Meta-analysis flowchart

Question

Is it possible to reproduce a meta-analysis type of flowchart as the one in the picture below using any R tool?

My attempt was using mermaid:

diagram = "
graph LR
  subgraph Screening
    b1-->b2
end
  subgraph Eligibility
    c1-->c2
  end
  subgraph Included
    d1-->d2
  end
  subgraph Identification
    a1-->a2
  end



"
mermaid(diagram)

Which generated:

But I cannot find a way of connect the nodes accross the subgraphs.

Is there another tool better fitting to this kind of job? I am thinking on any package that I could use from within my Rmarkdown document.

Pete · Accepted Answer · 2018-12-04T19:27:54.650

I have found the DiagrammeR package easiest to do this. The general idea would be something like:

library(glue)
library(DiagrammeR)

excluded <- glue('Full text articles excluded
             n = 1000
             Reasons for exclusion
             Reason 1
             Reason 2')
grViz("
digraph cohort_flow_chart
{
node [fontname = Helvetica, fontsize = 12, shape = box, width = 4]
a[label = 'Records identified in original search']
b[label = 'Records identified with update']
c[label = 'Records after duplicates removed']
d[label = 'Records screened']
e[label = 'Records excluded']
f[label = 'Full text articles assessed']
g[label = 'Studies included']
h[label = '@@1']



{ rank = same; a b}
{ rank = same; d, e}
{ rank = same; f, h}

a -> c;
b -> c;
c -> d;
d -> e [ minlen = 3 ];
d -> f;
f -> h [ minlen = 3 ];
f -> g;
}

[1]: excluded
")

Will look like:

Image with labels and empty nodes

grViz("
digraph cohort_flow_chart
{
node [fontname = Helvetica, fontsize = 12, shape = box, width = 4]
i[label = 'Identification', fillcolor = LightBlue, 
  style = filled,     width = 2]
j[label = 'Screening',fillcolor = LightBlue, style = filled, width = 2]
k[label = 'Eligibility', fillcolor = LightBlue, style = filled, 
  width = 2]
l[label = 'Included', fillcolor = LightBlue, style = filled, width = 2]

a[label = 'Records identified in original search']
b[label = 'Records identified with update']
c[label = 'Records after duplicates removed']
d[label = 'Records screened']
e[label = 'Records excluded']
f[label = 'Full text articles assessed']
g[label = 'Studies included']
h[label = '@@1']
blank_1[label = '', width = 0.01, height = 0.01]
blank_2[label = '', width = 0.01, height = 0.01]
blank_4[label = '', width = 4, color = White]

{ rank = same; a b i}
{ rank = same; blank_4 c j}
{ rank = same; f k}
{ rank = same; g l}
{ rank = same; blank_1 e}
{ rank = same; blank_2 h}

a -> c;
b -> c;
b -> blank_4 [ dir = none, color = White];
c -> d;
d -> blank_1 [ dir = none ];
blank_1 -> e [ minlen = 3 ];
blank_1 -> f;
f -> blank_2 [ dir = none ];
blank_2 -> h [ minlen = 3 ];
blank_2 -> g;

}

[1]: excluded
")

Thank you very much, this answers. Is it possible to add vertical labels to mark the distinct phases from identification to inclusion? — lf_araujo, Dec 04 '18 at 18:31
You could certainly add in labels, but not sure if you could get them vertical with the grViz function. — Pete, Dec 04 '18 at 19:00
Edited answer to include horizontal labels. Also can move the excluded boxes if wanted. Not sure there is a simple way with grViz to better allign the 'Records identified through update' box. — Pete, Dec 04 '18 at 19:29

Meta-analysis flowchart

1 Answers1