Streamline Your Research: Creating Publication-Ready Demographics Tables in R

Author

Jason Gavrilis

Introduction

When publishing studies involving human participants, a demographics table is a vital tool for summarizing the characteristics of the population under investigation.

Figure 1: Demographics are the characteristics of a population that have been categorized by distinct criteria.

However, creating a publication-ready demographics table can be a challenging and time-consuming task. If you’ve ever struggled with this, fear not! The following article will guide you through the process of creating a publication-ready demographics table in R, using the gtsummary and flextable packages. These tools will save you time and effort, allowing you to focus on the insights from your data rather than the intricacies of table formatting.

Prerequisites

Load required packages:

tidyverse for basic data manipulation (Wickham et al. 2019)
gtsummary for creating demographics tables (Sjoberg et al. 2021)
flextable for creating table objects (Gohel and Skintzos 2023)
labelled for adding labels to data (Larmarange 2022)
officer for exporting flextable table objects to Microsoft Word (Gohel 2023)

Code

if (!require(pacman)) install.packages("pacman")
library(pacman)
p_load(
  tidyverse, flextable, gtsummary, labelled, officer
)

Demographics Table

Figure 2 shows an example demographics table published by Ching et al. (2022) in the Ear and Hearing Journal.

Figure 2: *Image of Table 1 taken from Ching et al. (2022).*

Let’s create our own demographics table using a similar, but custom data set.

Data set

First we make a custom data frame which contains thirty observations of six demographic variables:

Code

data <- data.frame(
  sex = c(
    "Male", "Female", "Female",
    "Female", "Female", "Female", "Female", "Male", "Female",
    "Female", "Male", "Female", "Female", "Female", "Male",
    "Female", "Female", "Male", "Female", "Male", "Male",
    "Female", "Male", "Female", "Male", "Male", "Male", "Male",
    "Female", "Male"
  ),
    mode_hearing = c(
    "Bimodal", "Bilateral",
    "Bilateral", "Bimodal", "Bimodal", "Bilateral", "Bimodal",
    "Bilateral", "Unilateral", "Bimodal", "Unilateral",
    "Bilateral", "Unilateral", "Bilateral", "Unilateral", "Bilateral",
    "Bilateral", "Bilateral", "Bimodal", "Bimodal",
    "Bilateral", "Bimodal", "Bilateral", "Unilateral", "Bimodal",
    "Bilateral", "Bilateral", "Bimodal", "Bilateral",
    "Unilateral"
  ),
  age = c(
    71L, 52L, 47L, 44L, 68L, 57L,
    77L, 58L, 76L, 79L, 62L, 57L, 76L, 68L, 65L, 43L, 78L, 73L,
    58L, 45L, 76L, 44L, 66L, 42L, 76L, 63L, 76L, 54L, 50L,
    56L
  ),
  hearing_loss = c(
    98L, 108L, 66L, 99L, 115L, 95L,
    98L, 110L, 94L, 110L, 75L, 64L, 81L, 107L, 109L, 118L,
    75L, 79L, 119L, 67L, 73L, 73L, 86L, 97L, 72L, 82L, 91L,
    94L, 94L, 80L
  ),
  age_first_ci = c(
    63L, 42L, 36L, 41L, 57L, 51L,
    68L, 44L, 59L, 65L, 48L, 45L, 58L, 59L, 59L, 24L, 59L, 60L,
    52L, 36L, 73L, 24L, 48L, 37L, 64L, 46L, 62L, 38L, 40L,
    43L
  ),
  duration_use = c(
    8L, 10L, 11L, 3L, 11L, 6L, 9L,
    14L, 17L, 14L, 14L, 12L, 18L, 9L, 6L, 19L, 19L, 13L, 6L,
    9L, 3L, 20L, 18L, 5L, 12L, 17L, 14L, 16L, 10L, 13L
  )
)

flextable::flextable(head(data))

sex	mode_hearing	age	hearing_loss	age_first_ci	duration_use
Male	Bimodal	71	98	63	8
Female	Bilateral	52	108	42	10
Female	Bilateral	47	66	36	11
Female	Bimodal	44	99	41	3
Female	Bimodal	68	115	57	11
Female	Bilateral	57	95	51	6

Then we add nicely formatted labels to all variables in the data frame using the labelled package¹:

Code

labelled::var_label(data) <-
  list(
    sex = "Sex",
    mode_hearing = "Mode of hearing",
    age = "Age at assessment (Years)",
    hearing_loss = "Hearing Loss (4FA)",
    age_first_ci = "Age at first CI (Years)",
    duration_use = "Duration of use (Years)"
  )

flextable::flextable(var_label(data) %>%
  as.data.frame())

sex	mode_hearing	age	hearing_loss	age_first_ci	duration_use
Sex	Mode of hearing	Age at assessment (Years)	Hearing Loss (4FA)	Age at first CI (Years)	Duration of use (Years)

Basic Demographics Table

We then create the demographics table with the tbl_summary() function from the gtsummary package:

Code

# set gtsummary themes
gtsummary::reset_gtsummary_theme()
#theme_gtsummary_journal(journal = "nejm", set_theme = TRUE)

# create demographics table
basic <- data %>%
  tbl_summary(
    type = all_continuous() ~ "continuous2",
    statistic = list(
      all_continuous() ~ c(
        "{mean} ({sd})",
        "{median}",
        "{min}, {max}"
      ),
      all_categorical() ~ "{n} ({p}%)"
    ),
    digits = list(all_continuous() ~ 1),
    missing = "no",
  ) %>%
  bold_labels() %>%
  modify_header(
    label = "",
    all_stat_cols() ~ "**{level}**, N = {n} ({style_percent(p)}%)"
  ) %>%
  gtsummary::as_flex_table()

basic %>% 
  flextable::autofit() %>% 
  flextable::width(width = 3) %>% 
  flextable::line_spacing(space = 0.35, part = "body")

Table 1: Participant Demographics

	Overall, N = 30 (100%)1
Sex
Female	17 (57%)
Male	13 (43%)
Mode of hearing
Bilateral	14 (47%)
Bimodal	10 (33%)
Unilateral	6 (20%)
Age at assessment (Years)
Mean (SD)	61.9 (12.3)
Median	62.5
Range	42.0, 79.0
Hearing Loss (4FA)
Mean (SD)	91.0 (16.4)
Median	94.0
Range	64.0, 119.0
Age at first CI (Years)
Mean (SD)	50.0 (12.5)
Median	49.5
Range	24.0, 73.0
Duration of use (Years)
Mean (SD)	11.9 (4.9)
Median	12.0
Range	3.0, 20.0
1n (%)

And there we have a demographics table that is ready to publish!

Advanced Demographics Table

Sometimes, we we need to display demographic variables across a specific study group, such as comparing participant characteristics based on different modes of hearing.

With a few adjustments to the tbl_summary() function, we can make a new demographics table that allows us to quickly observe the differences in participant characteristics across the various hearing modes:

Code

# create demographics table with grouping variable
gtsummary::theme_gtsummary_compact(set_theme = TRUE, font_size = 10)

adv <- data %>%
  tbl_summary(
    by = mode_hearing,
    type = all_continuous() ~ "continuous2",
    statistic = list(
      all_continuous() ~ c(
        "{mean} ({sd})",
        "{median}",
        "{min}, {max}"
      ),
      all_categorical() ~ "{n} ({p}%)"
    ),
    digits = list(all_continuous() ~ 1),
    missing = "no",
  ) %>%
  add_overall() %>%
  add_p() %>%
  bold_labels() %>%
  modify_header(
    label = "",
    all_stat_cols() ~ "**{level}**, N = {n} ({style_percent(p)}%)"
  ) %>%
  gtsummary::as_flex_table()

adv %>% 
  flextable::line_spacing(space = 1.1, part = "body") %>% 
  flextable::autofit()

Table 2: Participant Demographics by Mode of Hearing

	Overall, N = 30 (100%)1	Bilateral, N = 14 (47%)1	Bimodal, N = 10 (33%)1	Unilateral, N = 6 (20%)1	p-value2
Sex					>0.9
Female	17 (57%)	8 (57%)	6 (60%)	3 (50%)
Male	13 (43%)	6 (43%)	4 (40%)	3 (50%)
Age at assessment (Years)					>0.9
Mean (SD)	61.9 (12.3)	61.7 (11.5)	61.6 (14.3)	62.8 (12.9)
Median	62.5	60.5	63.0	63.5
Range	42.0, 79.0	43.0, 78.0	44.0, 79.0	42.0, 76.0
Hearing Loss (4FA)					0.6
Mean (SD)	91.0 (16.4)	89.1 (17.1)	94.5 (18.3)	89.3 (12.9)
Median	94.0	88.5	98.0	87.5
Range	64.0, 119.0	64.0, 118.0	67.0, 119.0	75.0, 109.0
Age at first CI (Years)					>0.9
Mean (SD)	50.0 (12.5)	49.2 (12.5)	50.8 (15.1)	50.7 (9.4)
Median	49.5	47.0	54.5	53.0
Range	24.0, 73.0	24.0, 73.0	24.0, 68.0	37.0, 59.0
Duration of use (Years)					0.6
Mean (SD)	11.9 (4.9)	12.5 (4.8)	10.8 (5.0)	12.2 (5.5)
Median	12.0	12.5	10.0	13.5
Range	3.0, 20.0	3.0, 19.0	3.0, 20.0	5.0, 18.0
1n (%)
2Fisher's exact test; Kruskal-Wallis rank sum test

Exporting tables to Word

Typically, a demographics table will need to be inserted into a Microsoft Word document where it can join the rest of a manuscript or report.

The gtsummary package - in combination with the flextable and officer packages - makes exporting demographics tables to Microsoft Word very easy.

Let’s export our demographics tables to Microsoft word using flextable and officer packages:

Code

flextable::save_as_docx(adv,
  path = "articles/demographics_table_r/demo_table_adv.docx",
  pr_section =
    officer::prop_section(
      page_size = officer::page_size(
        orient = "landscape",
        width = 8.3, height = 11.7
      ),
      type = "continuous",
      page_margins = officer::page_mar()
    )
)

*Screenshot of the advanced demographics table exported to Microsoft Word.*

Once exported, the table can be edited just like any other Microsoft Word table.

Final Thoughts

When publishing study results, creating comprehensive and well-formatted demographics tables is essential, but it can also be time-consuming.
The gtsummary package in R drastically simplifies the process of generating publication-ready demographics tables.
With the help of flextable and officer packages, the gtsummary tables can be easily exported to Microsoft Word.

References

Ching, Teresa Y. C., Harvey Dillon, Sanna Hou, Mark Seeto, Ana Sodan, and Nicky Chong-White. 2022. “Development and Evaluation of a Language-Independent Test of Auditory Discrimination for Referrals for Cochlear Implant Candidacy Assessment.” Ear and Hearing 43 (4): 1151. https://doi.org/10.1097/AUD.0000000000001166.

Gohel, David. 2023. “Officer: Manipulation of Microsoft Word and PowerPoint Documents.” https://CRAN.R-project.org/package=officer.

Gohel, David, and Panagiotis Skintzos. 2023. “Flextable: Functions for Tabular Reporting.” https://CRAN.R-project.org/package=flextable.

Larmarange, Joseph. 2022. “Labelled: Manipulating Labelled Data.” https://CRAN.R-project.org/package=labelled.

Sjoberg, Daniel D., Karissa Whiting, Michael Curry, Jessica A. Lavery, and Joseph Larmarange. 2021. “Reproducible Summary Tables with the Gtsummary Package” 13: 570–80. https://doi.org/10.32614/RJ-2021-053.

Wickham, Hadley, Mara Averick, Jennifer Bryan, Winston Chang, Lucy D’Agostino McGowan, Romain François, Garrett Grolemund, et al. 2019. “Welcome to the Tidyverse” 4: 1686. https://doi.org/10.21105/joss.01686.

Footnotes

When creating the demographics table, these formatted labels will be shown in the table instead of the actual variable names. This is an important step which will save you a lot of effort renaming individual variables within the table.↩︎