Thuy Luu, PharmD, BCPS, MPH Candidate
Introduction
Human immunodeficiency virus (HIV) causes the body’s immune system to attack itself which can lead to infections and diseases that currently has no cure [1]. HIV impacts the individual, their families, and the communities where these individuals reside. Those with HIV oftentimes come from low to middle income countries with Sub-Saharan Africa being the most affected where 25.6 million people are living with HIV and where 66% of new HIV infections occurred in 2015[1].
Efforts to mitigate the burden of HIV include aid from the United States Agency for International Development (USAID) and President’s Emergency Plan for Aids Relief (PEPFAR). These organizations have goals to save the lives of those suffering from HIV/AIDS around the world, to end extreme poverty, and to enable societies to realize their potential [2.3].
This paper seeks to find out if there is a correlation between United States Federal aid towards HIV and prevalence of HIV in Sub-Saharan Africa.
Methods
To determine if there is a correlation between funding towards HIV and prevalence of HIV, a visual analysis using Tableau and statistical analysis using R was completed. Tableau is a software with a publicly available platform whose goal is to help people see and understand data through visual stories [4]. R is a free statistical and graphics software [5]. Data from the World Bank and USAID were used to determine prevalence and funding towards HIV. The World Back is an organization of 189 member countries whose goals are to end extreme poverty and promote shared prosperity [6].
The World Bank data was used to find prevalence of HIV reported for those aged 15-49 in 2015. After determining the top 10 counties with the highest prevalence from the World Bank data, the USAID data provided the amount in US dollars of funding by USAID towards HIV and funding by PEPFAR for those top 10 countries in 2015. This data is compiled in Table 1.
In a secondary analysis to determine if there is a correlation between United States Federal aid towards HIV on an individual level, total population data was obtained from the World Bank to determine dollars per capita. This data is compiled in Table 3 below. Please see table 5 for a compilation of both tables.
Tableau was then used to geographically locate the top 10 countries and visually represent the prevalence of HIV with circles and funding with color intensity. The greater the prevalence of HIV, the larger the circle. The greater the funding, the darker the shade of the circle.
R was used to statistically determine whether there was a correlation between funding towards HIV and prevalence of HIV.
Results and Discussion
An important note about the prevalence data obtained from the World Bank was it reported prevalence of HIV for those aged 15-49. This excludes those less than the age of 15 and those greater than the age of 49 who have HIV. The end outcome would be an underestimation of prevalence of HIV. This underestimation was consistently reported across all reported HIV prevalence data. In the secondary analysis, this underestimation of prevalence would lead to an overestimation of funding per capita in the HIV prevalent population. Again, this overestimation of funding would be consistent across all countries.
The country with the lowest prevalence of HIV, Uganda, did not have the highest or lowest funding, however it did have the second highest funding of the top 10 countries with highest prevalence of HIV. Of the top 10 countries with the highest prevalence of HIV, South Africa received the most funding while having the fourth highest prevalence of HIV. Lesotho has the least funding of the top 10 countries with the highest prevalence of HIV while having the second highest prevalence of HIV behind Swaziland.
Statistically, the data did not meet the all of the assumptions (independence, normal distribution, and linearity) to meet parametric correlation testing. The data was assumed to meet the assumption of independence since each observation is assumed to be independent of each other. The data was found not to be of normal distribution based on the histograms and plots shown on figure 2. Visually you can see both histograms do not have a standard normal distribution. The plots for prevalence and USAID/PEPFAR aid appear to have normal distribution since they move in the general direction of the line of normality, however the data points could be tighter around the line of normality.
Additionally, the data was also assessed for normal distribution using the Shapiro Wilk test. The Shapiro-Wilk test results for prevalence had a p value of 0.6446 (W=0.9480) and results for USAID/PEPFAR aid had a p value of 0.04064 (W=0.8370). The p value for prevalence was greater than 0.05 (alpha set at 0.05) and is not considered to be statistically significant, the null hypothesis that prevalence is normally distributed is not rejected. The p value for USAID/PEPFAR aid was less than 0.05 (alpha set at 0.05) and considered to be statistically significant, the null hypothesis that USAID/PEPFAR aid is normally distributed is rejected. Based on the Shapiro-Wilk test, prevalence data was considered to be normally distributed where USAID/PEPFAR aid was not normally distributed. Overall, it is assumed that prevalence and USAID/PEPFAR are not of normal distribution. The data also failed the assumption of linearity. The scatterplot in figure 3 does not exhibit linearity as the points do not lie closely to the line.
Since prevalence and USAID/PEPFAR did not meet assumptions for parametric testing, non-parametric correlation test was used to determine correlation between funding towards HIV and prevalence of HIV. Spearman’s rank correlation was used assess if there was a correlation between prevalence and USAID/PEPFAR aid. The Spearman rank correlation yielded a p value of 0.06647 (S=22, rho=-0.6121) which is greater than 0.05 (alpha set at 0.05) and considered to not be statistically significant, the null hypothesis that there is no monotonic correlation between prevalence and USAID/PEPFAR aid is not rejected. There is no monotonic correlation between prevalence and USAID/PEPFAR.
To further elucidate the lack of correlation between funding towards HIV and prevalence of HIV, funding towards HIV per capita and prevalence of HIV was compared. To determine funding towards HIV per capita, the amount of USAID HIV and PEPFAR funding in 2015 was divided the number of people who have HIV in 2015, which was determined by multiplying the reported prevalence of HIV in 2015 by the total population in 2015. This data is compiled in Table 3. A strong argument that the total population should be used rather than only those with HIV since, as previously mentioned, HIV effects not only the individual but also affect family and communities in which these individuals reside. For the purpose of this assessment, only those with HIV will be considered.
The visual story in figure 4 on Tableau for this second analysis appeared to also have no correlation between funding towards HIV and prevalence of HIV per capita, similar to funding towards HIV and prevalence of HIV. The country with the highest prevalence of HIV, Swaziland, did not have the most or least amount of funding. The country with the lowest prevalence of HIV, Uganda, did not have the most or least amount of funding. Of the top 10 countries with the highest prevalence of HIV, Namibia received the most funding while having the sixth highest prevalence of HIV. South Africa has the least funding of the top 10 countries with the highest prevalence of HIV while having the fourth highest prevalence of HIV.
Non-parametric testing was also used to assess correlation between funding towards HIV per capita and prevalence of HIV in 2015. As previously shown, prevalence of HIV was considered to have met the assumption of independence, however failed the assumption of normally distributed data and linearity. Funding towards HIV per capita is assumed to meet the assumption of independence since each observation is assumed to be independent of each other. Funding towards HIV per capita does not appear to have normally distributed data based on the histogram below. The data points on the plot do hug the line of normality however the upper tail trails away from the line.
The Shapiro Wilk test was also assessed for normal distribution. The results for funding towards HIV per capita had a p value of 0.7690 (W=0.9585). The p value for funding towards HIV per capita was greater than 0.05 (alpha set at 0.05) and is not considered to be statistically significant, the null hypothesis that funding towards HIV per capita is normally distributed is not rejected. Funding towards HIV per capita is normally distributed. Based on the Shapiro-Wilk test, prevalence data was normally distributed and funding towards HIV by USAID/PEPFAR per capita was normally distributed. Although the Shapiro Wilks tests showed prevalence and funding towards HIV per capita data to be normally distributed, the histograms and plots did not. The data is assumed to be not of normal distribution.
Lastly, prevalence and funding per capita data were not linear as show on figure 6. The data points do not closely hug the line.
The Spearman Rank Correlation test was used to assess for correlation between funding towards HIV per capita and prevalence of HIV. The Spearman rank correlation yielded a p value of 1.000 (S=164.0, rho=-0.006061) which is greater than 0.05 (alpha set at 0.05) and considered to not be statistically significant, the null hypothesis that there is no monotonic correlation between funding of HIV per capita and prevalence of HIV is not rejected. There is no monotonic correlation between funding of HIV per capita and prevalence of HIV.
Conclusion
In an effort to determine if there is a correlation between United States Federal aid towards HIV and prevalence of HIV in Sub-Saharan Africa 2015, a visual story of the data on Tableau did not appear to support a correlation. Statistically, the data also supports no correlation between United States Federal aid towards HIV and prevalence of HIV in Sub-Saharan Africa in 2015. Additionally, in an effort to explain lack of correlation between United States Federal aid towards HIV and prevalence of HIV in Sub-Saharan Africa in 2015, a second analysis comparing United Stated Federal aid towards HIV per capita and prevalence of HIV in Sub-Saharan Africa in 2015 was completed. Again, the data showed there was no correlation between United Stated Federal aid towards HIV per capita and prevalence of HIV in Sub-Saharan Africa in 2015.
References
[1] “Welcome to AIDS.gov.” [Online]. Available: https://www.aids.gov/. [Accessed: 20-Feb-2017].
[2] “USAID History | U.S. Agency for International Development.” [Online]. Available: https://www.usaid.gov/who-we-are/usaid-history. [Accessed: 19-Feb-2017].
[3] B. of P. A. Department Of State. The Office of Electronic Information, “About PEPFAR,” 20-Jan-2009. [Online]. Available: https://www.pepfar.gov/about/. [Accessed: 19-Feb-2017].
[4] “Tableau Software,” Tableau Software. [Online]. Available: https://www.tableau.com/. [Accessed: 21-Feb-2017].
[5] “R: The R Project for Statistical Computing.” [Online]. Available: https://www.r-project.org/. [Accessed: 21-Feb-2017].
[6] “About the World Bank.” [Online]. Available: http://www.worldbank.org/en/about. [Accessed: 19-Feb-2017].