1. Sankey diagram: 入院方式 → 急救責任分級 → 轉診型態
### 使用資料: EMOC_v7.csv
df = read.csv('EMOC_v7.csv', na = c('', 'NA'), stringsAsFactors = FALSE) # 21781 examples
### Columns to use
left_col = '入院方式'
cen_col = 'EMOC_轉出醫院急救責任醫院等級'
right_col = '轉診型態'
### Set the labels
unique_labs = do.call('c', sapply(c(left_col, cen_col, right_col), function(x) unique(df[, x])))
unique_labs_ref = seq(1:length(unique_labs)) - 1
names(unique_labs_ref) = unique_labs
### Set the names of the starting point,the ending point and the values of links
starts = c()
ends = c()
links = c()
## The left part
df_left = aggregate(as.formula(paste(right_col, '~', paste(left_col, cen_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_left[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_left[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_left[, 3])
## The righgt part
df_right = aggregate(as.formula(paste(left_col, '~', paste(cen_col, right_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_right[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_right[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_right[, 3])
### Set the colors
color_left_ref = c('rgba(213, 185, 0, 1)', 'rgba(135, 0, 0, 1)', 'rgba(0, 0, 0, 1)', 'rgba(17, 0, 140, 1)', 'rgba(0, 88, 191, 1)')
names(color_left_ref) = unique(df[, left_col])
color_cen_ref = c('rgba(53, 68, 73, 1)', 'rgba(235, 189, 135, 1)', 'rgba(57, 49, 47, 1)')
names(color_cen_ref) = unique(df[, cen_col])
color_right_ref = c('rgba(220, 20, 60, 1)', 'rgba(34, 139, 34, 1)', 'rgba(65, 105, 225, 1)')
names(color_right_ref) = unique(df[, right_col])
color_left = as.vector(color_left_ref[df_left[, 1]])
color_right = as.vector(color_cen_ref[df_right[, 1]])
color_node = as.vector(c(color_left_ref, color_cen_ref, color_right_ref)[unique_labs])
alpha_set = 0.3
### Plotting
library(plotly)
p <- plot_ly(
type = "sankey",
domain = c(
x = c(0,1),
y = c(0,1)
),
orientation = "h",
valueformat = ".0f",
valuesuffix = " 位",
node = list(
label = unique_labs,
color = color_node,
pad = 15,
thickness = 15,
line = list(
color = color_node,
width = 0.1
)
),
link = list(
source = starts,
target = ends,
value = links,
color = c( gsub('1)', paste(alpha_set, ')', sep = ''), color_left), gsub('1)', paste(alpha_set, ')', sep = ''), color_right))
)
) %>%
layout(
title = "入院方式 → 急救責任分級 → 轉診型態",
autosize = F, width = 1200, height = 600,
font = list(
size = 12
),
xaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE),
yaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE)
)
### Output
htmlwidgets::saveWidget(as_widget(p), "Sankey_v1.html")
p
2. Sankey diagram: 入院方式 → 急救責任分級 → 轉院原因 → 轉診型態
### Use the same data as the previous plot
### Columns to use
left_col = '入院方式'
cen_col = 'EMOC_轉出醫院急救責任醫院等級'
cen_2_col = 'EMOC轉院原因大類別'
right_col = '轉診型態'
### Set the labels
unique_labs = do.call('c', sapply(c(left_col, cen_col, cen_2_col, right_col), function(x) unique(df[, x])))
unique_labs_ref = seq(1:length(unique_labs)) - 1
names(unique_labs_ref) = unique_labs
### Set the names of the starting point, ending point and the values of links
starts = c()
ends = c()
links = c()
### The left part
df_left = aggregate(as.formula(paste(right_col, '~', paste(left_col, cen_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_left[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_left[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_left[, 3])
### The niddle part
df_cen = aggregate(as.formula(paste(left_col, '~', paste(cen_col, cen_2_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_cen[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_cen[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_cen[, 3])
### The righgt part
df_right = aggregate(as.formula(paste(left_col, '~', paste(cen_2_col, right_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_right[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_right[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_right[, 3])
### Set the colors
color_left_ref = c('rgba(213, 185, 0, 1)', 'rgba(135, 0, 0, 1)', 'rgba(0, 0, 0, 1)', 'rgba(17, 0, 140, 1)', 'rgba(0, 88, 191, 1)')
names(color_left_ref) = unique(df[, left_col])
color_cen_ref = c('rgba(53, 68, 73, 1)', 'rgba(235, 189, 135, 1)', 'rgba(57, 49, 47, 1)')
names(color_cen_ref) = unique(df[, cen_col])
color_cen_2_ref = c('rgba(35, 49, 73, 1)', 'rgba(78, 114, 39, 1)', 'rgba(57, 55, 51, 1)')
names(color_cen_2_ref) = unique(df[, cen_2_col])
color_right_ref = c('rgba(220, 20, 60, 1)', 'rgba(34, 139, 34, 1)', 'rgba(65, 105, 225, 1)')
names(color_right_ref) = unique(df[, right_col])
color_left = as.vector(color_left_ref[df_left[, 1]])
color_cen = as.vector(color_cen_ref[df_cen[, 1]])
color_right = as.vector(color_cen_2_ref[df_right[, 1]])
color_node = as.vector(c(color_left_ref, color_cen_ref, color_cen_2_ref, color_right_ref)[unique_labs])
alpha_set = 0.3
### Plotting
library(plotly)
p <- plot_ly(
type = "sankey",
domain = c(
x = c(0,1),
y = c(0,1)
),
orientation = "h",
valueformat = ".0f",
valuesuffix = " 人",
node = list(
label = unique_labs,
color = color_node,
pad = 15,
thickness = 15,
line = list(
color = color_node,
width = 0.1
)
),
link = list(
source = starts,
target = ends,
value = links,
color = c( gsub('1)', paste(alpha_set, ')', sep = ''), color_left),
gsub('1)', paste(alpha_set, ')', sep = ''), color_cen),
gsub('1)', paste(alpha_set, ')', sep = ''), color_right))
)
) %>%
layout(
title = "入院方式 → 急救責任分級 → 轉院原因 → 轉診型態",
autosize = F, width = 1200, height = 600,
font = list(
size = 12
),
xaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE),
yaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE)
)
### Output
htmlwidgets::saveWidget(as_widget(p), "Sankey_v2.html")
p
---
title: "急診轉診流動分析 Sankey Diagram"
output: html_notebook
---

<br>
<br>

###1. Sankey diagram: 入院方式 → 急救責任分級 → 轉診型態

```{r, echo=TRUE, message=TRUE, warning=FALSE}

### 使用資料: EMOC_v7.csv
df = read.csv('EMOC_v7.csv', na = c('', 'NA'), stringsAsFactors = FALSE) # 21781 examples

### Columns to use
left_col = '入院方式'
cen_col = 'EMOC_轉出醫院急救責任醫院等級'
right_col = '轉診型態'

### Set the labels
unique_labs = do.call('c', sapply(c(left_col, cen_col, right_col), function(x) unique(df[, x])))
unique_labs_ref = seq(1:length(unique_labs)) - 1
names(unique_labs_ref) = unique_labs

### Set the names of the starting point,the ending point and the values of links
starts = c()
ends = c()
links = c()

## The left part
df_left = aggregate(as.formula(paste(right_col, '~', paste(left_col, cen_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_left[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_left[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_left[, 3])

## The righgt part
df_right = aggregate(as.formula(paste(left_col, '~', paste(cen_col, right_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_right[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_right[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_right[, 3])

### Set the colors
color_left_ref = c('rgba(213, 185, 0, 1)', 'rgba(135, 0, 0, 1)', 'rgba(0, 0, 0, 1)', 'rgba(17, 0, 140, 1)', 'rgba(0, 88, 191, 1)')
names(color_left_ref) = unique(df[, left_col])
color_cen_ref = c('rgba(53, 68, 73, 1)', 'rgba(235, 189, 135, 1)', 'rgba(57, 49, 47, 1)')
names(color_cen_ref) = unique(df[, cen_col])
color_right_ref = c('rgba(220, 20, 60, 1)', 'rgba(34, 139, 34, 1)', 'rgba(65, 105, 225, 1)')
names(color_right_ref) = unique(df[, right_col])

color_left = as.vector(color_left_ref[df_left[, 1]])
color_right = as.vector(color_cen_ref[df_right[, 1]])
color_node = as.vector(c(color_left_ref, color_cen_ref, color_right_ref)[unique_labs])

alpha_set = 0.3

### Plotting
library(plotly)

p <- plot_ly(
  type = "sankey",
  domain = c(
    x =  c(0,1),
    y =  c(0,1)
  ),
  orientation = "h",
  valueformat = ".0f",
  valuesuffix = " 人",
  
  node = list(
    label = unique_labs,
    color = color_node,
    pad = 15,
    thickness = 15,
    line = list(
      color = color_node,
      width = 0.1
    )
  ),
  
  link = list(
    source = starts,
    target = ends,
    value =  links,
    color = c( gsub('1)', paste(alpha_set, ')', sep = ''), color_left),  gsub('1)', paste(alpha_set, ')', sep = ''), color_right))
  )
) %>% 
  layout(
    title = "入院方式 → 急救責任分級 → 轉診型態",
    autosize = F, width = 1200, height = 600,
    font = list(
      size = 12
    ),
    xaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE),
    yaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE)
  )

### Output
htmlwidgets::saveWidget(as_widget(p), "Sankey_v1.html")
p
```
<br>
<br>
<br>
<br>
<br>
<br>
<br>

###2. Sankey diagram: 入院方式 → 急救責任分級 → 轉院原因 → 轉診型態

```{r, echo=TRUE, message=TRUE, warning=FALSE}

### Use the same data as the previous plot

### Columns to use
left_col = '入院方式'
cen_col = 'EMOC_轉出醫院急救責任醫院等級'
cen_2_col = 'EMOC轉院原因大類別'
right_col = '轉診型態'

### Set the labels
unique_labs = do.call('c', sapply(c(left_col, cen_col, cen_2_col, right_col), function(x) unique(df[, x])))
unique_labs_ref = seq(1:length(unique_labs)) - 1
names(unique_labs_ref) = unique_labs

### Set the names of the starting point, ending point and the values of links
starts = c()
ends = c()
links = c()

### The left part
df_left = aggregate(as.formula(paste(right_col, '~', paste(left_col, cen_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_left[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_left[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_left[, 3])

### The niddle part
df_cen = aggregate(as.formula(paste(left_col, '~', paste(cen_col, cen_2_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_cen[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_cen[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_cen[, 3])

### The righgt part
df_right = aggregate(as.formula(paste(left_col, '~', paste(cen_2_col, right_col, sep = '+'))), df, function(x) length(x))
starts = c(starts, as.vector(sapply(as.vector(df_right[, 1]), function(x) unique_labs_ref[x])))
ends = c(ends, as.vector(sapply(as.vector(df_right[, 2]), function(x) unique_labs_ref[x])))
links = c(links, df_right[, 3])

### Set the colors
color_left_ref = c('rgba(213, 185, 0, 1)', 'rgba(135, 0, 0, 1)', 'rgba(0, 0, 0, 1)', 'rgba(17, 0, 140, 1)', 'rgba(0, 88, 191, 1)')
names(color_left_ref) = unique(df[, left_col])
color_cen_ref = c('rgba(53, 68, 73, 1)', 'rgba(235, 189, 135, 1)', 'rgba(57, 49, 47, 1)')
names(color_cen_ref) = unique(df[, cen_col])
color_cen_2_ref = c('rgba(35, 49, 73, 1)', 'rgba(78, 114, 39, 1)', 'rgba(57, 55, 51, 1)')
names(color_cen_2_ref) = unique(df[, cen_2_col])
color_right_ref = c('rgba(220, 20, 60, 1)', 'rgba(34, 139, 34, 1)', 'rgba(65, 105, 225, 1)')
names(color_right_ref) = unique(df[, right_col])

color_left = as.vector(color_left_ref[df_left[, 1]])
color_cen = as.vector(color_cen_ref[df_cen[, 1]])
color_right = as.vector(color_cen_2_ref[df_right[, 1]])
color_node = as.vector(c(color_left_ref, color_cen_ref, color_cen_2_ref, color_right_ref)[unique_labs])

alpha_set = 0.3

### Plotting
library(plotly)

p <- plot_ly(
  type = "sankey",
  domain = c(
    x =  c(0,1),
    y =  c(0,1)
  ),
  orientation = "h",
  valueformat = ".0f",
  valuesuffix = " 人",
  
  node = list(
    label = unique_labs,
    color = color_node,
    pad = 15,
    thickness = 15,
    line = list(
      color = color_node,
      width = 0.1
    )
  ),
  
  link = list(
    source = starts,
    target = ends,
    value =  links,
    color = c( gsub('1)', paste(alpha_set, ')', sep = ''), color_left),
               gsub('1)', paste(alpha_set, ')', sep = ''), color_cen),
               gsub('1)', paste(alpha_set, ')', sep = ''), color_right))
  )
) %>% 
  layout(
    title = "入院方式 → 急救責任分級 → 轉院原因 → 轉診型態",
    autosize = F, width = 1200, height = 600,
    font = list(
      size = 12
    ),
    xaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE),
    yaxis = list(showgrid = F, zeroline = F, showticklabels = FALSE)
  )

### Output
htmlwidgets::saveWidget(as_widget(p), "Sankey_v2.html")
p
```
<br>
<br>
