World Bank Health Indicators in Africa (2000-2020): An R Analysis

R
Programming
Healthcare
Data Analysis
A project analyzing World Bank health indicators in Africa, focusing on maternal and neonatal care bottlenecks.
Published

July 9, 2025

Project Summary

The project focuses on analyzing World Bank health indicators in Africa from 2000 to 2020, with a specific emphasis on maternal and neonatal care bottlenecks.

Problem Statement

The high rates of maternal and neonatal mortality in Africa are a significant public health concern. Identifying and addressing the systemic bottlenecks in healthcare delivery is crucial for improving health outcomes in the region.

Approach

  1. Data Collection: Gather health indicators from the World Bank’s World Development Indicators (WDI) database.
  2. Data Cleaning: Handle missing values and ensure data quality.
  3. Exploratory Data Analysis (EDA): Perform univariate, bivariate, and multidimensional analyses to uncover patterns and relationships.
  4. Key Insights Report: Summarize findings and provide actionable recommendations.

Implementation

Load data

We will use the World Bank’s World Development Indicators (WDI) to gather data on maternal and neonatal health indicators. The following indicators will be used:

  1. SH.MED.BEDS.ZS – Hospital beds per 1,000 people
  2. SH.XPD.CHEX.PC.CD – Health expenditure per capita
  3. SP.DYN.IMRT.IN – Infant mortality rate
  4. SH.STA.MMRT – Maternal mortality ratio
  5. SH.H2O.BASW.ZS – Access to basic drinking water
## SECTION 1: DATA COLLECTION ----
# Fetch data from World Bank API for African countries (SSA region)
health_indicators <- WDI(
  country = "all",
  indicator = c(
    "hospital_beds" = "SH.MED.BEDS.ZS",
    "health_expenditure" = "SH.XPD.CHEX.PC.CD",
    "infant_mortality" = "SP.DYN.IMRT.IN",
    "maternal_mortality" = "SH.STA.MMRT",
    "water_access" = "SH.H2O.BASW.ZS"
  ),
  start = 2000,
  end = 2020,
  extra = TRUE
)

# Filter for African countries (using region classification)
africa_data <- health_indicators %>%
  filter(region == "Sub-Saharan Africa" | country %in% c("Egypt, Arab Rep.", "Libya", "Tunisia", "Algeria", "Morocco")) %>%
  select(country, year, hospital_beds, health_expenditure, 
         infant_mortality, maternal_mortality, water_access, region, income)

# Remove regional aggregates
africa_data <- africa_data %>% 
    filter(!is.na(country) & 
    !country %in% c("Africa Eastern and Southern", "Africa Western and Central"))

## SECTION 2: DATA CLEANING ----
# Check for missing values
missing_summary <- africa_data %>%
  summarise(across(everything(), ~sum(is.na(.))))

kable(missing_summary, caption = "Missing Values Count by Variable")
Missing Values Count by Variable
country year hospital_beds health_expenditure infant_mortality maternal_mortality water_access region income
0 0 675 40 0 0 19 0 0 
# Impute missing values using linear interpolation (time-series aware)
africa_clean <- africa_data %>%
  group_by(country) %>%
  arrange(year) %>%
  mutate(
    hospital_beds = ifelse(sum(!is.na(hospital_beds)) >= 2, 
                         na_interpolation(hospital_beds),
                         hospital_beds),
    health_expenditure = ifelse(sum(!is.na(health_expenditure)) >= 2,
                               na_interpolation(health_expenditure),
                               health_expenditure),
    water_access = ifelse(sum(!is.na(water_access)) >= 2,
                         na_interpolation(water_access),
                         water_access)
  ) %>%
  ungroup()

# For mortality rates, we'll leave NAs as they may represent true missing data
# rather than something we should interpolate

# Remove countries with >50% missing data after interpolation
country_missing <- africa_clean %>%
  group_by(country) %>%
  summarise(na_count = sum(is.na(infant_mortality) | is.na(maternal_mortality))) %>%
  filter(na_count <= 10)  # More than 10 years missing

africa_final <- africa_clean %>%
  filter(country %in% country_missing$country)

# Final check
summary(africa_final)
   country               year      hospital_beds   health_expenditure
 Length:1113        Min.   :2000   Min.   :0.100   Min.   :  5.29    
 Class :character   1st Qu.:2005   1st Qu.:0.700   1st Qu.: 12.33    
 Mode  :character   Median :2010   Median :1.100   Median : 17.81    
                    Mean   :2010   Mean   :1.473   Mean   : 51.21    
                    3rd Qu.:2015   3rd Qu.:2.000   3rd Qu.: 59.54    
                    Max.   :2020   Max.   :5.010   Max.   :340.24    
                                   NA's   :126                       
 infant_mortality maternal_mortality  water_access      region         
 Min.   :  9.00   Min.   :  19.0     Min.   :18.68   Length:1113       
 1st Qu.: 38.50   1st Qu.: 210.0     1st Qu.:44.44   Class :character  
 Median : 53.20   Median : 427.0     Median :55.30   Mode  :character  
 Mean   : 55.73   Mean   : 459.7     Mean   :57.62                     
 3rd Qu.: 72.40   3rd Qu.: 604.0     3rd Qu.:71.87                     
 Max.   :234.90   Max.   :1662.0     Max.   :99.32                     
                                                                       
    income         
 Length:1113       
 Class :character  
 Mode  :character

Some of the data points are missing, especially for maternal and neonatal health indicators. We will use linear interpolation to fill in the gaps for hospital beds, health expenditure, and water access, as these are time-series data that can be reasonably interpolated. Mortality rates will be left as is since they may represent true missing data rather than something we should interpolate.

SECTION 3: EXPLORATORY DATA ANALYSIS —-

### UNIVARIATE ANALYSIS ----
# Calculate summary statistics for annotations
stats <- africa_final %>% 
  summarise(
    median = median(hospital_beds, na.rm = TRUE),
    q75 = quantile(hospital_beds, 0.75, na.rm = TRUE),
    who_threshold = 2.3
  )

ggplot(africa_final, aes(x = hospital_beds)) +
  # Enhanced histogram with density curve
  geom_histogram(
    aes(y = ..density..),
    fill = "#3574b2",  # Professional blue
    color = "white",
    bins = 20,
    alpha = 0.9
  ) +
  geom_density(
    color = "#e6550d",  # Complementary orange
    linewidth = 1.2,
    adjust = 1.5        # Smoothing parameter
  ) +
  # Reference lines and annotations
  geom_vline(
    xintercept = stats$who_threshold,
    linetype = "dashed",
    color = "#2ca02c",  # WHO green
    linewidth = 0.8
  ) +
  geom_vline(
    xintercept = stats$median,
    linetype = "dotted",
    color = "#333333",
    linewidth = 0.8
  ) +
  # Professional annotations
  annotate(
    "text",
    x = stats$q75, y = 0.7,
    label = paste0("75% of countries\nhave < ", round(stats$q75, 1), " beds"),
    color = "#333333",
    size = 3.5,
    lineheight = 0.9
  ) +
  annotate(
    "text",
    x = stats$who_threshold + 0.3, y = 0.85,
    label = "WHO recommended\nminimum (2.3)",
    color = "#2ca02c",
    size = 3.5,
    lineheight = 0.9
  ) +
  annotate(
    "text",
    x = stats$median - 0.3, y = 0.85,
    label = paste0("Median: ", round(stats$median, 1)),
    color = "#333333",
    size = 3.5
  ) +
  # Scales and labels
  scale_x_continuous(
    breaks = seq(0, 6, by = 0.5),
    limits = c(0, 6),
    expand = c(0, 0)
  ) +
  scale_y_continuous(
    labels = scales::comma_format(accuracy = 0.1),
    expand = expansion(mult = c(0, 0.1))
  ) +
  labs(
    title = "Deficient Hospital Bed Capacity Across Africa (2000-2020)",
    subtitle = paste0(
      round(mean(africa_final$hospital_beds < stats$who_threshold, na.rm = TRUE) * 100),
      "% of country-years fall below WHO's recommended minimum of 2.3 beds/1,000 population"
    ),
    x = "Hospital Beds per 1,000 Population",
    y = "Density",
    caption = "Source: World Bank Development Indicators | WHO Guidelines"
  ) +
  # Professional theme
  theme_minimal(base_size = 12) +
  theme(
    plot.title = element_text(face = "bold", size = 16, hjust = 0.5),
    plot.subtitle = element_text(size = 12, hjust = 0.5, color = "gray40", margin = margin(b = 20)),
    panel.grid.minor = element_blank(),
    panel.grid.major.x = element_blank(),
    axis.line = element_line(color = "gray60"),
    axis.title.y = element_text(angle = 0, vjust = 0.5),
    plot.caption = element_text(color = "gray50", hjust = 0, margin = margin(t = 10)),
    plot.margin = margin(1, 1, 1, 1, "cm"),
    plot.background = element_rect(fill = "white", color = NA)
  )

75% of country-years fall below WHO’s recommended minimum of 2.3 beds per 1,000 population. This indicates a significant shortfall in healthcare infrastructure across the continent, which is a critical bottleneck for improving health outcomes.

# Health expenditure (log scale due to right skew)
# Calculate key statistics
stats <- africa_final %>%
  summarise(
    median_exp = median(health_expenditure, na.rm = TRUE),
    mean_exp = mean(health_expenditure, na.rm = TRUE),
    q75 = quantile(health_expenditure, 0.75, na.rm = TRUE)
  )

# Enhanced visualization
ggplot(africa_final, aes(x = health_expenditure)) +
  # Histogram with density overlay
  geom_histogram(
    aes(y = ..density..),
    fill = "#2e8b57",  # Professional sea green
    color = "white",
    bins = 25,
    alpha = 0.9
  ) +
  geom_density(
    color = "#36454F",  # Charcoal for contrast
    linewidth = 1.2,
    adjust = 1.5
  ) +
  # Reference lines
  geom_vline(
    xintercept = stats$median_exp,
    linetype = "dashed",
    color = "#d62728",  # Contrasting red
    linewidth = 0.8
  ) +
  geom_vline(
    xintercept = stats$mean_exp,
    linetype = "dotted",
    color = "#333333",
    linewidth = 0.8
  ) +
  # Professional annotations
  annotate(
    "text",
    x = stats$median_exp * 1, y = 1.8,
    label = paste0("Median: $", round(stats$median_exp)),
    color = "#d62728",
    size = 5
  ) +
  annotate(
    "text",
    x = stats$mean_exp * 1, y = 1.8,
    label = paste0("Mean: $", round(stats$mean_exp)),
    color = "#333333",
    size = 5
  ) +
  # Scales
  scale_x_log10(
    labels = dollar_format(accuracy = 1),
    breaks = c(10, 30, 100, 300, 1000),
    expand = c(0, 0)
  ) +
  scale_y_continuous(
    expand = expansion(mult = c(0, 0.1))
  ) +
  # Enhanced labels
  labs(
    title = "Highly Skewed Distribution of Health Expenditure in Africa (2000-2020)",
    subtitle = "75% of country-years spent less than $60 per capita annually on health",
    x = "Health Expenditure per Capita (USD, log scale)",
    y = "Density",
    caption = "Source: World Bank Development Indicators | Inflation-adjusted values"
  ) +
  # Professional theme
  theme_minimal(base_size = 12) +
  theme(
    plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
    plot.subtitle = element_text(color = "gray40", hjust = 0.5, margin = margin(b = 15)),
    panel.grid.minor = element_blank(),
    axis.line = element_line(color = "gray60"),
    plot.caption = element_text(color = "gray50", hjust = 0),
    plot.margin = margin(1, 1, 1, 1, "cm"),
    plot.background = element_rect(fill = "white", color = NA)
  )

On average, African countries spent only $51 per capita on health in 2020, with 75% of country-years spending less than $60. This low expenditure is a significant bottleneck for improving health outcomes, particularly in maternal and neonatal care. This means that many countries are unable to invest adequately in healthcare infrastructure, leading to poor health outcomes.

# 3. Water access vs health outcomes
# Define WHO water access threshold
who_water_threshold <- 80

africa_final %>%
  filter(year == 2020) %>%
  ggplot(aes(x = water_access, y = infant_mortality)) +
  # Enhanced points with better visual encoding
  geom_point(
    aes(size = health_expenditure, color = region, fill = region),
    alpha = 0.8,
    shape = 21,  # Allows both fill and color
    stroke = 0.5  # Border thickness
  ) +
  # Improved regression line
  geom_smooth(
    method = "lm",
    formula = y ~ x,
    color = "#e6550d",  # Professional orange
    fill = "#fdae6b",   # Confidence interval fill
    se = TRUE,         # Show confidence interval
    level = 0.95,
    linewidth = 1.2
  ) +
  # WHO reference line
  geom_vline(
    xintercept = who_water_threshold,
    linetype = "dashed",
    color = "#3182bd",  # Professional blue
    linewidth = 0.8
  ) +
  # Professional color and size scales
  scale_color_brewer(
    palette = "Dark2",
    name = "Region"
  ) +
  scale_fill_brewer(
    palette = "Dark2",
    guide = "none"  # Only use one legend for color
  ) +
  scale_size_continuous(
    range = c(2, 10),
    name = "Health Expenditure\n(USD per capita)",
    labels = scales::dollar_format(),
    breaks = c(50, 100, 200, 400)  # Specific break points
  ) +
  # Axis scales
  scale_x_continuous(
    limits = c(40, 100),
    breaks = seq(40, 100, by = 10)
  ) +
  scale_y_continuous(
    limits = c(0, 100),
    breaks = seq(0, 100, by = 20)
  ) +
  # Enhanced labels and annotations
  labs(
    title = "Water Access and Infant Mortality in African Nations (2020)",
    subtitle = "Size represents health expenditure | Dashed line shows WHO SDG target for water access",
    x = "Population with Access to Basic Drinking Water Services (%)",
    y = "Infant Mortality Rate\n(per 1,000 live births)",
    caption = "Source: World Bank Development Indicators | WHO SDG Target: 80% water access"
  ) +
  # Professional theme
  theme_minimal(base_size = 12) +
  theme(
    plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
    plot.subtitle = element_text(color = "gray40", hjust = 0.5, size = 10),
    panel.grid.minor = element_blank(),
    axis.title.y = element_text(angle = 0, vjust = 0.5),
    legend.position = "right",
    legend.box = "vertical",
    legend.spacing.y = unit(0.5, "cm"),
    plot.caption = element_text(color = "gray50", hjust = 0),
    plot.margin = margin(1, 1, 1, 1, "cm")
  ) +
  # Highlight key threshold
  annotate(
    "text",
    x = who_water_threshold - 5,
    y = 95,
    label = "WHO\nSDG Target",
    color = "#3182bd",
    size = 3.5,
    lineheight = 0.8
  ) +
  # Highlight key countries
  ggrepel::geom_text_repel(
    data = . %>% filter(infant_mortality > 70 | water_access > 90),
    aes(label = country),
    size = 3,
    box.padding = 0.5,
    min.segment.length = 0
  )

Only a handful of countries in Africa meet the WHO standards for both water access and health expenditure. Countries with more than 80% water access consistently show infant mortality rates below 40 per 1,000 live births. This highlights the foundational importance of basic infrastructure in improving health outcomes. Notably, these countries were Mauritius, Seychelles, and Egypt, which have made significant investments in water infrastructure.

### MULTIDIMENSIONAL ANALYSIS ----

# Create correlation matrix with proper NA handling
cor_matrix <- africa_final %>%
  select(
    `Hospital Beds` = hospital_beds,
    `Health Expenditure` = health_expenditure,
    `Infant Mortality` = infant_mortality,
    `Maternal Mortality` = maternal_mortality,
    `Water Access` = water_access
  ) %>%
  cor(use = "pairwise.complete.obs")  # More robust NA handling

# Custom color palette
corr_colors <- brewer.pal(n = 9, name = "RdYlBu")  # Red-Yellow-Blue diverging palette

# Enhanced correlation plot
corrplot(
  cor_matrix,
  method = "color",
  type = "upper",
  col = corr_colors,
  tl.col = "black",
  tl.srt = 45,  # Diagonal text rotation
  addCoef.col = "black",
  number.cex = 0.8,
  number.digits = 2,
  mar = c(1, 1, 2, 1),  # Plot margins
  title = "Correlation Matrix of African Health Indicators (2000-2020)",
  bg = "white",
  is.corr = TRUE,
  diag = FALSE,
  outline = "gray",
  addgrid.col = "gray90"
)

# Add subtitle and source
mtext("Strong negative correlations suggest better health outcomes with increased resources",
      side = 1, line = 3, cex = 0.8, col = "gray40")
mtext("Source: World Bank Development Indicators", 
      side = 1, line = 4, cex = 0.7, col = "gray50")

Strong negative correlations suggest better health outcomes with increased resources. The strongest correlation found was between hospital beds and infant mortality (r = -0.57), highlighting the foundational importance of basic infrastructure. This was followed by water access and infant mortality (r = -0.55), indicating that increased spending on water access is associated with lower infant mortality rates.

# Facet plot by income level

# Define a custom color palette
income_palette <- c(
  "Low income" = "#1f77b4",
  "Lower middle income" = "#ff7f0e",
  "Upper middle income" = "#2ca02c",
  "High income" = "#d62728"
)

ggplot(africa_final, aes(x = health_expenditure, y = infant_mortality)) +
  # Points colored by income with better aesthetics
  geom_point(
    aes(fill = income),
    shape = 21,  # Allows both fill and border
    color = "white",
    size = 3.5,
    alpha = 0.8,
    stroke = 0.3  # Border thickness
  ) +
  # Regression lines with income-matched colors
  geom_smooth(
    aes(color = income),
    method = "lm",
    se = FALSE,
    linewidth = 1.2,
    show.legend = FALSE
  ) +
  # Professional scales
  scale_x_log10(
    labels = dollar_format(accuracy = 1),
    breaks = c(10, 30, 100, 300, 1000)
  ) +
  scale_fill_manual(
    values = income_palette,
    name = "Income Level"
  ) +
  scale_color_manual(
    values = income_palette,
    guide = "none"
  ) +
  # Facet organization
  facet_wrap(
    ~income,
    scales = "free",
    labeller = labeller(income = label_wrap_gen(15))
  ) +
  # Clean labels
  labs(
    title = "Health Expenditure Efficiency by National Income Level",
    subtitle = "African Nations 2000-2020 | Size represents population",
    x = "Health Expenditure per Capita (USD, log scale)",
    y = "Infant Mortality Rate\n(per 1,000 live births)",
    caption = "Source: World Bank Development Indicators"
  ) +
  # Professional theme
  theme_minimal(base_size = 12) +
  theme(
    plot.title = element_text(face = "bold", size = 14, hjust = 0.5),
    plot.subtitle = element_text(color = "grey40", hjust = 0.5),
    panel.grid.minor = element_blank(),
    strip.background = element_rect(fill = "#f7f7f7", color = NA),
    strip.text = element_text(face = "bold"),
    legend.position = "none",  # Using facets instead
    axis.title.y = element_text(angle = 0, vjust = 0.5),
    plot.caption = element_text(color = "grey50")
  ) +
  # Strategic labeling - only most recent year per country
  geom_text_repel(
    data = africa_final %>% 
      group_by(country, income) %>% 
      filter(year == max(year)),
    aes(label = country),
    size = 3,
    box.padding = 0.3,
    min.segment.length = 0.5,
    seed = 123,  # For reproducible positioning
    color = "grey30",
    max.overlaps = 20
  ) +
  # Add efficiency metrics
   geom_text(
    data = africa_final %>% 
      group_by(income) %>% 
      do({
        mod <- lm(infant_mortality ~ log(health_expenditure), data = .)
        data.frame(
          income = first(.$income),
          label = paste0("Slope: ", round(coef(mod)[2], 1), "\n",
                        "R² = ", round(summary(mod)$r.squared, 2))
        )
      }),
    aes(x = Inf, y = Inf, label = label),
    hjust = 1.1,
    vjust = 1.5,
    size = 3,
    color = "grey40",
    lineheight = 0.9
  )

Negative correlation between health expenditure and infant mortality rates is observed, with lower-middle income (Kenya, Zambia, Zimbabwe etc) and Upper middle (South Africa, Algeria, Libya etc) countries showing the most efficient health spending. This suggests that these countries achieve better health outcomes relative to their expenditure compared to low-income countries. This might indicate that lower-middle & upper middle income countries are more efficient in converting health expenditure into improved health outcomes, while low income countries have health outcomes disproportionate to their expenditure, indicating potential inefficiencies.

SECTION 4: KEY INSIGHTS REPORT —-

  1. Resource Availability vs. Outcomes:
    • 75% of African countries fall below WHO’s recommended minimum of 2.3 hospital beds per 1,000 population.
    • Only 25% of country-years spent more than $60 per capita on health, indicating a significant bottleneck in healthcare infrastructure.
    • There is a moderate negative correlation (r = -0.57) between hospital beds and infant mortality, suggesting that increased healthcare infrastructure is associated with better child health outcomes.
    • However, the relationship is not uniform across countries, indicating other factors like healthcare quality, nutrition, and education play significant roles.
  2. Health Expenditure Trends:
    • 75% of country-years spent less than $60 per capita on health, with a median expenditure of only $51 in 2020.
    • The distribution of health expenditure is highly skewed, with a few countries spending significantly more than others.
  3. Water Access Impact:
    • Countries with >80% water access consistently show infant mortality rates below 40 per 1,000 live births.
    • A strong correlation found was between water access and infant mortality (r = -0.55), highlighting the foundational importance of basic infrastructure.
  4. Income-Level Disparities:
    • Lower-middle and Upper-middle income countries show the most efficient health spending, with steeper declines in mortality per dollar spent compared to low-income countries. Every 10% increase in health spending → -12.74*log(1.1) ≈ -1.2 fewer infant deaths/1,000.
    • Low income countries have health outcomes disproportionate to their expenditure, indicating potential inefficiencies. Positive coefficient (1.95): Higher spending correlates with worse outcomes in this group
  5. Regional Variations:
    • North African countries consistently outperform sub-Saharan Africa on all metrics despite similar expenditure levels.
    • Southern Africa shows the most improvement over time, while Central Africa lags behind.

RECOMMENDATIONS:

  1. Targeted Infrastructure Investment: Prioritize water access improvements as they show strong correlation with multiple health outcomes.
  2. Efficiency Focus: Higher spending doesn’t automatically mean better outcomes - need to examine healthcare delivery quality.
  3. Regional Collaboration: Central African countries could benefit from adopting strategies that worked in Southern Africa and Middle-East & North African Countries.
  4. Data Improvement: Significant data gaps exist, particularly for fragile states - better monitoring is needed.