Clustering Countries by Healthy Life Expectancy and Current Health Expenditure

ACTEX Learning - AFDP: R Session

Introduction

Clustering is a powerful technique used in data analysis to group similar data points together based on their characteristics. In this example, we will use K-means clustering to group countries based on two important health indicators: Healthy Life Expectancy (HLE) and Current Health Expenditure (CHE) as a percentage of GDP.

Data Preparation

We will use two datasets from the World Health Organization (WHO).

Healthy Life Expectancy (HLE)

Healthy life expectancy (HALE) at birth (years) is an estimate of the average number of years that a person can expect to live in “full health” by taking into account years lived in less than full health due to disease and/or injury.

# Load the Healthy Life Expectancy (HLE) dataset
HLE_raw <- readRDS("data/HLE_raw.rds")
HLE_raw %>% dim()
[1] 24420    34

In this dataset, HALE is calculated from life tables and the prevalence of diseases and injuries that cause health loss. The data frame contains 34 columns and 24420 rows belonging to 185 countries/territories across 22 from 2000 to 2021 for 2 indicators:

  • Healthy life expectancy (HALE) at birth (years)
  • Healthy life expectancy (HALE) at age 60 (years)

We select only Healthy life expectancy (HALE) at birth (years) and Both sexes, and rename columns to be more user-friendly. This is done to simplify the analysis, and to consider the level of Healthy life expectancy (HALE) for the entire population of each country, across all available years, but you can explore other indicators and dimensions in the dataset.

HLE <- HLE_raw %>% 
  filter(Indicator=="Healthy life expectancy (HALE) at birth (years)",
         Dim1=="Both sexes")  %>%  
  select(Location, FactValueNumeric) %>%
  rename(country = Location, hle = FactValueNumeric) 

HLE %>% head()
                   country   hle
1                  Lesotho 44.63
2 Central African Republic 45.43
3                  Somalia 47.42
4                 Eswatini 47.47
5               Mozambique 49.72
6              Afghanistan 50.45

Next, we rename some countries for consistency with other datasets, and calculate the average healthy life expectancy for each country by grouping the data by country and summarizing the healthy life expectancy with the reframe() function.

HLE <- HLE %>%
  mutate(country = recode(country, 
                          "United Arab Emirates" = "UAE",
                          "occupied Palestinian territory, including east Jerusalem" = "Palestine", 
                          "Syrian Arab Republic" = "Syria")) %>%
  group_by(country) %>% 
  reframe(hle=round(mean(hle)))

HLE %>%
  head()
# A tibble: 6 × 2
  country               hle
  <chr>               <dbl>
1 Afghanistan            50
2 Albania                67
3 Algeria                65
4 Angola                 50
5 Antigua and Barbuda    66
6 Argentina              66

Current health expenditure (CHE)

Current health expenditure (CHE) as percentage of gross domestic product (GDP) (%) is the share of a country’s gross domestic product (GDP) that is spent on health care.

CHE_raw <- readRDS("data/CHE_raw.rds")
CHE_raw %>% dim()
[1] 4384   34

In this dataset, Current health expenditure (CHE) as percentage of gross domestic product (GDP) (%) is the share of a country’s gross domestic product (GDP) that is spent on health care. The data frame contains 34 columns and 4384 rows belonging to 194 countries/territories across 23 from 2000 to 2022 for 1 indicators:

CHE <- CHE_raw %>% 
  filter(Indicator=="Current health expenditure (CHE) as percentage of gross domestic product (GDP) (%)")  %>%  
  select(Location, FactValueNumeric) %>%
  rename(country = Location, che = FactValueNumeric) 
  

CHE %>% head
                           country   che
1                Brunei Darussalam  1.82
2                      New Zealand 10.03
3                            Chile 10.06
4     Netherlands (Kingdom of the) 10.10
5 Micronesia (Federated States of) 10.34
6                         Portugal 10.47
CHE <- CHE %>% 
  mutate(Country = recode(country, 
                          "United Arab Emirates"="UAE",
                          "occupied Palestinian territory, including east Jerusalem"="Palestine", 
                          "Syrian Arab Republic"="Syria")) %>%
  group_by(country) %>% 
  reframe(che=round(sum(che),3))

CHE %>% head
# A tibble: 6 × 2
  country               che
  <chr>               <dbl>
1 Afghanistan         246. 
2 Albania             144. 
3 Algeria             106. 
4 Andorra             152. 
5 Angola               68.3
6 Antigua and Barbuda 117.

Jointed dataframe HLE and CHE

CHEHLE <- merge(x = HLE, y = CHE, by = "country") %>% 
  select(country, hle, che)

CHEHLE %>% head
              country hle    che
1         Afghanistan  50 245.83
2             Albania  67 143.64
3             Algeria  65 106.28
4              Angola  50  68.29
5 Antigua and Barbuda  66 117.26
6           Argentina  66 208.09

Clustering with K-means

What is K-means clustering?

The goal of the K-means clustering algorithm is to find groups in the data, with the number of groups represented by the variable K. The algorithm works iteratively to assign each data point to one of K groups based on the features that are provided. Data points are clustered based on feature similarity.

To know a bit more about the function, you can check the documentation with the command below:

?kmeans()
A <- CHEHLE[,2:3]

km <- kmeans(A, centers = 5, iter.max = 1000)
CHEHLE$group <- km$cluster

CHEHLE %>% head
              country hle    che group
1         Afghanistan  50 245.83     2
2             Albania  67 143.64     5
3             Algeria  65 106.28     5
4              Angola  50  68.29     1
5 Antigua and Barbuda  66 117.26     5
6           Argentina  66 208.09     2

We select a random group of countries (50), to avoid overplotting in the visualization.

set.seed(1111)  # for reproducibility

CHEHLE_sample <- CHEHLE[sample(nrow(CHEHLE), 50) ,]

Data Visualization

In this scatter plot, each point represents a country, with the x-axis showing healthy life expectancy (HLE) and the y-axis showing current health expenditure (CHE) as a percentage of GDP. The points are colored based on their cluster group, which was determined using the K-means clustering algorithm. The plot also includes vertical and horizontal lines that divide the data into quadrants, which can help to identify patterns or trends in the relationship between HLE and CHE.

ggplot(CHEHLE_sample %>% 
         mutate(group = as.factor(group)),
       aes(x = hle, y = che, label = country)) +
  geom_text(aes(y = che, x = hle, color = group),
            check_overlap = T,
            size = 3, 
            fontface = "bold",
            show.legend = FALSE)  + 
  scale_color_brewer(palette = "Set1") +
  labs(title = "Healthy life expectancy vs. Current health expenditure",
       x = "Healthy Life Expectancy (years)", 
       y = "Current Health Expenditure (% GDP per cap)") +
  theme_bw() +
  theme(plot.title = element_text(face = "bold", hjust = 0.5, size = 16))

Data sources

References

Back to top