Visualizing Malaria

Malaria is a public and global health problem that has placed heavy burden on individual health, community wellbeing, and economic growth. Malaria’s effects are ever-present, as an approximate 409,000 individuals died of malaria in 2019.

We use open source datasets from the MalariaAtlas package to visualize Malaria’s impact on the globe. Three visualizations are constructed. This blog-post will include the images of each of the visualizations and walk the reader through ideas and inference drawn from the graphs. Note that two of the visualizations are interactive. To maximize impact of the visualizations and allow for interactivity, click here.
I use the plotly library to make the two of the visualization and interactive plots. I chose plotly for its ease of use in python, flexibility, dynamic features, interactive capabilities, and mapping abilities.
The final plot is made with seaborn. I selected seaborn since it allows for faceting and coloring of longitudinal line plots with relative ease.
All computing is done in Python 3.7.8.

Data Pre-Processing:

The datasets provided required a few adjustments before plotting began. The first adjustment was using a Country / ISO Code / World Region crosstable on github to assign world regions to each country in the dataset. The crosstable can be found here. Below we read in all the data:

#read in the three datasets provided as raw csv files
deaths = pd.read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2018/2018-11-13/malaria_deaths.csv", 
                    names =["Country", "Code", "Year", "DPH"], header = 0)
deaths_age = pd.read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2018/2018-11-13/malaria_deaths_age.csv",
                    names = ["Country","Code","Year","Age_group","Deaths"], header = 0)
inci = pd.read_csv("https://raw.githubusercontent.com/rfordatascience/tidytuesday/master/data/2018/2018-11-13/malaria_inc.csv",
                  names = ["Country", "Code", "Year", "Incidence"], header = 0)
#read in a cross-table also found on github
crostab = pd.read_csv("https://raw.githubusercontent.com/lukes/ISO-3166-Countries-with-Regional-Codes/master/all/all.csv")
#rename columns
crostab = crostab.rename(columns = {'name':'Country', 'alpha-3':'Code'})

We then merge the crosstable to each of the three datasets of interest:

#merge columns from crosstable: 
deaths = deaths.merge(crostab[['Code','region','sub-region']], on = 'Code', how = 'left')
deaths_age = deaths_age.merge(crostab[['Code','region','sub-region']], on = 'Code', how = 'left')
inci = inci.merge(crostab[['Code','region','sub-region']], on = 'Code', how = 'left')

The data is ready to be plotted. Final adjustments are requied for each plot, which will be explained below.

Mapping Malaria

I first visualized longitudinal Malaria death rates (Maralia Deaths per 100,000 people) across the globe. Below is a snippet of code that first processed the countries in the dataset, and uses plotly to output a dynamic/interactive plot:

#remove data with no country code since we are interested in only the countries 
deathsdat = deaths.dropna(subset=['Code'])
deathsdat = deathsdat[~deathsdat.Code.isin(['OWID_WRL'])]
fig = px.choropleth(
    deathsdat, 
    locations="Code",                    
    color="DPH",
    hover_name="Country",
    animation_frame="Year",
    range_color = (0,250),
    color_continuous_scale=px.colors.sequential.matter, 
    title = "Map of Maralia Deaths per 100k people in Africa (1990-2016)", 
    labels={'DPH': 'Deaths per 100k'}
)
fig.show()

The heat map shows which regions of the world are most affected by Malarial deaths by year. Below is a gif of the plot. While the plot changes color, the user can hover over countries of interest to show Deaths per 100k. The user can also drag the slider to visualize a specific year.

The interactive/dynamic version of the plot can be found here. Use the play button to watch the map change with year through 1990 - 2016.
The plot shows a substantial concentration of high malarial death rates in Africa, particularly in West Africa.

This plot motivated a more specific look of Malaria in Africa. I subsetted the data to African countries to obtain a more granular view of the problem.

Pinpointing Malaria in Africa

African countries were isolated from the malarial deaths dataset. I created a barplot where each bar represents each country’s death rates. The plot is ordered from least deaths per 100k on the very left to most deaths per 100k on the very right. We must again process the data a bit, and then we can plot the bars dynamically and interactively calling plotly:

#subset to africa data
africa_deathdat = deathsdat[deathsdat.region.isin(['Africa'])]
africa_deathdat = africa_deathdat.sort_values(by = ['Year', 'DPH'], ascending = True)

#change values to be shortened for plotting aesthetics
africa_deathdat["Country"].replace({"Sao Tome and Principe": "STP",
                                    "Central African Republic": "CAR",
                                   "Democratic Republic of Congo" : "DRC",
                                   "Equatorial Guinea": "Eqt.Guinea"}, inplace=True)
fig2 = px.bar(africa_deathdat, x="Country", y="DPH",
             animation_frame="Year", animation_group="Country",
             range_y = [0,250],
            labels={'Country': 'Country (Africa)', 'DPH':'Malaria Deaths per 100,000 people'},
            title="Maralia Deaths per 100k people in Africa (1990-2016)")
fig2.layout.updatemenus[0].buttons[0].args[1]['frame']['duration'] = 1000
fig2.layout.updatemenus[0].buttons[0].args[1]['transition']['duration'] = 100
fig2.update_layout(margin=dict(l=20, r=20, t=30, b=200),paper_bgcolor="white")
fig2['layout']['updatemenus'][0]['pad']=dict(r= 10, t= 150)
fig2['layout']['sliders'][0]['pad']=dict(r= 5, t= 120,)
fig2.show()

Below is a gif of the plot. The user can hover over countries of interest while the plot changes. The user can also drag the slider on the bottom to visualize a specific year. The interactive/dynamic version of the plot can be found here.

When the play button is clicked, as the years progress, the bars shift to represent the countries with the lowest and highest malarial death rates per 100k from 1990 - 2016. During the early 90’s Uganda, Sierrra Leone, Burundi, and Burkina Faso were affected the worst by malarial death rates. Burkina Faso and Sierra Leone continued to struggle through the 2000’s and into 2016. Overall malarial death rates decreased from 1990 - 2016 as seen by the drop in bar heights.

Now that the visualizations show which countries in Africa are most affected, I was curious which groupings of individuals were most affected by malarial deaths.
The death rate dataset stratified by age was one source that would help with this investigation.

Who is Most Affected?

I graphed longitudinal death rates in Africa faceted by age groupings. The age groupings provided in the dataset are the following: Under 5, 5-14, 15-49, 50-69, 70 or older. Below is code that processes the data and plots the facets using seaborn:

#remove data with no country code after merging 
#since we are interested in only the countries 
deaths_agedat = deaths_age.dropna(subset=['Code'])
#subset to africa data
africa_deaths_agedat = deaths_agedat[deaths_agedat.region.isin(['Africa'])]
#create new column on the top three countries
africa_deaths_agedat = (africa_deaths_agedat.assign(Country_group = np.where(africa_deaths_agedat.loc[:,'Country']  == "Nigeria", "Nigeria", 
                                np.where(africa_deaths_agedat.loc[:,'Country'] == "Democratic Republic of Congo", "Democratic Republic of Congo", 
                                np.where(africa_deaths_agedat.loc[:,'Country']  == "Uganda", "Uganda", "All Other Countries"))))
                        .rename(columns={"Age_group": "Age Group", "Country_group": "Country Group"})
                       )
fig3 = sns.relplot(
    data=africa_deaths_agedat,
    x="Year", y="Deaths", col="Age Group",
    col_wrap = 2,estimator=None, color='.7',units = "Country",
    hue = "Country Group",
    col_order = ['Under 5', '5-14', '15-49', '50-69','70 or older'],
    kind="line", facet_kws={'sharey': False, 'sharex': False}
)
fig3.fig.subplots_adjust ( wspace=.05, hspace=.3 )

While graphing, I noticed three countries that stood out from the rest of the African countries: Uganda, Nigeria, and the Democratic Republic of Congo. These countries are colored separately.

During the first pass of this graph, I used the “ ‘sharey’ : False” parameter in seaborn. This gave all plots different y axes. In these plots, it was clear that Uganda, Nigeria, and the Democratic Republic of Congo were affected most in nearly all age groupings. I then tweaked this parameter to be “ ‘sharey’ : True”, meaning all the plots shared the same y axis maximum and minimum.

This change yields a different story to the data. While we know that Uganda, Nigeria, and the Democratic Republic of Congo were affected most, by making the y axis shared, it is clear that the Under 5 Age Grouping is affected most by Malarial Deaths across all countries in Africa.