Calculate percent from total observations in r gtsummary::tbl_summary?

Question

Issue: In gtsummary the tbl_summary function calculates column percent out of the total non-missing observations. I would like gtsummary to calculate percent from the total of missing and non-missing observations.

Example from the gtsummary Table Gallery at http://www.danieldsjoberg.com/gtsummary/articles/gallery.html

trial[c("trt", "age", "grade")] %>%
  tbl_summary(
    by = trt, 
    missing = "no",
    statistic = all_continuous() ~ "{median} ({p25}, {p75}) [N = {N_nonmiss}]"
  ) %>%
  modify_header(stat_by = md("**{level}**<br>N =  {n} ({style_percent(p)}%)")) %>%
  add_n() %>%
  bold_labels() %>%
  modify_spanning_header(starts_with("stat_") ~ "**Chemotherapy Treatment**")

Grade has no missing observations so 35 people with Grade 1 disease in the Drug A group is 35/98 (36%).

Now, recoding Grade 3 to missing:

trial$grade[trial$grade %in% "III"] <- NA
trial$grade <- droplevels(trial$grade)

Re-run tbl_summary:

trial[c("trt", "age", "grade")] %>%
  tbl_summary(
    by = trt, 
    missing = "no",
    statistic = all_continuous() ~ "{median} ({p25}, {p75}) [N = {N_nonmiss}]"
  ) %>%
  modify_header(stat_by = md("**{level}**<br>N =  {n} ({style_percent(p)}%)")) %>%
  add_n() %>%
  bold_labels() %>%
  modify_spanning_header(starts_with("stat_") ~ "**Chemotherapy Treatment**")

Grade 1 is now expressed as n = 35 out of 67 (52%) non-missing obs. in the Drug A group. I would still like the percent to be expressed as 36% of 98 people. Is there a way to do this in gtsummary?

Daniel D. Sjoberg Daniel D. Sjoberg · Accepted Answer · 2020-08-28T20:50:40

I think the best way to get what you're looking for is to make the missing values explicit NA using the forcats::fct_explicit_na() function. When the NA value is a level of a factor, it'll be included in the denominator for percentage calculations.

library(gtsummary)
library(tidyverse)


trial %>%
  select(response, trt) %>%
  # make missing value explicit for categorical variables, using fct_explicit_na
  mutate(response = factor(response) %>% fct_explicit_na()) %>%
  # summarize data
  tbl_summary(by = trt)

Does this solution work for you?

Calculate percent from total observations in r gtsummary::tbl_summary?

1 Answers