How many people are living with HIV? How has this number evolved in the last 20 years?

Looking at the following plot showing the number of people in the world living with HIV can be really alarming:

Infected population has been continuously growing in the last 20 years. But this other one looks much better:

The number of new infections is decreasing and the number of deaths related with AIDS is also decreasing since 2005. In a sense the first graph shows good news, because infected population is living more years.

Let’s see an analysis by country.

Which countries are the most affected?

Let’s look at the percent of population living with HIV in 2017:

## # A tibble: 10 x 2
##    country_name      aids.prop
##    <chr>                 <dbl>
##  1 Eswatini              17.8 
##  2 Botswana              17.2 
##  3 Lesotho               16.3 
##  4 South Africa          12.8 
##  5 Zimbabwe               9.13
##  6 Namibia                8.32
##  7 Mozambique             7.33
##  8 Zambia                 7.12
##  9 Malawi                 5.66
## 10 Equatorial Guinea      4.60

Africa has been the most affected continent. It’s even clearer when we look the following interactive choropleth generated with leaflet:

+−

% living with HIV

0.006 – 1.981
1.981 – 3.956
3.956 – 5.931
5.931 – 7.906
7.906 – 9.881
9.881 – 11.856
11.856 – 13.831
13.831 – 15.807
15.807 – 17.782
No data

Leaflet | © Mapbox © OpenStreetMap contributors Improve this map

The most affected area is in particular the South of Africa, with countries from the North of Africa with one of the lowest HIV-infection shares.

The Shiny App can be used to explore the choropleth for different HIV indicators and years.

How are countries grouped based on HIV indicators?

Let’s see how countries are grouped depending only on HIV related features. In order to do this I only considered the commensurable features, that is, relative values like percent or rates. Firstly, I embedded the dataset to 2D having into account HIV related features. Like that, I could visualize how countries group and see how clusters are related to the geographical location. I had to impute some missing values with the missMDA package prior to the embedding transformation.

UMAP is a new (2018) embedding algorithm that has a similar scheme to t-SNE, the well-known state of the art non-linear embedding algorithm. Their objective is to visualize high-dimensional datasets in 2D or 3D preserving neighborhoods in terms of distances. I decided to use it because it’s much faster than t-SNE and all algorithm decisions are mathematically justified.

Let’s see the resulting embedding for 2015:

We can visualize 6/7 clear groups, two of which are composed only by African countries (pink groups). These two African groups are the most affected countries. All the other clusters have an heterogeneous composition in terms of continents.

Secondly, I clustered the embedded data with the HDBSCAN density-based clustering algorithm from the dbscan library.

I plotted the resulting clustering encoded with colors in the world map:

+−

Leaflet | © Mapbox © OpenStreetMap contributors Improve this map

Like this, we can see the clusters from the embedding but in a more tidy way.

It’s interesting to compare to the same visualization from 20 years before:

All sub-Saharan African countries were in the same group before.

+−

Leaflet | © Mapbox © OpenStreetMap contributors Improve this map

The shiny app can be used to explore how these plots change for different years.

ART treatment

Antiretroviral treatment is the main therapy to fight the HIV infection. There is no cure or vaccine; however, antiretroviral treatment can slow the course of the disease and may lead to a near-normal life expectancy.

Some information about Antiretroviral treatment (ART) from avert.org:

ART works by keeping the level of HIV in your body low (your viral load). This lets your immune system recover and stay strong. Keeping your viral load low also helps to prevent HIV being passed on.

This means that the treatment positively affects non-only the number of AIDS-related deaths, but it also decreases the new infections.

The increasing slope of the plot indicates that every year the good treatment coverage is increasing. Good news.

There is an important aspect to take care about: the health of pregnant women. Some information about the Prevention of mother-to-child transmission (PMTCT):

PMTCT programmes provide a range of services to women and infants. These include preventing HIV infections among women of reproductive age (15–49 years), preventing unwanted pregnancies among women living with HIV, and providing women living with HIV with lifelong ART to maintain their health and prevent transmission during pregnancy, labour and breastfeeding. PMTCT programmes also support safe childbirth practices and appropriate infant feeding, as well as providing infants exposed to HIV with virological testing after birth and during the breastfeeding period, ART for prevention and effective treatment. These actions can reduce the risk from 45% to below 5%.

Let’s look how the PMTCT coverage in Africa evolves:

We can also visualize that mother treatment coverage is increasing since we start to have values. Let’s see how new children infections evolve:

We can see that the closest in time the steeper the slope.

Knowledge about HIV

Of course education, and in particular HIV knowledge is highly related with the spread of the virus.

From Wikipedia: AIDS stigma exists around the world and it prevents many people from seeking HIV testing, returning for their results, or securing treatment, possibly turning what could be a manageable chronic illness into a death sentence and perpetuating the spread of HIV.

This is the why it is so important to provide good information to the people so that they know the actual methods of prevention and stop discriminating people living with HIV.

Let’s look at the percent of population that has reasonable knowledge about HIV. In this case, a person is considered to have knowledge about HIV if they correctly identify the two major ways of preventing the sexual transmission of HIV (using condoms and limiting sex to one faithful, uninfected partner), reject the two most common local misconceptions about HIV transmission, and know that a healthy-looking person can have HIV.

Even though this particular feature is really interesting, it was too difficult to analyze because of the huge proportion of missing values:

As I encountered difficulties because of of the missing data, I decided to plot the evolution of the variable for some of the countries from the risk area that had the biggest amount of measures:

I did the same for males and then plot them together:

Looking at both charts I couldn’t affirm much about this indicator.

Use of condom

As I mentioned in the introduction the methods to prevent HIV infection are safe sex, not sharing needles and male circumcision. From all these prevention actions we have access to the safe sex practice indicator. This indicator measure the percent of the population (ages 15-49) who used a condom at last intercourse with a non-marital and non-cohabiting sexual partner in the last 12 months. Let’s see these indicator by gender:

We observe an increasement in the percent for both available gender information. It is also important to notice that male values are in general much bigger than female values.