Introduction to robvis, a visualization tool for risk-of-bias assessments

Luke A McGuinness

2024-04-25

Source: vignettes/Introduction_to_robvis.Rmd

Introduction

The robvis package provides functions to convert a risk-of-bias assessment summary table into a summary plot or a traffic-light plot, formatted based on the specific risk-of-bias assessment tool used.

robvis currently contains templates for the following tools:

  • ROB2
  • ROBINS-I
  • QUADAS-2
  • ROB1

Users can find the exact assessment tool name expected by the tool argument of the rob_summary() and rob_traffic_light() functions by running:

rob_tools()
#> Note: the "ROB2-Cluster" template is only available for the rob_traffic_light() function.
#> [1] "ROB2"         "ROB2-Cluster" "ROBINS-I"     "ROBINS-E"     "QUADAS-2"    
#> [6] "QUIPS"        "Generic"

Loading your data

robvis expects certain facts about the data you provide it.

  1. The first column contains the study identifier
  2. The last column will contain the overall risk-of-bias judgments
  3. The first row of the data does not contain column headings. This can be achieved using the header = TRUE option (which indicates that the first line contains column headings) when reading in your summary table:
data <- read.csv("path/to/summary_table.csv", header = TRUE)

All other columns are expected to contain the results of the risk-of bias assessments for a specific domain. To elaborate, consider as an example the ROB2.0 tool which has 5 domains. The resulting data set that robvis would expect for this tool would have 7 columns:

  • Column 1: Study identifier
  • Column 2-6: One RoB2 domain per column
  • Column 7: Overall risk-of-bias judgments

The only exception to this is the "Generic" template, which is discussed below.

Example data sets

To help users explore robvis, we have included an example data set for each tool template that exists in the package. For example, the data_rob2 data set, which contains example risk-of-bias assessments performed using the RoB2.0 tool for randomized controlled trials, is presented below:

Study D1 D2 D3 D4 D5 Overall
Study 1 Low Low Low Low Low Low
Study 2 Some concerns Low Low Low Low Some concerns
Study 3 No information Low Some concerns Low Low Some concerns
Study 4 Low Low High Low Some concerns High
Study 5 High High Low Low Some concerns High
Study 6 Low High Some concerns Low Low High
Study 7 Low Some concerns Some concerns High Low High
Study 8 Low Some concerns Low Low Low Some concerns
Study 9 Low Low High Low Low High


Summary plots (rob_summary())

This function returns a ggplot object displaying a weighted bar-chart of the distribution of risk-of-bias judgments across the domains of the specified tool.

Examples:

RoB2.0 tool for randomized controlled trials 

rob_summary(data_rob2, tool = "ROB2")

ROBINS-I tool for non-randomized studies of interventions 

rob_summary(data_robins_i, tool = "ROBINS-I")

QUADAS-2 tool for diagnostic test accuracy studies 

rob_summary(data_quadas, tool = "QUADAS-2")

rob_summary() options

Overall risk-of-bias judgments (overall)

By default, a bar representing the overall risk-of-bias judgments is not included in the plot. If you would like to include this, set overall = TRUE. For example:

rob_summary(data_rob2, tool = "ROB2", overall = TRUE)

Weighted or un-weighted bar plots (weighted)

By default, the barplot is un-weighted, in that it shows the number of studies at each level of risk of bias. However, if you provide a risk-of-bias dataset that has an additional “Weights” column containing some measure of study precision, you can produce a barplot showing the amount of statistical information at each level of bias:

data_rob2_weighted <- data_rob2

# Assign a random number between 1 and 10 as the weight for each study
data_rob2_weighted$Weights <- sample(1:10,9)

# Produce a weighted barplot
rob_summary(data_rob2_weighted, tool = "ROB2", weighted = TRUE)

See the “Working with metafor” article for more information on how to extract the weights assigned to each study during a meta-analysis performed with metafor

Colour scheme (colour)

NB: Please note the non-US English spelling of colour

The colour argument of both plotting functions allows users to select from two predefined colour schemes, “cochrane” (default) or “colourblind”, or to define their own palette by providing a vector of hex codes.

For example, to use the predefined “colourblind” palette:

rob_summary(data = data_rob2, tool = "ROB2", colour = "colourblind")

And to define your own colour scheme:

rob_summary(data = data_rob2, tool = "ROB2", colour = c("#f442c8","#bef441","#000000","#557925"))

When defining your own colour scheme, you must ensure that the number of discrete judgments (e.g. “Low”/“Moderate”/“High”/“Critical”) and the number of colours specified are the same. Additionally, colours must be specified in order of ascending risk-of-bias (e.g. “Low” -> “Critical”), with the first hex corresponding to “Low” risk of bias.



Traffic light plots (rob_traffic_light())

This function returns a ggplot object displaying the risk-of-bias judgment in each domain for each study, as well as the overall risk-of-bias judgement for that study.

Examples:

RoB2.0 tool for randomized controlled trials 

rob_traffic_light(data_rob2, tool = "ROB2")

ROBINS-I tool for non-randomized studies of interventions 

rob_traffic_light(data_robins_i, tool = "ROBINS-I")

QUADAS-2 tool for diagnostic test accuracy studies 

rob_traffic_light(data_quadas, tool = "QUADAS-2")

rob_traffic_light() options

Point size (psize)

By default, the size of each point is set to 20. However, if you have a large number of studies, it is useful to be able to reduce the point size so that the resulting graphic is not too large.

# Generate larger dataset
data <- rbind(data_rob2, data_rob2)
data$Study <- paste("Study",seq(1,18))

# Plot with reduced point size
rob_traffic_light(data, tool = "ROB2", psize = 10)

Colour scheme (colour)

NB: Please note the non-US English spelling of colour

The colour argument of both plotting functions allows users to select from two predefined colour schemes, “cochrane” (default) or “colourblind”, or to define their own palette by providing a vector of hex codes.

For example, to use the predefined “colourblind” palette:

rob_traffic_light(data = data_rob2, tool = "ROB2", colour = "colourblind")

And to define your own colour scheme:

rob_traffic_light(data = data_rob2, tool = "ROB2", colour = c("#f442c8","#bef441","#000000","#557925"))

When defining your own colour scheme, you must ensure that the number of discrete judgments (e.g. “Low”/“Moderate”/“High”/“Critical”) and the number of colours specified are the same. Additionally, colours must be specified in order of ascending risk-of-bias (e.g. “Low” -> “Critical”), with the first hex corresponding to “Low” risk of bias.



The “Generic” template

Motivation

This template offers increased flexibility in the domains that are included in the plot. It can handle any number of user defined domains and uses the user defined column headings as domains titles in the resulting figures.

Varying numbers of domains

The "Generic" option can handle varying numbers of columns, as authors using the ROB1 assessment tool frequently add or remove bias domains within this tool. However, we would discourage authors from doing so with any tool other than ROB1. Authors using other published tools (ROB2, QUADAS-2, ROBINS-I) should use the stricter templates presented above to ensure they conform with the guidance.

Domain names

For the other tools listed above, the names of the columns containing the domain-level risk of bias judgments are not important. However, this is not the case when using the "Generic" template.

Compare for example, the first row of the data_rob2 and the data_rob1, and the resulting figures.

colnames(data_rob2)
#> [1] "Study"   "D1"      "D2"      "D3"      "D4"      "D5"      "Overall"

colnames(data_rob1)
#> [1] "Study"                                 
#> [2] "Random.sequence.generation"            
#> [3] "Allocation.concealment"                
#> [4] "Blinding.of.participants.and.personnel"
#> [5] "Blinding.of.outcome.assessment"        
#> [6] "Incomplete.outcome.data"               
#> [7] "Selective.reporting"                   
#> [8] "Other.sources.of.bias"                 
#> [9] "Overall"

The domain columns (Columns 2-6) in the ROB2 example have been given arbitrary names of D1-D5, as they will be overwritten by the function to correspond to the correct domain titles as per the ROB2 guidance.

In contrast, the domain columns (Columns 2-8) in the ROB1 example use their true title as the column heading, as these will be used in the figures produced by rob_summary() and rob_traffic_light(). As an example, suppose we change the name of the “Random.sequence.generation” column to something else. In the rob_summary() figure, the title of the first bar is changed, while in the rob_traffic_light() figure, the caption is updated to reflect this change.