
xtabs(~ cycle + DMDEDUC3, demo_combined)[, 1:4]
DMDEDUC3
cycle 10th grade 10th Grade 11th grade 11th Grade
1999-2000 0 307 0 281
2001-2002 0 324 0 319
2003-2004 0 284 0 295
2005-2006 0 307 0 277
2007-2008 0 154 0 165
2009-2010 0 193 0 185
2011-2012 167 0 150 0
2013-2014 162 0 186 0
2015-2016 160 0 153 0
2017-2018 139 0 155 0
Cross-cycle Consistency checks using variable codebooks
In our experience, inconsistencies such as the change in capitalization described above occur quite often, and in a variety of different ways, requiring attention to detail when combining data from across cycles. These inconsistencies are not necessarily mistakes, as NHANES questionnaires and variables are often modified from cycle to cycle. The primary source that must be consulted to identify such inconsistencies are the per-table documentation, and in particular the variable codebooks giving details of how each variable is recorded.
The NHANES database contains the variable codebooks for all tables
across all cycles in a single database table called
Metadata.VariableCodebook. Once imported into R, this information
can be manipulated in various ways to glean information of
interest. See the codebook
diagnostics
vignette for a different set of diagnostics that are enabled by having
easy access to this information.
library("phonto")
dim(all_cb <- nhanesQuery("select * from Metadata.VariableCodebook"))
[1] 202018 7
## dim(all_cb <- metaData("Codebook")) # alt interface
all_cb <- dplyr::filter(all_cb, !startsWith(TableName, "P_")) # skip pre-pandemic tables
An analyst would typically be interested in some specific variables that are relevant to their study. The first step is to identify how many cycles these variables were recorded in. To this end, we may start by examining the number of tables each variable appears in across all cycles of continuous NHANES.
var_freq <-
all_cb[c("Variable", "TableName")] |> unique() |>
xtabs(~ Variable, data = _) |>
sort(decreasing = TRUE)
table(var_freq)
var_freq
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
4720 3861 2358 564 373 696 776 529 266 365 49 46 263 50 11 23
17 18 19 20 21 22 23 25 28 29 30 32 40 56 58 66
2 33 3 48 1 4 1 3 8 1 1 2 1 1 1 1
74 111
1 1
Not surprisingly, many variables appear only once or twice across all
cycles, and for such variables combining data across cycles would not
be interesting. Variables that appear in multiple cycles, but only
once per cycle, may usually be merged as in the DEMO example above,
but note that they may appear in different tables. One must be careful
about ensuring that the variable is measuring the same quantity in all
cycles. This is usually true, but not always. For example, the range
of values for the LBCBHC variable in different cycles seem to
exhibit some discrepancy.
subset(all_cb, Variable == "LBCBHC")[1:5]
Variable TableName CodeOrValue ValueDescription Count
150088 LBCBHC PSTPOL_D 0.004 to 2.3147 Range of Values 247
150089 LBCBHC PSTPOL_D . Missing 0
150168 LBCBHC PSTPOL_E 0.0044 to 5.2617 Range of Values 264
150169 LBCBHC PSTPOL_E . Missing 0
150249 LBCBHC PSTPOL_F 3.465 to 4801 Range of Values 295
150250 LBCBHC PSTPOL_F . Missing 6
150324 LBCBHC PSTPOL_G 3.536 to 5401 Range of Values 251
150325 LBCBHC PSTPOL_G . Missing 0
150399 LBCBHC PSTPOL_H 3.536 to 5648 Range of Values 281
150400 LBCBHC PSTPOL_H . Missing 3
150467 LBCBHC PSTPOL_I 3.536 to 1330 Range of Values 265
150468 LBCBHC PSTPOL_I . Missing 0
More useful information about the variables may be obtained from the
Metadata.QuestionnaireVariables table in the database, which
contains one row for each variable in each table containing its
description, target group, etc., obtained from the HTML documentation
of NHANES tables.
dim(all_var <- phonto::nhanesQuery("select * from Metadata.QuestionnaireVariables"))
[1] 50714 11
## dim(all_var <- metaData("Variables"))
subset(all_var, Variable == "LBCBHC")[1:5]
Variable TableName Description
43244 LBCBHC PSTPOL_D Beta-hexachlorocyclohexane (ng/g)
43277 LBCBHC PSTPOL_E Beta-hexachlorocyclohexane (ng/g)
43310 LBCBHC PSTPOL_F Beta-hexachlorocyclohexane (pg/g)
43340 LBCBHC PSTPOL_G Beta-hexachlorocyclohexane (pg/g)
43370 LBCBHC PSTPOL_H Beta-hexachlorocyclohexane (pg/g)
43397 LBCBHC PSTPOL_I Beta-hexachlorocyclohexane (pg/g)
Target
43244 Both males and females 12 YEARS - 150 YEARS
43277 Both males and females 12 YEARS - 150 YEARS
43310 Both males and females 12 YEARS - 150 YEARS
43340 Both males and females 12 YEARS - 150 YEARS
43370 Both males and females 12 YEARS - 150 YEARS
43397 Both males and females 12 YEARS - 150 YEARS
SasLabel
43244 Beta-hexachlorocyclohexane (ng/g)
43277 Beta-hexachlorocyclohexane (ng/g)
43310 Beta-hexachlorocyclohexane (pg/g)
43340 Beta-hexachlorocyclohexane (pg/g)
43370 Beta-hexachlorocyclohexane (pg/g)
43397 Beta-hexachlorocyclohexane (pg/g)
This shows that the unit of measurement was changed from the 2009–2010 cycle, explaining the discrepancy. Without a careful check, such changes may be overlooked, leading to errors in interpretation.
It is not easy to systematically detect such changes without manual
inspection of variables of interest. One way to shortlist possible
candidate variables are to identify those for whom the Description
or SasLabel field has changed. Unfortunately, such changes happen
frequently for completely benign reasons, leading to many false
positives.
Within-cycle consistency
Somewhat more surprisingly, several variables appear in more tables than there are cycles, which means that they must appear in multiple tables within the same cycle. The following variables appear in more than 20 tables.
var_freq[ var_freq > 20 ]
Variable
URXUCR WTSA2YR WTDRD1 DRDINT WTDR2D RIAGENDR DRABF WTSB2YR
111 74 66 58 56 40 32 32
WTSC2YR WTSAF2YR DR1DAY DR1DRSTZ DR1EXMER DR1LANG DR2DAY DR2DRSTZ
30 29 28 28 28 28 28 28
DR2EXMER DR2LANG RIANSMP RIDAGGRP WTFSM WTSVOC2Y DR1DBIH DR2DBIH
28 28 25 25 25 23 22 22
PHAFSTHR PHAFSTMN DSDSUPP
22 22 21
For such variables, selecting the corresponding subset of all_cb
shows all entries in the codebook tables for that variable, across all
tables and cycles. For example,
subset(all_cb, Variable == "PHAFSTMN")[1:5]
Variable TableName CodeOrValue ValueDescription Count
74503 PHAFSTMN FASTQX_D 0 to 59 Range of Values 8903
74504 PHAFSTMN FASTQX_D . Missing 537
74546 PHAFSTMN FASTQX_E 0 to 59 Range of Values 8832
74547 PHAFSTMN FASTQX_E . Missing 475
74589 PHAFSTMN FASTQX_F 0 to 59 Range of Values 9557
74590 PHAFSTMN FASTQX_F . Missing 278
74632 PHAFSTMN FASTQX_G 0 to 59 Range of Values 8528
74633 PHAFSTMN FASTQX_G . Missing 428
74675 PHAFSTMN FASTQX_H 0 to 59 Range of Values 9182
74676 PHAFSTMN FASTQX_H . Missing 240
74718 PHAFSTMN FASTQX_I 0 to 59 Range of Values 8911
74719 PHAFSTMN FASTQX_I . Missing 254
74761 PHAFSTMN FASTQX_J 0 to 59 Range of Values 7996
74762 PHAFSTMN FASTQX_J . Missing 370
84545 PHAFSTMN GLU_D 0 to 59 Range of Values 3251
84546 PHAFSTMN GLU_D . Missing 101
84559 PHAFSTMN GLU_E 0 to 59 Range of Values 3223
84560 PHAFSTMN GLU_E . Missing 92
84573 PHAFSTMN GLU_F 0 to 59 Range of Values 3535
84574 PHAFSTMN GLU_F . Missing 46
84587 PHAFSTMN GLU_G 0 to 59 Range of Values 3167
84588 PHAFSTMN GLU_G . Missing 72
84598 PHAFSTMN GLU_H 0 to 59 Range of Values 3291
84599 PHAFSTMN GLU_H . Missing 38
92762 PHAFSTMN INS_H 0 to 59 Range of Values 3291
92763 PHAFSTMN INS_H . Missing 38
92776 PHAFSTMN INS_I 0 to 59 Range of Values 3135
92777 PHAFSTMN INS_I . Missing 56
110377 PHAFSTMN OGTT_D 0 to 59 Range of Values 3251
110378 PHAFSTMN OGTT_D . Missing 101
110395 PHAFSTMN OGTT_E 0 to 59 Range of Values 3223
110396 PHAFSTMN OGTT_E . Missing 92
110430 PHAFSTMN OGTT_F 0 to 59 Range of Values 3163
110431 PHAFSTMN OGTT_F . Missing 0
110465 PHAFSTMN OGTT_G 0 to 59 Range of Values 2815
110466 PHAFSTMN OGTT_G . Missing 0
110505 PHAFSTMN OGTT_H 0 to 59 Range of Values 2909
110506 PHAFSTMN OGTT_H . Missing 0
148145 PHAFSTMN PH 0 to 59 Range of Values 8350
148146 PHAFSTMN PH . Missing 482
148188 PHAFSTMN PH_B 0 to 59 Range of Values 9480
148189 PHAFSTMN PH_B . Missing 449
148231 PHAFSTMN PH_C 0 to 59 Range of Values 9013
148232 PHAFSTMN PH_C . Missing 166
Inspection of this table shows that the PHAFSTMN variable was
initially recorded in the PH table for the first three cycles, after
which it was recorded in three different tables (FASTQX, GLU, and
OGTT) for several cycles, before being dropped again from the latter
two tables. It is natural to wonder whether all these tables contain
the same data. This can only be verified by comparing the actual data,
which we will not do for this example, but some hints are provided by
the data counts included in the codebook. For example, for the
2005–2006 cycle, we have
subset(all_cb, Variable == "PHAFSTMN" & endsWith(TableName, "_D"))[1:5]
Variable TableName CodeOrValue ValueDescription Count
74503 PHAFSTMN FASTQX_D 0 to 59 Range of Values 8903
74504 PHAFSTMN FASTQX_D . Missing 537
84545 PHAFSTMN GLU_D 0 to 59 Range of Values 3251
84546 PHAFSTMN GLU_D . Missing 101
110377 PHAFSTMN OGTT_D 0 to 59 Range of Values 3251
110378 PHAFSTMN OGTT_D . Missing 101
From the variable metadata table, we see that
subset(all_var, Variable == "PHAFSTMN" & endsWith(TableName, "_D"))[c(1, 2, 4, 5)]
Variable TableName Target
24274 PHAFSTMN FASTQX_D Both males and females 1 YEARS - 150 YEARS
25529 PHAFSTMN GLU_D Both males and females 12 YEARS - 150 YEARS
31383 PHAFSTMN OGTT_D Both males and females 12 YEARS - 150 YEARS
SasLabel
24274 Total length of food fast minutes
25529 Total length of 'food fast' minutes
31383 Total length of 'food fast' minutes
While not definitive, this suggests that the PHAFSTMN variable
measures the same quantity in all three tables, and the difference in
number of observations may be due to the difference in target age
group.
Even if a preliminary inspection suggests no obvious problems, one
should verify by comparing the actual recorded data. For example,
consider the ENQ100 variable, which appears in both ENX and SPX
tables.
subset(all_var, Variable == "ENQ100" & endsWith(TableName, "_E"))[c(1, 2, 4, 5)]
Variable TableName Target
23932 ENQ100 ENX_E Both males and females 6 YEARS - 79 YEARS
45401 ENQ100 SPX_E Both males and females 6 YEARS - 79 YEARS
SasLabel
23932 Cough cold resp illness 7 days?
45401 Had respiratory illness?
merge(nhanes("ENX_E")[c("SEQN", "ENQ100")],
nhanes("SPX_E")[c("SEQN", "ENQ100")], by = "SEQN") |>
xtabs(~ ENQ100.x + ENQ100.y, data =_, addNA = TRUE)
ENQ100.y
ENQ100.x Don't know No Yes <NA>
Don't know 3 0 0 0
No 0 5013 0 0
Yes 0 0 1354 0
<NA> 0 217 63 1093
Comparing the records in the two tables after matching by SEQN, the
participant identifier, we see that even though most records are
consistent, several records with Yes or No answers in the SPX
tables are recorded as NA (missing) in the ENX tables.
A more egregious example, where the same variable is clearly measuring
two different things, is provided by the LBXHCT variable.
subset(all_cb, Variable == "LBXHCT" & endsWith(TableName, "_H"))[1:5]
Variable TableName CodeOrValue ValueDescription Count
24512 LBXHCT CBC_H 17.9 to 56.5 Range of Values 8544
24513 LBXHCT CBC_H . Missing 878
33809 LBXHCT COT_H 0.011 to 1150 Range of Values 8029
33810 LBXHCT COT_H . Missing 884
subset(all_var, Variable == "LBXHCT" & endsWith(TableName, "_H"))[c(1, 2, 3, 4, 5)]
Variable TableName Description
9538 LBXHCT CBC_H Hematocrit (%)
11531 LBXHCT COT_H Hydroxycotinine Serum (ng/mL)
Target SasLabel
9538 Both males and females 1 YEARS - 150 YEARS Hematocrit (%)
11531 Both males and females 3 YEARS - 150 YEARS Hydroxycotinine Serum (ng/mL)
merge(nhanes("CBC_H")[c("SEQN", "LBXHCT")],
nhanes("COT_H")[c("SEQN", "LBXHCT")], by = "SEQN") |> head()
SEQN LBXHCT.x LBXHCT.y
1 73557 45.4 1.330
2 73558 36.7 4.480
3 73559 49.9 0.055
4 73560 37.8 0.025
5 73561 43.8 0.011
6 73562 41.5 0.011
Quality control reports
Getting a comprehensive picture of the issues relevant for a
particular variable may take some work, as seen in the above
examples. However, much of the preliminary work can be automated, and
the qc_var() function in the phonto package dowa precisely this.
The report for the LBCBHC variable can be generated using the
following call.
qc_var("LBCBHC")
Variable: LBCBHC
Mismatch in Description:
description Frequency
1 Beta-hexachlorocyclohexane (ng/g) 2
2 Beta-hexachlorocyclohexane (pg/g) 4
Mismatch in Saslabel:
saslabel Frequency
1 Beta-hexachlorocyclohexane (ng/g) 2
2 Beta-hexachlorocyclohexane (pg/g) 4
Similarly, the reports for PHAFSTMN (where the issues flagged are
relatively benign) and LBXHCT (where they are not) are obtained as
follows.
qc_var("PHAFSTMN")
Variable: PHAFSTMN
Appears in multiple tables within same cycle:
cycle Tables
1 1999-2000 PH
2 2001-2002 PH_B
3 2003-2004 PH_C
4 2005-2006 FASTQX_D / GLU_D / OGTT_D
5 2007-2008 FASTQX_E / GLU_E / OGTT_E
6 2009-2010 FASTQX_F / GLU_F / OGTT_F
7 2011-2012 FASTQX_G / GLU_G / OGTT_G
8 2013-2014 FASTQX_H / GLU_H / INS_H / OGTT_H
9 2015-2016 FASTQX_I / INS_I
10 2017-2018 FASTQX_J
11 2017-2020 P_FASTQX
Mismatch in Description:
description
1 The time (in minutes) between when the examinee last ate or drank anything other than water and the time of the venipuncture.
2 The time (in minutes) between when the examinee last ate or drank anything other than water and the time of the venipuncture.
3 Total length of 'food fast' minutes
Frequency
1 1
2 12
3 10
Mismatch in Saslabel:
saslabel Frequency
1 Total length of 'food fast' minutes 12
2 Total length of food fast minutes 11
Mismatch in Target:
target Frequency
1 Both males and females 1 YEARS - 150 YEARS 11
2 Both males and females 12 YEARS - 150 YEARS 12
qc_var("LBXHCT")
Variable: LBXHCT
Appears in multiple tables within same cycle:
cycle Tables
1 1999-2000 LAB25
2 2001-2002 L25_B
3 2003-2004 L25_C
4 2005-2006 CBC_D
5 2007-2008 CBC_E
6 2009-2010 CBC_F
7 2011-2012 CBC_G
8 2013-2014 CBC_H / COT_H
9 2015-2016 CBC_I / COT_I
10 2017-2018 CBC_J / COT_J
11 2017-2020 P_CBC
Mismatch in Description:
description Frequency
1 Hematocrit (%) 11
2 Hydroxycotinine Serum (ng/mL) 3
Mismatch in Saslabel:
saslabel Frequency
1 Hematocrit (%) 11
2 Hydroxycotinine Serum (ng/mL) 3
Mismatch in Target:
target Frequency
1 Both males and females 1 YEARS - 150 YEARS 11
2 Both males and females 3 YEARS - 150 YEARS 3
As we can see, these reports give an incomplete but suggestive initial list of potential issues, which can be followed up on by the analyst if necessary.
Other useful checks that have not yet been implemented but are planned include:
-
Special codes in numeric variables: For numeric variables, special codes are sometimes used to indicate unusual values. A common use is to have special codes to disambiguate various types of missing responses (e.g.,
"Refused"and"Don't know"). Some of these are already converted toNAvalues by default, but others may remain. -
One particular instance of special codes represent data coarsening, that is, clubbing of numeric values over a range (e.g., ages over 80 are usually replaced by 80).
The plan is to report the presence of such codes in variable QC reports.
Variables that are potentially skipped
Some variables in a table may be skipped depending on the response to
a previous question. This information is recorded in the variable
codebooks. For example, in table WHQ, if the response to WHD060 is
“Yes”, then subsequent questions are skipped until WHQ090 is
reached.
nhanesCodebook("WHQ", "WHD060")
$WHD060
$WHD060$`Variable Name:`
[1] "WHD060"
$WHD060$`SAS Label:`
[1] "Weight change intentional"
$WHD060$`English Text:`
[1] "Was the change between {your/SP's} current weight and {your/his/her} weight a year ago intentional?"
$WHD060$`Target:`
[1] "Both males and females 16 YEARS - 150 YEARS"
$WHD060$WHD060
Code.or.Value Value.Description Count Cumulative Skip.to.Item
1 1 Yes 508 508 WHQ090
2 2 No 376 884 <NA>
3 7 Refused 0 884 <NA>
4 9 Don't know 0 884 <NA>
5 . Missing 5160 6044 <NA>
This information, while useful, does not directly tell us which
variables were skipped. This may be important for the analyst to know,
because a variable that was skipped is recorded as a missing value,
whereas we may actually have additional information based on whether
it was skipped or not. In the above example, the immediately following
variable is WHQ070.
nhanesCodebook("WHQ", "WHQ070")
$WHQ070
$WHQ070$`Variable Name:`
[1] "WHQ070"
$WHQ070$`SAS Label:`
[1] "Tried to lose weight in past year"
$WHQ070$`English Text:`
[1] "During the past 12 months {have you/has SP} tried to lose weight?"
$WHQ070$`Target:`
[1] "Both males and females 16 YEARS - 150 YEARS"
$WHQ070$WHQ070
Code.or.Value Value.Description Count Cumulative Skip.to.Item
1 1 Yes 1618 1618 <NA>
2 2 No 3889 5507 WHQ090
3 7 Refused 8 5515 WHQ090
4 9 Don't know 11 5526 WHQ090
5 . Missing 518 6044 <NA>
This codebook gives no indication of whether this question could have
been skipped for some participants. However, from the previous
codebook, was can infer that out of the 518 missing values, 508 were
presumably from those that said “Yes” to WHD060, and these are
qualitatively different from the remaining 10. While the question of
how to address such situations is for the analyst to answer, it
would be useful to flag variables that are potentially skipped.
The get_skip_info() function in the phonto package reports the
variables that are potentially skipped in any given table. The
following code applies it on the WHQ table.
get_skip_info("WHQ")
Table Variable MaybeSkipped SkippedDueTo
1 WHQ WHD010 FALSE
2 WHQ WHD020 FALSE
3 WHQ WHQ030 FALSE
4 WHQ WHD040 FALSE
5 WHQ WHD050 FALSE
6 WHQ WHD060 FALSE
7 WHQ WHQ070 TRUE WHD060
8 WHQ WHD080A TRUE WHD060,WHQ070,WHQ070,WHQ070
9 WHQ WHD080B TRUE WHD060,WHQ070,WHQ070,WHQ070
10 WHQ WHQ080C TRUE WHD060,WHQ070,WHQ070,WHQ070
11 WHQ WHD080D TRUE WHD060,WHQ070,WHQ070,WHQ070
12 WHQ WHQ080E TRUE WHD060,WHQ070,WHQ070,WHQ070
13 WHQ WHD080F TRUE WHD060,WHQ070,WHQ070,WHQ070
14 WHQ WHQ080G TRUE WHD060,WHQ070,WHQ070,WHQ070
15 WHQ WHQ080H TRUE WHD060,WHQ070,WHQ070,WHQ070
16 WHQ WHQ080I TRUE WHD060,WHQ070,WHQ070,WHQ070
17 WHQ WHQ080J TRUE WHD060,WHQ070,WHQ070,WHQ070
18 WHQ WHQ080K TRUE WHD060,WHQ070,WHQ070,WHQ070
19 WHQ WHD080L TRUE WHD060,WHQ070,WHQ070,WHQ070
20 WHQ WHD080M TRUE WHD060,WHQ070,WHQ070,WHQ070
21 WHQ WHD080N TRUE WHD060,WHQ070,WHQ070,WHQ070
22 WHQ WHQ090 FALSE
23 WHQ WHQ100A TRUE WHQ090,WHQ090,WHQ090
24 WHQ WHQ100B TRUE WHQ090,WHQ090,WHQ090
25 WHQ WHD100C TRUE WHQ090,WHQ090,WHQ090
26 WHQ WHD100D TRUE WHQ090,WHQ090,WHQ090
27 WHQ WHQ100E TRUE WHQ090,WHQ090,WHQ090
28 WHQ WHD100F TRUE WHQ090,WHQ090,WHQ090
29 WHQ WHQ100G TRUE WHQ090,WHQ090,WHQ090
30 WHQ WHQ100H TRUE WHQ090,WHQ090,WHQ090
31 WHQ WHQ100I TRUE WHQ090,WHQ090,WHQ090
32 WHQ WHQ100J TRUE WHQ090,WHQ090,WHQ090
33 WHQ WHQ100K TRUE WHQ090,WHQ090,WHQ090
34 WHQ WHD100L TRUE WHQ090,WHQ090,WHQ090
35 WHQ WHD100M TRUE WHQ090,WHQ090,WHQ090
36 WHQ WHD100N TRUE WHQ090,WHQ090,WHQ090
37 WHQ WHD110 FALSE
38 WHQ WHD120 FALSE
39 WHQ WHD130 FALSE
40 WHQ WHD140 FALSE
41 WHQ WHD150 FALSE
42 WHQ WHD160 FALSE
43 WHQ WHD170 FALSE
Searching variables
Having access to all variable descriptions in a single dataset makes search operations convenient. For example,
all_var |> within({ Description <- tolower(Description) }) |>
subset(endsWith(TableName, "_C") &
(grepl("hypertension", Description) |
grepl("blood pressure", Description)),
select = c(1, 2, 5))
Variable TableName SasLabel
8894 BPQ010 BPQ_C Last blood pressure reading by doctor
8895 BPQ020 BPQ_C Ever told you had high blood pressure
8896 BPQ030 BPQ_C Told had high blood pressure - 2+ times
8897 BPQ040A BPQ_C Taking prescription for hypertension
8898 BPQ040B BPQ_C Told to control weight for hypertension
8899 BPQ040C BPQ_C Told to reduce sodium for hypertension
8900 BPQ040D BPQ_C Told to exercise more for hypertension
8901 BPQ040E BPQ_C Told to reduce alcohol for hypertension
8902 BPQ040F BPQ_C Told to do other things for hypertension
9097 PEASCST1 BPX_C Blood Pressure Status
9098 PEASCTM1 BPX_C Blood Pressure Time in Seconds
9099 PEASCCT1 BPX_C Blood Pressure Comment
9112 BPXSY1 BPX_C Systolic: Blood pres (1st rdg) mm Hg
9113 BPXDI1 BPX_C Diastolic: Blood pres (1st rdg) mm Hg
9115 BPXSY2 BPX_C Systolic: Blood pres (2nd rdg) mm Hg
9116 BPXDI2 BPX_C Diastolic: Blood pres (2nd rdg) mm Hg
9118 BPXSY3 BPX_C Systolic: Blood pres (3rd rdg) mm Hg
9119 BPXDI3 BPX_C Diastolic: Blood pres (3rd rdg) mm Hg
9121 BPXSY4 BPX_C Systolic: Blood pres (4th rdg) mm Hg
9122 BPXDI4 BPX_C Diastolic: Blood pres (4th rdg) mm Hg
11863 CVQ220C CVX_C Priority 2 Stop excessive BP
11864 CVQ220E CVX_C Priority 2 Stop significant drop in SBP
11880 CVAARM CVX_C Arm selected for blood pressure monitor
11881 CVACUFF CVX_C Cuff size for blood pressure monitor
11888 CVDWSY CVX_C Warm-up systolic BP (mm Hg)
11889 CVDWDI CVX_C Warm-up diastolic BP (mm Hg)
11895 CVDS1SY CVX_C Stage 1 systolic BP (mm Hg)
11896 CVDS1DI CVX_C Stage 1 diastolic BP (mm Hg)
11902 CVDS2SY CVX_C Stage 2 systolic BP (mm Hg)
11903 CVDS2DI CVX_C Stage 2 diastolic BP (mm Hg)
11906 CVDR1SY CVX_C Recovery 1 systolic BP (mm Hg)
11907 CVDR1DI CVX_C Recovery 1 diastolic BP (mm Hg)
11910 CVDR2SY CVX_C Recovery 2 systolic BP (mm Hg)
11911 CVDR2DI CVX_C Recovery 2 diastolic BP (mm Hg)
11914 CVDR3SY CVX_C Recovery 3 systolic BP (mm Hg)
11915 CVDR3DI CVX_C Recovery 3 diastolic BP (mm Hg)
29638 LEXBRP1 LEXAB_C Brachial SBP 1 (mm Hg)
29639 LEXBRP2 LEXAB_C Brachial SBP 2 (mm Hg)
29640 LEXBRPM LEXAB_C Mean Brachial SBP (mm Hg)
29641 LEXLPTS1 LEXAB_C Left Posterior Tibial SBP 1 (mm Hg)
29642 LEXLPTS2 LEXAB_C Left Posterior Tibial SBP 2 (mm Hg)
29643 LEXLPTSM LEXAB_C Left Mean Posterior Tibial SBP (mm Hg)
29645 LEXRPTS1 LEXAB_C Right Posterior Tibial SBP 1 (mm Hg)
29646 LEXRPTS2 LEXAB_C Right Posterior Tibial SBP 2 (mm Hg)
29647 LEXRPTSM LEXAB_C Right Mean Posterior Tibial SBP (mm Hg)
29651 LEALAPNC LEXAB_C Left ankle SBP > 255 mm Hg?
29652 LEARAPNC LEXAB_C Right ankle SBP > 255 mm Hg?
30223 MCQ250F MCQ_C Blood relatives w/hypertension/stroke
42272 PFD069J PFQ_C Hypertension or high blood pressuredays
Information in this table can be supplemented using the
nhanesTableSummary() function, which computes further information
such as the variable type and number of non-missing observations; see
this vignette for more details and links to
publicly accessible online search interfaces.