World Malaria Report

About the data

The WHO World malaria report, released on an annual basis, provides a comprehensive and up-to-date assessment of trends in malaria control and elimination across the globe. The annexes to the report provide useful information on the data sources and methods used to compile the World Malaria Report, a set of regional malaria profiles and tables of historical time series of case data, commodity distribution, funding amounts and policy adoption.

These datasets provide information on long term trends in malaria control and enable comparison between countries. The data provided are on an annual time scale at the national level, and would not be suitable for subnational analysis or to explore seasonal trends in malaria.

Accessing the data

The reports can be accessed on the WHO website. Each report has a web page that includes useful contextual information and additional resources. The “Annexes in Excel format” will download as a zip file that should be extracted. The annexes can be viewed in the original report in PDF format, though for modelling it is usually preferable to have an excel spreadsheet.

The most recent report may not contain all years of estimates. Should earlier years be required, one can access the annexes from earlier reports and merge these.

What do the data look like?

Description

The list of annexes may vary annually. The information included typically looks like this. Data are available for each country and annually. Some datasheets contain a few years of data, while others report data for the most recent year. To make a longer time series of data, you can access older world malaria reports and annexes from the WHO. Be careful to check the data definitions and information on methodology to be sure that datasets can be concatenated.

The annexes can be found in the report itself as in the following examples:

The annexes can also be downloaded as Excel documents from the report page.

Example of the annex layout. This example is from the 2023 World malaria report. Note that the annexes are paired with the report and should be interpreted as such.
Annex 1	Data sources and methods
Annex 2	Number of ITNs distributed through campaigns in malaria endemic countries
Annex 3	WHO Regional profiles
Annex 4	Data tables and methods
	A. Policy adoption
	B. Antimalarial drug policy
	C. Funding for malaria control
	D. Commodities distribution and coverage
	E. Household survey results
	F. Population denominator for case incidence and mortality rate, and estimated malaria cases and deaths
	G. Population denominator for case incidence and mortality rate, and reported malaria cases by place of care
	H. Reported malaria cases by method of confirmation
	I. Reported malaria cases by species
	J. Reported malaria deaths
	K. Methods for Tables A-D-G-H-I-J

Key points to consider

Pitfalls of Annual Data:

Using annual data for your analysis can lead to several pitfalls:

Seasonal variations: Annual data may not capture the seasonal patterns of malaria transmission, which can be influenced by factors such as rainfall, temperature, and vector abundance. This can lead to an over- or underestimation of the disease burden and the impact of interventions.
Temporal aggregation: Aggregating data annually may mask important short-term dynamics, such as outbreaks or the immediate impact of interventions.

Reporting - Incidence vs. Prevalence Data:

It is important to distinguish between incidence and prevalence data when performing your analysis:

Incidence data: Incidence refers to the number of new cases of malaria occurring in a population over a specified period. It provides information on the rate at which new infections occur and is more sensitive to changes in transmission dynamics.
Prevalence data: Prevalence refers to the proportion of the population that has malaria at a given point in time. It provides a snapshot of the disease burden but does not capture the rate of new infections.

Difference Between Policy and Implementation:

There can be discrepancies between the intended policy for ITN distribution and the actual implementation on the ground. Consider the following:

Coverage: The planned ITN coverage may differ from the actual coverage achieved due to factors such as logistics, access to remote areas, and population acceptance.
Timing: The timing of ITN distribution campaigns may vary from the planned schedule. This can affect the impact of ITNs on malaria transmission, particularly if the distribution does not align with peak transmission seasons.
Effectiveness: The effectiveness of ITNs may differ from the ideal scenario due to factors such as improper usage, wear and tear, and insecticide resistance.

National Level - Ignoring Spatial Heterogeneity:

Modelling malaria transmission at a national level can overlook important spatial heterogeneities:

Regional variations: Malaria transmission can vary significantly between different regions within a country due to differences in climate, vector ecology, and socioeconomic factors.
Local hotspots: Even within regions, there can be local hotspots of malaria transmission due to factors such as proximity to breeding sites, population density, and human behavior.
Population movement: The movement of people between different areas can introduce or reintroduce malaria parasites, affecting transmission dynamics.

Examples of data use in literature

The report was released on Jan 31, 2024 and, as such, few studies have used the data. However, here are a few that are either under review or have been published recently:

How to use this data?

For this example, the goal is to determine the reported malaria deaths for the elimination 8 (E8) countries, but your specific analysis may require different data.

We can either manually download, unzip and read in annex 4J or use the whowmr package to access the data.

Warning

The annexes contain references to footnotes. The whowmr package leaves these in the data - the idea being that one should engage with the footnote and proceed with context.

After inspecting the data to understand how it is structured and whether there are any footnotes relevant to our analysis, we can display our data in a table as follows:

Show the code

# View Annex 4J of the 2023 World Malaria Report. Note that we remove the footnotes here.
# The gt package is used for better formatting, but this is optional.
whowmr::wmr2023$wmr2023j |>
  # Hide the footnotes
  mutate(Country=`Country/area` |> stringr::str_remove("\\d.*$")) |>
  select(Country, `2010`:`2022`) |>
  head() |>
  gt() |>
  tab_options(table.align = "left")

Country	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Algeria	1	0	0	0	0	0	0	0	0	0	0	0	0
Angola	8114	6909	5736	7300	5714	7832	15997	13967	11814	18691	11757	13676	12474
Benin	964	1753	2261	2288	1869	1416	1646	2182	2138	2589	2440	2990	2955
Botswana	8	8	3	7	22	5	3	17	9	7	11	5	6
Burkina Faso	9024	7001	7963	6294	5632	5379	3974	4144	4294	1060	3983	4355	4243
Burundi	2677	2233	2263	3411	2974	3799	5853	4414	2481	3316	2276	2292	2374

The E8 countries are Botswana, Eswatini, Namibia, South Africa, Angola, Mozambique, Zambia, and Zimbabwe. To filter the data for these countries, you can use the following code:

Show the code

# Define the E8 countries
e8_countries <- c("Botswana", "Eswatini", "Namibia", "South Africa", "Angola", "Mozambique", "Zambia", "Zimbabwe")

# Filter the data for the E8 countries
wmr2023j_e8 <- whowmr::wmr2023$wmr2023j |>
  filter(`Country/area` %in% e8_countries) |>
  rename(Country="Country/area") |>
  select(!`WHO Region`) # Remove the WHO Region column

# View the filtered dataset
gt(wmr2023j_e8)|>
  tab_options(table.align = "left")

Country	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Angola	8114	6909	5736	7300	5714	7832	15997	13967	11814	18691	11757	13676	12474
Botswana	8	8	3	7	22	5	3	17	9	7	11	5	6
Eswatini	8	1	3	4	4	5	3	20	2	3	2	7	4
Mozambique	3354	3086	2818	2941	3245	2467	1685	1114	968	734	563	408	423
Namibia	63	36	4	21	61	32	65	57	58	6	35	14	28
South Africa	83	54	72	105	174	110	34	301	69	79	38	56	29
Zambia	4834	4540	3705	3548	3257	2389	1827	1425	1209	1339	1972	1503	1361
Zimbabwe	255	451	351	352	406	200	351	527	192	266	400	131	177

How to plot this data?

Show the code

library(stringr)

# Create the lollipop plot
ggplot(wmr2023Ea_Africa, aes(x = fct_reorder(Country, ITNCovPercent, .na_rm = FALSE),
                             y = ITNCovPercent,
                             color = Source)) +
  geom_segment(aes(x = Country, xend = Country, y = 0, yend = ITNCovPercent), color = "grey") +  # Keep constant grey color for segments
  geom_point(size = 4) +
  coord_flip() +  # Adjust the aspect ratio to make the plot taller
  scale_colour_manual_health_radar() +  # Correct color scale for the discrete 'Source' variable
  theme_health_radar() +
  labs(title = "Percentage of Households with at least one ITN",
       subtitle = "Data for selected African countries",
       caption = str_wrap("The percentage of households with at least one insecticide-treated net (ITN) in selected African countries, highlighting variations in ITN coverage. Benin,and Mozambique exhibit the highest coverage, while Kenya and Nigeria have the lowest. Most countries have coverage between 50% and 60%, which supports malaria control efforts if ITNs are properly used and resistance to insecticides remains low. Source: WMR 2023 Annex 4Ea.", width = 75),
       x = "Country",
       y = "Percentage")

Show the code

library(ggrepel)
library(sf)
library(rnaturalearth)
library(rnaturalearthdata)
library(whowmr)

# Define the elimination 8 countries
e8_countries <- c("Botswana", "Eswatini", "Namibia", "South Africa", "Angola", "Mozambique", "Zambia", "Zimbabwe")

# Load the map data for Africa
e8_africa <- ne_countries(continent = "Africa", returnclass = "sf") |>
  mutate(name = if_else(name=="eSwatini", "Eswatini", name)) |>
  filter(name %in% e8_countries)

# Merge the map data with your data
e8_data <- e8_africa |>
  left_join(wmr2023$wmr2023j |> mutate(), by = c("name" = "Country/area"))

# Create the choropleth map
ggplot(data = e8_data) +
  geom_sf(aes(fill = `2022`), color = "lightgrey", size = 0.3) +  # Use fill for the continuous 2022 data
  theme_health_radar() +
  scale_fill_continuous_health_radar(name = "Deaths") +  # Correct fill scale for continuous data
  labs(title = "Reported malaria deaths in the Elimination 8 countries (2022)",
       caption = str_wrap("The choropleth map highlights reported malaria deaths in the Elimination 8 countries in 2022, with Angola, Mozambique, and Zambia showing higher mortality rates. These reported figures may not reflect the full scale of malaria deaths due to potential underreporting and misclassification. Considering estimated deaths alongside reported figures would provide a more comprehensive picture. Source: WMR 2023 Annex 4J.", width = 80)
,
       x = "Latitude",
       y = "Longitude") +
  
  geom_sf_text(aes(label = name, color = ifelse(`2022` > mean(`2022`), "black", "white")), size = 3) +  # Conditional text color
  scale_color_identity() +  
  theme(
    plot.caption.position = "plot",  # Keep the caption outside the plot area
  )

Show the code

whowmr::wmr2023$wmr2023f |>
  # Filter for the frontline E8 countries
  filter(`Country/area` |> stringr::str_detect("Botswana|Eswatini|Namibia|South Africa"),
         # We couldn't get opt_interactive to work with grouped rows so we'll just display fewer columns
         Year>=2020) |>
  # Rename columns to something more readable
  rename(Country="Country/area", `Population at risk`="Population denominator for incidence and mortality rate") |>
  # Hide the footnotes
  mutate(Country=Country |> stringr::str_remove("\\d.*$")) |> 
  # Remove the WHO Region column
  select(!`WHO Region`) |>
  gt(
    groupname_col = "Country",
    row_group_as_column = TRUE
  ) |>
  tab_header(title="Population at risk and estimated malaria cases and deaths in the E8 frontline countries",
             subtitle="Data from WMR 2023 Annex 4F") |>
  # Control how NA values are displayed
  sub_missing(columns = everything(), missing_text = "-") |>
  # Group the Cases and Deaths columns and rename, eg, Cases_Lower to just Lower
  cols_label_with(columns = starts_with("Cases")|starts_with("Deaths"), fn=~stringr::str_remove_all(., ".+_")) |>
  tab_spanner(label="Cases", columns=starts_with("Cases")) |> 
  tab_spanner(label="Deaths", columns=starts_with("Deaths")) |>
  # Format the numbers to be more readable
  fmt_number(columns=c(`Population at risk`,starts_with("Cases")|starts_with("Deaths")), suffixing = T) |>
  # Country column should be bold:
  tab_style(
    style = cell_text(weight = "bold"),
    locations = cells_row_groups()
  ) |>
  tab_style(
    style = cell_borders(
      sides = c("left"),
      weight = px(1)),
    locations = cells_body(
      columns = c(Year, `Population at risk`, Cases_Lower, Deaths_Lower)
    )
  ) |>
  tab_style(
    style = cell_borders(
      sides = c("right"),
      weight = px(1)),
    locations = cells_body(
      columns = c(Deaths_Upper)
    )
  )|>
  tab_options(table.align = "left")

	Year	Population at risk	Cases			Deaths
Population at risk and estimated malaria cases and deaths in the E8 frontline countries
Data from WMR 2023 Annex 4F
	Year	Population at risk	Lower	Point	Upper	Lower	Point	Upper
Botswana	2020	1.69M	1.20K	1.76K	2.70K	-	11.00	-
	2021	1.72M	820.00	1.06K	1.50K	-	5.00	-
	2022	1.74M	420.00	542.00	740.00	-	6.00	-
Eswatini	2020	330.58K	-	233.00	-	-	2.00	-
	2021	333.83K	-	505.00	-	-	5.00	-
	2022	336.47K	-	214.00	-	-	4.00	-
Namibia	2020	1.98M	16.00K	20.19K	25.00K	-	35.00	-
	2021	2.01M	17.00K	21.32K	26.00K	-	14.00	-
	2022	2.04M	13.00K	16.89K	21.00K	-	28.00	-
South Africa	2020	5.88M	-	4.46K	-	-	38.00	-
	2021	5.94M	-	2.97K	-	-	56.00	-
	2022	5.99M	-	2.04K	-	-	29.00	-

Show the code

# Create the plot for Deaths with low and high bands
whowmr::wmr2023$wmr2023f |>
  # Filter for the E8 countries
  filter(`Country/area` |> stringr::str_detect("Angola|Botswana|Eswatini|Malawi|Mozambique|Namibia|South Africa|Zambia|Zimbabwe")) |>
  # Rename columns to something more readable
  rename(Country="Country/area", `Population at risk`="Population denominator for incidence and mortality rate") |>
  # Hide the footnotes
  mutate(Country=Country |> stringr::str_remove("\\d.*$")) |> 
  # Remove the WHO Region column
  select(!`WHO Region`) |> 
  ggplot(aes(x = Year, y = Cases_Point, group = Country, color = Country)) +
    geom_line(linewidth = 1) +
    geom_ribbon(aes(ymin = Cases_Lower, ymax = Cases_Upper, fill = Country), alpha = 0.2) +  # Corrected to use 'fill' for the ribbon
    scale_colour_manual_health_radar() +  # Apply manual color scale for 'color' aesthetic (lines)
    scale_fill_manual_health_radar() +    # Apply manual color scale for 'fill' aesthetic (ribbon fill)
    theme_health_radar() +
    labs(
      title = "Estimated Malaria Cases (2000-2022)",
      subtitle = "With confidence bands",
      x = "Year",
      y = "Number of Cases (thousands)",
      color = "Country",
      fill = "Country",  # Added 'fill' label for the ribbon shading
      caption = str_wrap("The plot displays estimated malaria cases in selected countries from 2000 to 2022, with lines representing case estimates and shaded areas showing the confidence bands. Variations in case numbers and uncertainty across countries are evident, such as stable case numbers in Mozambique, while Angola saw fluctuating cases with increasing uncertainty after 2015. Source: WMR 2023 Annex 4F.", width = 85)

    )

Show the code

# Read in the costing data
costing_df <- read_csv("tutorial_data/costing_df.csv")

# Reshape data for plotting
plot_df <- costing_df |>
  pivot_longer(cols = starts_with("Country_"), names_to = "Donor", values_to = "Amount") |>
  mutate(Donor = gsub("Country_", "", Donor)) # Remove "Country_" from the Donor names

# Set the order of the donors so that "Govt_NMP" is last
plot_df$Donor <- factor(plot_df$Donor, levels = c(
  "GlobalFund", "WorldBank", "PMI_USAID", "OtherBilaterals", 
  "WHO", "UNICEF","Govt_NMP", "OtherContributions"
))

# Create the plot
ggplot(plot_df, aes(x = Year, y = Amount / 1000, fill = Donor)) +
  geom_bar(stat = "identity", position = "stack") +
  facet_wrap(~ Country, scales = "free_y") +
  scale_fill_manual_health_radar() +  # Apply the custom fill scale for 'Donor'
  labs(
    title = "Stacked Bar Plot of Country Contributions (2020-2022)",
    subtitle = "For Selected Countries",
    x = "Year",
    y = "Contributions (in thousands)",
    fill = "Donor",
    caption = str_wrap("The stacked bar plot shows donor contributions to malaria control efforts in selected countries from 2020 to 2022. Contributions vary by country and donor, with Mozambique receiving significant support from the Global Fund and PMI/USAID, while in South Africa, the Global Fund and local government are key contributors. Understanding the distribution of funding is essential for evaluating the sustainability and impact of malaria control programs. Source: WMR 2023 4C.", width = 100)
  ) +
  theme_health_radar()  # Apply the custom radar theme

How can this data be used in disease modelling?

Preparing the data

We obtain estimated case data (Annex 4-F) from Angola for the years 2000 to 2022. For high transmission countries in the WHO African Region, estimates are derived from a spatiotemporal Bayesian geostatistical model, using parasite prevalence data from household surveys. This methodology is further described in Annex 1 of the World Malaria Report.

We can use this data to calibrate our disease models, by fitting model outputs to estimated malaria cases from the World Malaria Report. Throughout the calibration process we adjust the initial parameters we input into the model, making the model more accurate and representative of actual disease dynamics in a specific context. These parameters represent specific dynamics occurring the in the modelled context, for instance ITN usage (\(itn_{use}\)) may be really high in Angola, or the dominant Anopheles vector in this region may have a higher human biting rate (\(a\)) than other Anopheles species. Country-specific parameter values should be obtained from existing datasets such as DHS surveys, literature review, expert knowledge, and published reports from national control programs and other stakeholders. This data carries a level of uncertainty that may prevent the model from making accurate projections, but this can be accounted for in uncertainty and sensitivity analyses to increase the model results’ credibility and robustness.

For this example, we fit a simple malaria model to incidence rates calculated from estimated case data from Angola. While it is simpler to fit to case numbers, fitting to incidence rates allows us to capture the disease burden relative to a changing population size, and is better for comparing burden across different populations. Here, we calculate incidence rate per 1000 of the population at risk with the following formula: \[ \textrm{Incidence rate} = \frac{\textrm {Estimated Cases}}{\textrm {Population at risk}} \times 1000 \]

Show the code

estimated_data <- whowmr::wmr2023$wmr2023f |>
  # Filter for Angola
  filter(`Country/area` == "Angola") |>
  rename(Population_at_risk ="Population denominator for incidence and mortality rate") |>
  select(Year, Cases_Lower, Cases_Point, Cases_Upper, Population_at_risk) |> 
  # Calculate the incidence rate
  mutate(inci_rate = Cases_Point/Population_at_risk*1000,
         inci_rate_lower = Cases_Lower/Population_at_risk*1000,
         inci_rate_upper = Cases_Upper/Population_at_risk*1000) 

estimated_data |> 
  ggplot() +
  #geom_pointrange(aes(x = Year, y = Cases_Point, ymin = Cases_Lower, ymax = Cases_Upper), colour = theme_health_radar_colours[15]) + # to see case numbers
  geom_pointrange(aes(x = Year, y = inci_rate, ymin = inci_rate_lower, ymax = inci_rate_upper), colour = theme_health_radar_colours[9]) +
  labs(title = "Malaria incidence rate in Angola from 2000 to 2022",
       subtitle = "Estimated case data as at 23 October 2023",
       y = "Estimated Cases per 1000",
       caption = stringr::str_wrap("We see that the incidence rate for malaria in Angola has generally decreased over time, but rose again after 2014. We also note wide lower and upper ranges of the data, indicating a high level of uncertainty in the exact estimate of the incidence rate. Source: World Malaria Report 2023, Annex 4-F.")) +
  theme_health_radar()

Model assumptions

We assume that the vector population is at equilibrium, and that ITNs are a widely-used control intervention in Angola. For more details on how to incorporate ITNs and calculating effective coverage, see the DHS page. In a more complex model, we would also account for other Angola-specific interventions, such as larviciding, as well as the treatment cascade illustrated below.

Source: WHO (2014). From malaria control to malaria elimination: a manual for elimination scenario planning.

Simulated calibration

Following calibration and fine-tuning of the parameters, we anticipate model outputs that mimic the pattern and overall trajectory of the estimated data. We simulate an example below:

Show the code

# Simulated model output with 95% Confidence intervals of incidence rates
modelled_incidence_rate <- data.frame(
  Year = 2000:2022,
  incidence_rate = estimated_data$inci_rate + rnorm(length(estimated_data$inci_rate), mean = 30, sd = 25)) |>  
  mutate(
    std_error = 0.04, # for example
    lower_bound = incidence_rate - (1.96 * std_error * 1000),
    upper_bound = incidence_rate + (1.96 * std_error * 1000),
    Type = "Model output")
  
  ggplot() +
      geom_ribbon(data = modelled_incidence_rate, aes(x = Year, y = incidence_rate, ymin = lower_bound, ymax = upper_bound), fill = theme_health_radar_colours[20]) +
     geom_line(data = modelled_incidence_rate, aes(x = Year, y = incidence_rate, colour = Type)) +
  geom_pointrange(data = estimated_data |> mutate(Type = "WHO estimates"), aes(x = Year, y = inci_rate, ymin = inci_rate_lower, ymax = inci_rate_upper, colour = Type)) +
  labs(title = "Malaria incidence rate in Angola from 2000 to 2022",
       subtitle = stringr::str_wrap("Estimated cases plotted alongside calibrated model output with 95% Confidence Intervals"),
       y = "Estimated Cases per 1000",
       colour = "Variable Type", 
       caption = stringr::str_wrap("The calibrated model output projects incidence close to the WHO estimates. The yellow ribbon represents the model's uncertainty, and its overlap with the confidence intervals of the data indicates a good fit. Source: Model output & World Malaria Report 2023, Annex 4-F.")) +
  theme_health_radar()

The next steps are to validate the model, ensuring that it continues to replicate data outside of the calibration period. A sensitivity analysis would help identify which parameters have the most influence on model outputs, and quantify the level of uncertainty in the parameters, further strengthening the reliability of the model for projections going forward.

Policy implications

Calibrated models that produce outputs aligned with real-world incidence or prevalence data are relevant and credible. This increases confidence and utility for decision makers in planning for malaria programming. In addition, models that are tailored to a specific context can better answer questions regarding the impact of certain interventions, or the implementation of other policies in the specified area.