Data Analytics Portfolio

Causes of Death in Africa Dashboard Hosted on R Shiny

Introduction

This section involves analysing data from a publicly available dataset and building a dashboard on R shiny with the data. The "Causes of Death in Africa from 1990 - 2019" dashboard is designed to provide an interactive visualization of the causes of death over the specified period.

Motivation

The dashboard tries to provide an interactive platform for exploring causes of death across African countries, from 1990-2019. The dashboard allows users to understand causes of death in more detail by exploring specific causes of death in any African country over chosen periods.

The dashboard serves users with different backgrounds; individuals involved in planning and implementing health policies can use the dashboard to identify critical areas of concern and to monitor the impact of health interventions over time. Also, professionals engaged in health research can explore the data for trends and insights that may inform academic studies, or further investigations into specific causes of death. In addition, NGOs focused on improving health outcomes in Africa can use this dashboard to target their efforts more effectively, understanding where their resources and programs might have the most significant impact.

Data and Transformation

Data used in this analysis was gotten from two sources:

Causes of death- https://ourworldindata.org/causes-of-death

List of countries - https://statisticstimes.com/geography/countries-by-continents.php

After installing and loading the packages, the initial dataset, “annual-number-of-deaths-by-cause.csv”, was read into the R environment using the read.csv() function. This dataset contains annual deaths segmented by different causes of death all countries in the world over a range of years. I used the head() function to look at the first few rows of the dataset, providing insight into its structure, column names, and the nature of the data contained within. I used gsub() function to simplify the column names to extract only the names of the disease to enhance readability, specifically "Deaths...|...Sex..Both...Age..All.Ages..Number." in the column names were removed.

The next thing I did was to transformed the dataset from a wide format to a long format using the pivot_longer() function from the tidyverse package. This restructuring process consolidates the various causes of death into a single column (cause_of_death), with the corresponding death counts placed in another column (death_count). Then I checked for null values in each column using colSums(is.na()) and there was none.

The next step was to scrap the web for Additional data pertaining to African countries, their ISO-alpha3 codes, and respective regions using the rvest package. I used read_html(), html_table(), and filter() to extract and filter the relevant table from the webpage, focusing specifically on African countries. Then I merged the death dataset and the supplementary country data using the inner_join() function. I then used the select() function to exclude redundant columns, ensuring a clean, integrated dataset.

The final step in data preparation involved renaming the "Entity" column to "Country" to more accurately reflect its contents. This is accomplished using the rename() function, resulting in the final dataset, african_death_causes, which was used for analysis and visualization in the Shiny app.

Click to see the R Shiny Dashboard

R Code for Building the R shiny Dashboard

#install and load package

install.packages("tidyverse")

install.packages("shiny")

install.packages("shinydashboard")

install.packages("leaflet")

install.packages("plotly")

library(tidyverse)

library(shiny)

library(rvest)

library(shinydashboard)

library(leaflet)

library(plotly)

#DATA TRANSFORMATION

#read the dataset

death_causes <- read.csv("~/annual-number-of-deaths-by-cause.csv")

#view the first few rows

head(death_causes)

# Trim column names to simplify them

names(death_causes) <- gsub("Deaths...|...Sex..Both...Age..All.Ages..Number.", "", names(death_causes))

names(death_causes) <- gsub ("\\.s ", "'s", names(death_causes))

names(death_causes) <- gsub ("\\.", " ", names(death_causes))

# Convert to long format to put the causes of death in one column and the no of deaths in another

death_causes <- death_causes %>%

pivot_longer(

cols = -c(Entity, Code, Year),

names_to = "cause_of_death",

values_to = "death_count"

)

# View the first few rows of the transformed dataset

head(death_causes)

#check for null values in each column

colSums(is.na(death_causes))

#WEBSCRAPPING TO GET THE LIST OF AFRICAN COUNTRIES, CODE AND REGIONS

# The URL of the page to scrape

url <- "https://statisticstimes.com/geography/countries-by-continents.php"

# Read the HTML content of the page

page <- read_html(url)

# to extract the table containing the countries and their details

countrytable <- page %>%

html_table(fill = TRUE) %>%

.[[3]] # To get the 3rd table as it is the one I need

# to filter for African countries

african_countries <- countrytable %>%

filter(Continent == "Africa")

# to select the required columns

african_countries <- african_countries %>%

select(`Country or Area`, `ISO-alpha3 Code`, `Region 1`)

african_countries <- african_countries %>%

rename(

Country = `Country or Area`,

Code = `ISO-alpha3 Code`,

Region = `Region 1`

)

# to view the results

african_countries

#check for null values in each column

colSums(is.na(african_countries))

#NOW I HAVE 2 TABLES death_causes and african countries, but i want to filter

#down to only African countries

african_death_causes <- death_causes %>%

inner_join(african_countries, by = "Code") %>%

select(-Country, -Code) # To remove the duplicate Country column if present

#change Entity column to Country

african_death_causes <- african_death_causes %>%

rename(Country = Entity)

# View the first columns of the final table

head(african_death_causes)

#BUILDING THE SHINY APP

# Define UI

ui<- dashboardPage(

dashboardHeader(title = "Causes of Deaths in Africa"),

dashboardSidebar(

sliderInput("yearRange", "Select Year Range:",

min = 1990,

max = 2019,

value = c(1990, 2019),

step = 1,

sep = ""),

selectInput("countryInput", "Select a Country:",

choices = c("All African Countries", unique(african_death_causes$Country)),

selected = "All African Countries"),

selectInput("diseaseInput", "Select a Disease:",

choices = c("All", unique(african_death_causes$cause_of_death)),

selected = "All")

),

dashboardBody(

fluidRow(

column(12, uiOutput("dynamicTextbox")) # This will display the dynamic text box

),

fluidRow(

column(12,

box(title = "Summary",

status = "primary",

width = 12,

solidHeader = TRUE,

valueBoxOutput("totalDeaths", width = 6),

valueBoxOutput("avgDeathsPerYear", width = 6)

)

),

fluidRow(

column(12,

box(title = "Trend of Deaths From 1990 - 2019",

status = "primary",

solidHeader = TRUE,

width = 12,

plotlyOutput("trendPlot", height = 200)

)

,

fluidRow(

column(12,

box(title = "Top 5 Causes of Death",

status = "primary",

width = 12,

solidHeader = TRUE,

plotlyOutput("topCausesPlot", height = 200)

)

),

fluidRow(

column(12,

box( title = "Distribution of Deaths by Region",

status = "primary",

solidHeader = TRUE,

width = 12,

plotlyOutput("deathsByRegionPlot", height = 200) # Use plotOutput for a chart or tableOutput for a table

)

)))))

# Server Logic

server <- function(input, output) {

# Filter data based on inputs

filteredData <- reactive({

african_death_causes %>%

filter(Year >= input$yearRange[1], Year <= input$yearRange[2],

if(input$countryInput != "All African Countries") Country == input$countryInput else TRUE,

if(input$diseaseInput != "All") cause_of_death == input$diseaseInput else TRUE)

})

# Total Deaths card

output$totalDeaths <- renderValueBox({

totalDeathsInMillions <- sum(filteredData()$death_count) / 1e6 # Convert to millions

# format the number with two decimal places

formattedTotal <- sprintf("%.2f", totalDeathsInMillions)

valueBox(paste(formattedTotal, "Million"), "Total Deaths", color = "blue")

})

#Average Yearly Deaths

output$avgDeathsPerYear <- renderValueBox({

dataByYear <- filteredData() %>%

group_by(Year) %>%

summarise(YearlyDeaths = sum(death_count))

avgDeaths <- mean(dataByYear$YearlyDeaths) / 1e6 # Convert to millions for consistency

# format the number with two decimal places

formattedAvg <- sprintf("%.2f", avgDeaths)

valueBox(paste(formattedAvg, "Million"), "Average Deaths per Year", color = "blue")

})

# Death Trend Plot

output$trendPlot <- renderPlotly({

# Prepare the data for plotting

trendData <- filteredData() %>%

group_by(Year) %>%

summarise(TotalDeaths = sum(death_count))

# Plotting the trend of deaths over time using a line chart

p <- ggplot(trendData, aes(x = Year, y = TotalDeaths)) +

geom_line(color = "steelblue", size = 1) + # Adjust the line thickness with 'size'

geom_point(color = "steelblue", size = 1.5, alpha = 0.6) + # Adjust point size and transparency with 'size' and 'alpha'

theme_minimal() +

theme(panel.grid.major = element_blank(), panel.grid.minor = element_blank()) + # Remove gridlines

labs( x = "Year", y = "Total Deaths") +

scale_y_continuous(labels = scales::comma) # Format the y-axis labels with commas

ggplotly(p)

})

# Top 5 Causes of Death Plot

output$topCausesPlot <- renderPlotly({

topCauses <- filteredData() %>%

group_by(cause_of_death) %>%

summarise(TotalDeaths = sum(death_count)) %>%

top_n(5, TotalDeaths) %>%

arrange(desc(TotalDeaths))

# Generate the bar chart

p <- ggplot(topCauses, aes(x = reorder(cause_of_death, -TotalDeaths), y = TotalDeaths, fill = cause_of_death)) +

geom_bar(stat = "identity", fill ="steelblue") +

coord_flip() +

theme_minimal() +

theme( panel.grid.major.y = element_blank()) + # Remove gridlines

labs(x = "", y = "Number of Deaths") +

scale_y_continuous(labels = scales::comma) # Use comma formatting for the y-axis labels

ggplotly(p)

})

#Distribution of death by region

output$deathsByRegionPlot <- renderPlotly({

dataByRegion <- filteredData() %>%

group_by(Region) %>%

summarise(TotalDeaths = sum(death_count)) %>%

arrange(desc(TotalDeaths))

# Generate the bar chart

p <- ggplot(dataByRegion, aes(x = reorder(Region, -TotalDeaths), y = TotalDeaths)) +

geom_bar(stat = "identity", fill = "steelblue") +

theme_minimal() +

theme(panel.grid.major.x = element_blank()) + # Remove gridlines

labs(x = "Region", y = "Total Deaths") +

scale_y_continuous(labels = scales::comma) # Use comma formatting for the y-axis labels

ggplotly(p)

})

output$dynamicTextbox <- renderUI({

# Construct the dynamic text based on the current filter settings

selectedYearRange <- paste(input$yearRange[1], "to", input$yearRange[2])

selectedCountry <- ifelse(input$countryInput == "All", "all countries", input$countryInput)

selectedDisease <- ifelse(input$diseaseInput == "All", "all diseases", input$diseaseInput)

dynamicText <- paste("Analyzing causes of death in", selectedCountry,

"from", selectedYearRange, "focusing on", selectedDisease, ".")

# Return a box or wellPanel with the constructed text

box(

title = "CAUSES OF DEATH IN AFRICA FROM 1990 - 2019",

status = "primary",

solidHeader = TRUE,

width = 12,

dynamicText # dynamic text based on selected filters

)

})

}

# Run the app

shinyApp(ui = ui, server = server)