Behavioural Surveillance Surveys - The Wisdom of Whores

The following is one example of how to
calculate the 95 percent confidence interval
from the above data.
Male STD patients: n = 435, p = 15.
95% CI = 15 ± 1.96 * √(15*85)
= 15 ± 3.36
95% CI = 11.6 - 18.4
√435
From these data, it appears that condom
use during the most recent commercial sex act
is relatively low among male STI patients in
Tamil Nadu, and since the sample size is large,
this can be asserted with some degree of
confidence. It is also entirely plausible:
recent unprotected sex with a sex worker may
well be the source of the STI which led to the
respondent’s inclusion in the survey sample.
At the other end of the spectrum lie male
students, 80 percent of whom used a condom
the last time they had sex with a sex worker,
according to the survey findings. However
because only 18 students reported sex with a
sex worker in the last year, the denominator
for this indicator is very small and the
confidence interval is very wide. The range
of the proportion of students using condoms
the last time they had sex with a sex worker,
according to these calculations, lies anywhere
between 62 and 99 percent. In other words,
anywhere from under two thirds of students to
almost all students used condoms the last time
they had sex with a sex worker.
However, this very large range makes
it difficult for program planners to respond
with appropriate prevention programs for this
group and to gauge how successful previous
efforts have been. If the true value lies
closer to 99 percent, then condom use is
very high and the battle has already been
partially won. In this case, prevention
strategies (and corresponding funding) could
be best geared for the maintenance of desired
behaviors. But since there is a large confidence
interval, the true value may just as likely lie
near 62 percent, which indicates that a
significant proportion of the population is still
putting themselves at risk for HIV infection.
This illustrates the difficulty of interpreting
data that have wide confidence intervals.
Bivariate analysis
Bivariate analysis is used to investigate
the relationship between two different variables
(in the case of BSS usually categorical
variables) that may be associated. Variables
are associated if the value of one tells you
something about the value of another.
For example, level of education is generally
associated with income levels. If you know
that a person has a university education,
you could guess that they earn more money
than a person who did not finish primary
school. Obviously there will be exceptions.
The role of statistical tests in bivariate analysis
is to determine the extent to which the
association is a real one at a population level,
and the extent to which it may have occurred
just by chance.
The most common test used in this context
is known as the Chi-square test, usually
written using the annotation: χ 2 77
B EHAV I OR A L S U R V EI L L A NC E SURV EY S CHAPTER 7