|Ken Shirriff -> Java -> AIDS statistics|
This graph is a Java interface to the CDC AIDS data set and shows the number of new AIDS cases in the US each year (among people older than 13). You may need to enlarge your viewer to see everything. Click here to load the applet without frames.
Using the graph: This applet lets you select different categories and view them in different ways. The top M and F checkboxes can be used to select male only, female only, male and female separately, or male and female combined. The vertical checkboxes can be used to select which CDC categories are displayed; the colored boxes match the graph line colors. The graph will rescale after each change. Normally the graph displays the number of new cases each year for each category. You can also see cumulative totals (the total number of cases through each year), a log scale, the percentage change in the number of cases compared to the previous year, the values as their fraction (percentages) of the year's total. You can also "stack" the categories on top of each other, and can enter minimum and maximum y-scale bounds in the text entries. Pre-programmed button selections can be retrieved through "Presets", and are referenced below to illustrate various points.
About the data: The initial graph (preset 1) shows the total number of new cases each year from 1981 to mid-1997 and the number split into the categories of men who have sex with men, IV drug users, men who have sex with men and use IV drugs, hemophiliacs, people who were infected heterosexually, transfusion recipients, and other/unknown. (1997 data is mid-1996 to mid-1997.) Note that there are many new AIDS cases every year, with over 68,000 in 1996.
Important: the steep rise in 1993 and later drop is a statistical artifact, due to the CDC's looser case definition introduced in 1993. Under the new definition, many cases were reported in 1993 that wouldn't have been reported under the old definition, causing the temporary jump, with smaller effects in later years. Thus, the statistics from 1993 onwards are largely useless.
The number of new cases (preset 2) forms a roughly S shaped curve (until the 1993 definition change). The number of cases rose rapidly and then started to level off by 1992. The log graph (preset 3) makes it clear that AIDS has never been growing exponentially; exponential growth would be a straight line on the log graph. The number of new cases each year is very high however.
The percent change graph (preset 4) shows the percent increase in the number of new cases each year compared to the previous year. Note that the percent increase has dropped fairly steadily since the start of the epidemic (except for the mess from 1993 onwards). This also shows that the growth is not exponential.
It is very important to understand the difference between the percent change from year to year and the number of cases each year. News articles usually report percent change: "the number of new cases increased x% over last year", but this can be misleading. A category with no increase can be much more important than the fastest-growing category. If there were 50000 new AIDS cases last year and 50000 new cases this year, the percent increase is 0%, but this is no real victory, since there were still 50000 new cases. On the other hand, if a different group had 100 new cases last year and 150 this year, the growth rate would be 50%. Despite the much higher growth rate for the second group, the first group is clearly much worse off. If the rates remain constant (which is very doubtful from preset 4), the second group would eventually catch up, so the percent increase is somewhat important for future predictions, but often the percent increase is just used as a "sound bite".
There are way, way more male AIDS cases than female cases (preset 2). Much media attention is focused on the fact that the female growth rate has generally been higher than the male rate (preset 4), but since the number of male cases started so much higher, the victims have remained predominantly male. A graph of the percentage breakdown (preset 5) shows the gap between male and female cases is slowly narrowing, but male cases are likely to remain the majority for many years.
Looking at the separate categories (preset 6) shows that the vast majority of cases occur in men who have sex with men and in IV drug users. This can also be seen in the stacked view (preset 7) or percent view (preset 8).
Nonetheless, there are a significant and growing number of heterosexual cases. The most important heterosexual risk is from an IV drug using partner, especially for women (preset 9). The "other partner" category is unclear. This could be due to partners who don't fall into a risk category, partners who are in a risk category but the AIDS victim didn't know, or cases that haven't been investigated yet.
A plot of the growth rates (preset 10) is a bit of a mess. In general, the growth rates are declining (up to 1992). The fastest growing category switched around often, which is another reason why the media reports on the fastest-growing category are somewhat meaningless. Heterosexual spread was generally the fastest growing category from 1989 onwards, while homosexual spread had dropped to nearly zero growth. The number of cases (preset 1), however, shows that despite the low growth rate, the majority of new cases are homosexual.
Finally, a cumulative graph (preset 11) shows that there were over 570,000 AIDS cases reported in the US by the end of 1996, illustrating the enormous scope of the AIDS problem.
You can experiment with the graphs and form your own opinions. AIDS statistics are very often misused, so I hope that by providing this tool, people can look at the statistics for themselves and get an accurate view of what's happening. Note that this web page is not connected with or endorsed by the CDC, and I have no connection with the CDC.
Send me email at firstname.lastname@example.org if you find any problems, errors, confusing things, or think there are more interpretations that I should describe.
The source is StatsGraph.java and Graph.java. The data file is here. I have some AIDS pages that discuss various AIDS theories.