Pajongjit Asavikul. Biostatistical methods on HIV infection among Thai military conscripts . Master's Degree(Biostatistics). Mahidol University. : Mahidol University, 2007.
Biostatistical methods on HIV infection among Thai military conscripts
Abstract:
The first objective of this study was to investigate high risk geographical
distribution of HIV infection among Thai military conscripts by applying the mixture
model in disease mapping. This was a cross-sectional study. The secondary data from
periodic surveillance reporting of HIV infection among Thai military conscripts
inducted in November 2005 from Armed Forces Research Institute of Medical
Sciences, Royal Thai Army Medical Department, Ministry of Defense, was used in
this study. There were 28,648 conscripts recruited from 76 provinces of Thailand.
Non-parametric maximum likelihood estimation (NPMLE) was calculated to estimate
the parameters of the mixture model. The smallest number of components was tested
by likelihood ratio statistic (LRS) at α=0.05. Both methods could be accomplished by
the computer package C.A.MAN., developed by Böhning and colleagues. The result
indicated that HIV infection among Thai military conscripts consisted of one
component. This implies that HIV infection among Thai military conscripts is
homogeneous within 76 provinces (LRS =0.00, p-value = 1.00).
The second objective was to determine factors associated with HIV infection
among Thai military conscripts by using logistic regression analysis. This casecontrol
study (1:8) used secondary data which were collected from a number of the
conscripts inducted in November 2005. Cases were conscripts who were HIV seropositive
and agreed to participate in this study, while controls were selected by multistage
random sampling from conscripts who were HIV sero-negative. There were 86
cases and 688 controls. The finding of this study illustrated that there were four
important factors associated with HIV infection among Thai military conscripts, these
were age > 21 years old (OR = 5.15, 95%CI = 3.11-8.54), having ever used noninjecting
illicit drugs (OR = 2.26, 95% CI = 1.36-3.76), having ever had sex with men
(OR = 5.65, 95% CI = 1.98-16.07), and having had a history of sexually transmitted
disease (OR = 2.87, 95% CI = 1.57-5.25). These four important factors could account
for the variation of HIV infection among Thai military conscripts of about 16.04
percent. The split-group method was used to test model validation and it was shown
that the logistic model which obtained in this study was likely valid since it could
reflect the true outcome, 87.91 percent in generalizability.
For further study, other methods of model validation such as bootstrap and
jackknife, should be used for testing and comparison of the accuracy of the model.