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

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()
#> [1] "ROB2"
#> [1] "ROBINS-I"
#> [1] "QUADAS-2"
#> [1] "ROB1"

Loading your data

robvis expects certain facts about the data you provide it.

1. The first column contains the study identifier
2. The second-to-last column will contain the overall risk-of-bias judgments
3. The last column will contain the weights, which can all be set to 1 if the relevant weights are not available.
4. 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 8 columns:

• Column 1: Study identifier
• Column 2-6: One RoB2 domain per column
• Column 7: Overall risk-of-bias judgments
• Column 8: Weights

The only exception to this is the "ROB1" 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	Weight
Study 1	Low	Low	Low	Low	Low	Low	33.3333333
Study 2	Some concerns	Low	Low	Low	Low	Low	33.3333333
Study 3	Some concerns	Low	Some concerns	Low	Low	Some concerns	0.1428571
Study 4	Low	Low	High	Low	Some concerns	High	9.0909091
Study 5	High	High	Low	Low	Some concerns	Low	12.5000000
Study 6	Low	High	Some concerns	Low	Low	Some concerns	25.0000000
Study 7	Low	Some concerns	Some concerns	High	Low	Some concerns	200.0000000
Study 8	Low	Some concerns	Low	Low	Low	Low	11.1111111
Study 9	Low	Low	High	Low	Low	High	1.1111111

---

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, 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 weighted by some measure of study precision (see Example data sets). You can turn off this option by setting weighted = FALSE. For example, compare this plot with that produced by the based rob_summary() function using the data_rob2 data set.

rob_summary(data_rob2, tool = "ROB2", weighted = FALSE)

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"))

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, 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"))

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 “ROB1” 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 "ROB1" 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 "ROB1" 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"
#> [8] "Weight"

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"                                
#> [10] "Weight"

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.