Stratified Table 1 using a svydesign object and tbl_svysummary? - r

Question

I'm trying to create a Table 1 for NHANES survey data, first stratified by a binary variable for obese vs non-obese, then stratified again by a binary variable for control/trt group status ("wlp_yn"). I want to get counts (%) for categorical characteristics and means (SE) for continuous baseline characteristics. For these counts and means, I am trying to get p-values as well.
I've tried using tbl_svysummary(), svyby(), tbl_strata(), and CreateTableOne() without any success.
In the code below, I subset the full dataset into a smaller dataset of only control group data ("obese_adults") to divide up the table first. I am also starting out with age for the characteristics ("age_group" is categorical version of "RIDAGEYR" continuous variable). I couldn't figure it out, but I'm curious if there's another way to code this?
add_p_svysummary_ex1 <-
obese_adults %>%
tbl_svysummary(by = wlp_yn, percent = "row", include = c(age_group, RIDAGEYR),
statistic = list(all_continuous() ~ "{mean} ({sd})")) %>%
add_p()
add_p_svysummary_ex1
svyby(~RIDAGEYR, ~age_group+wlp_yn, obese_adults, svymean) # avg age of each age group
Thanks in advance! Would really appreciate any help.
Edit: This is a simplified version of the code for reproducibility
# DEMO
demo <- nhanes('DEMO')
demo_vars <- names(demo)
demo2 <- nhanesTranslate('DEMO', demo_vars, data = demo)
# PRESCRIPTION MEDICATIONS
rxq_rx <- nhanes('RXQ_RX')
rxq_rx_vars <- names(rxq_rx)
rxq_rx2 <- nhanesTranslate('RXQ_RX', rxq_rx_vars, data = rxq_rx)
rxq_rx2 <- rxq_rx2 %>% select("SEQN", "RXD240B") %>% filter(!is.na(RXD240B)) %>% group_by(SEQN) %>% dplyr::summarise(across(everything(), ~toString(na.omit(.))))
nhanesAnalysis = join_all(list(demo2, rxq_rx2), by = "SEQN", type = "full")
# Reconstructing survey weights for combining 1999-2018 - Combining ten survey cycles (twenty years)
nhanesAnalysis$wtint20yr <- ifelse(nhanesAnalysis$SDDSRVYR %in% c(1,2), (2/10 * nhanesAnalysis$WTINT4YR), # for 1999-2002
(1/10 * nhanesAnalysis$WTINT2YR)) # for 2003-2018
# sample weights
nhanesDesign <- svydesign(id = ~SDMVPSU,
strata = ~SDMVSTRA,
weights = ~wtint20yr,
nest = TRUE,
data = nhanesAnalysis)
# subset
obese_adults <- subset(nhanesDesign, (obesity == 1 & !is.na(BMXBMI) & RIDAGEYR >= 60))

Answer 1

Is this what you are looking for. A double dummy split:
library(gtsummary)
library(tidyverse)
data(mtcars)
mtcars %>%
select(am, cyl, hp, vs) %>%
dplyr::mutate(
vs = factor(vs, labels = c("Obese", "Non-Obese")),
am = factor(am, labels = c("Control", "Treatment")),
cyl = paste(cyl, "Cylinder")
) %>%
tbl_strata(
strata = vs,
~.x %>%
tbl_summary(
by = am,
type = where(is.numeric) ~ "continuous"
) %>%
modify_header(all_stat_cols() ~ "**{level}**")
)
I'm not sure why you like to use tbl_svysummary() here, it's made for survey weights.