Risk of Bias Plots

mars authors

2026-04-02

This vignette shows how to create two base-R risk-of-bias displays with risk_of_bias_plot():

1. A study-level traffic-light panel
2. A component-level stacked percentage bar panel (Low/Unclear/High, summing to 100)

library(mars)

Required Input Structure

The function expects long-format data with one row per study x component judgment:

• study: study identifier
• component: risk-of-bias domain/component name
• decision: judgment (Low, Unclear, High)

Common case/spelling variants are accepted, including low risk, high risk, some concerns, and uncler.

rob_data <- data.frame(
  study = rep(paste0("Study_", 1:6), each = 5),
  component = rep(
    c("Randomization", "Allocation Concealment", "Blinding", "Incomplete Data", "Reporting"),
    times = 6
  ),
  decision = c(
    "Low", "Low", "Unclear", "High", "Low",
    "Low", "Unclear", "Unclear", "High", "Low",
    "High", "High", "Unclear", "Low", "Low",
    "Low", "Low", "Low", "Low", "Low",
    "Uncler", "Low", "High", "Unclear", "Low",
    "Some concerns", "Low risk", "High risk", "Unclear", "Low"
  ),
  stringsAsFactors = FALSE
)

head(rob_data, 10)
#>      study              component decision
#> 1  Study_1          Randomization      Low
#> 2  Study_1 Allocation Concealment      Low
#> 3  Study_1               Blinding  Unclear
#> 4  Study_1        Incomplete Data     High
#> 5  Study_1              Reporting      Low
#> 6  Study_2          Randomization      Low
#> 7  Study_2 Allocation Concealment  Unclear
#> 8  Study_2               Blinding  Unclear
#> 9  Study_2        Incomplete Data     High
#> 10 Study_2              Reporting      Low

Default Plot

out <- risk_of_bias_plot(
  data = rob_data,
  study = "study",
  component = "component",
  decision = "decision"
)

Custom Ordering and Colors

You can control row/column order and traffic-light colors.

risk_of_bias_plot(
  data = rob_data,
  study = "study",
  component = "component",
  decision = "decision",
  study_order = paste0("Study_", 6:1),
  component_order = c(
    "Randomization",
    "Allocation Concealment",
    "Blinding",
    "Incomplete Data",
    "Reporting"
  ),
  low_col = "#3C8D2F",
  unclear_col = "#E6B91E",
  high_col = "#B4362D",
  main_traffic = "Study-Level Risk of Bias",
  main_aggregate = "Component Summary (%)"
)

Separate Plots and Horizontal Aggregate Bars

You can request one panel at a time, and draw the aggregate panel horizontally.

risk_of_bias_plot(
  data = rob_data,
  study = "study",
  component = "component",
  decision = "decision",
  plot = "study"
)

risk_of_bias_plot(
  data = rob_data,
  study = "study",
  component = "component",
  decision = "decision",
  plot = "aggregate",
  aggregate_horizontal = TRUE
)

Returned Values

The function returns useful derived objects invisibly:

names(out)
#> [1] "decision_matrix" "counts"          "percentages"
out$counts
#>                        Low Unclear High
#> Randomization            3       2    1
#> Allocation Concealment   4       1    1
#> Blinding                 1       3    2
#> Incomplete Data          2       2    2
#> Reporting                6       0    0
round(out$percentages, 1)
#>                          Low Unclear High
#> Randomization           50.0    33.3 16.7
#> Allocation Concealment  66.7    16.7 16.7
#> Blinding                16.7    50.0 33.3
#> Incomplete Data         33.3    33.3 33.3
#> Reporting              100.0     0.0  0.0