[Final Program, International Conference on AIDS (4th: 1988: Stockholm, Sweden), Book 2]

4638 OCCUPATIONAL CORRELATES OF HIV INFETION
Charles Magruder, D Cowan, P Kelley
Walter Reed Army Institute of Research,
Washington, DC
Objective. To evaluate possible occupational risk of
HIV infection and to prioritize occupational groups
for prevention efforts including health education.
Methods. The U.S. Department of Defense directive
that reserve forces be screened for HIV antibody
provided an opportunity to study the distribution of
HIV by civilian occupation. Reservists primarily work
in civilian settings. This analysis utilizes job
categories as defined by the Department of Labor and
represents the screening of 106,000 persons. Doubly
Elisa positive sera are confirmed by Western Blot.
Results. The overall prevalence rate was 1.97/1000
[95% CI=1.71,2.25]. Rates were highest for those in
the medical [3.4/1000; 95% CI=2.1,5.3], entertainment/
recreation [3.3/1000; 95% CI=1.9,5.4], secretarial and
clerical [2.6/1000;1.7,4.3], and protective services
[2.4/1000;95% CI=1.5,4.2] fields. All other rates
were < 1.9/1000. The rate for male medical personnel
was 5.7/1000 [95% CI=3.4,8.8] vs. 2.2/1000 [95%
CI=1.9,2.5] for all males. Female medical workers had
a rate of 0.7/1000 [95% CI=0.l,2.6] vs. 0.9/1000 [95%
CI=0.4,1.4] for all females. Rates for single and
married medical workers were 6.4/1000 [95% CI=3.4,
10.6] and 1.8/1000 [95% CI=0.7,3.7] respectively.
Conclusion. Data collection and analysis is ongoing
but preliminary analysis suggests that prevalence
among medical workers is greatly influenced by sex and
mar ital status. Nevertheless, prevalence appears
higher in certain job categories. These groups should
be targeted in future health education efforts.

4639 RACE SPECIFIC TREND ANALYSES OF HIV
ANTIBODY PREVALENCE IN THE UNITED STATES
Lytt I. Gardner, J Brundage, J McNeil,
D Burke, R Miller.
Walter Reed Army Institute of Research, Washington, DC
Objective. To describe HIV antibody (HIV-Ab)
prevalence trends by race for the U.S.
Methods. Over 920,000 white and 234,000 black civilian
applicants for military service (92.1% of the first
1.25 million applicants) have been screened for
HIV-Ab since October 1985. Logistic regression and an
empirical Bayes smoothing technique were used to
examine prevalence trends.
Results. Significant increases in prevalence between
the 1st and 2nd years of the testing program occurred
in non-epidemic, non-metropolitan areas of 6 of the 7
most populous states. From logistic regression we have
found year 1 to year 2 increases most consistently
among applicants aged 25 or younger. For individuals
<= 25 years of age the prevalence increases in
Florida, Texas, and Illinois were most attributable to
increases in the black rates; in Pennsylvania and Ohio
both races contributed. In non-metropolitan New York
State, no increase was seen for either race; for
California, whites contributed more than blacks.
Conclusion. Race specific HIV-Ab trends in
non-epidemic areas reflect a greater contribution to
increases by blacks in 3 of the 7 largest states, and
about the same for whites and blacks in one state. In
another state, whites contributed more to the increase
than blacks. The effect of self-deferral biases on
these trends is discussed.

Notes:

4640 SURVEILLANCE OF HIV-Ab TRENDS USING
A LOCALLY SMOOTHED PREVALENCE ESTIMATE
Lytt I. Gardner, J Brundage, J McNeil.
Walter Reed Army Institute of Research, Washington, DC
Objective. To develop a non-parametric empirical Bayes
methodology for representing HIV antibody (HIV-Ab)
prevalence trends in U.S. counties.
Methods. Over 1.25 million civilian applicants for
military service have been screened for HIV-Ab since
October, 1985. Analysis of geographic and temporal
trends has been uninformative at a geographic level
smaller than that of the state; state rates are
precise but lacking in detail which could be of use to
health planners and epidemiologists. An empirical
Bayes estimation method was developed employing an
adjacency matrix; this improved the precision of the
individual county's prevalence estimate without
sacrificing validity. The result is a locally smoothed
estimate of prevalence which is helpful for
representing time trends at the county level.
Results. These locally smoothed prevalence estimates
supply informative trend information in populous
states such as California, Texas, Florida and New
York. Unexpectedly, increases were seen more
consistently in non-epidemic, non-metropolitan areas
of these large states than in the epidemic
metropolitan areas. There were significant increases
(p<.05) between the Ist and 2nd years in nonmetropolitan areas of Texas, Florida and California.
Conclusion. Analysis of HIV-Ab prevalence using this
method permits a small, precise scale of analysis of
spatial-temporal trends. We will discuss the effect of
potential deferral biases in this population.

4641 A DIRECT ESTIMATION OF THE RATE OF NEW HIV
INFECTIONS IN U.S. ARMY PERSONNEL
John G. McNeil*, F. Wann**, D. Burke*,
J. Brundage*, M. Peterson
*Walter Reed Army Institute of Research, Washington DC
**SRA Technologies, ***Office of the Secretary of
Defense(Health Affairs), Alexandria Virginia
Objective. To describe the rate of incident HIV
infection in a large, well-defined segment of the general
U.S. population.
Methods. Since October 1985, all civilian applicants for
military service, as well as active duty soldiers, have
been tested for the presence of HIV-Ab. As of September
1987, 1.5 million individuals have been tested through
these two routine screening programs. Of.these, 132,920
individuals representing 88,160 person years of follow-up
have been tested more than once for HIV-Ab. Incident
cases must meet specific surveillance definitions
indicating HIV-Ab seroconversion and be clinically
confirmed within 6 months of their serological diagnosis.
Results. Sixty-eight individuals have fulfilled specific
surveillance criteria and are considered incident cases.
The observed incidence rate in this population is 0.77
new infections per 1000 persons per year (0.077%; 95%
Poisson CI: 0.60 - 0.98/1000 per year). Forty-four
additional HIV-Ab seroconverters are not included in this
analysis because seroconversion was discovered through
nonroutine testing (e.g. VD clinic, blood donation), or
because clinical follow-up of a serological diagnosis is
incomplete. Incident cases are demographically similar to
prevalent cases in this population in terms of age, race
and gender specific risk.
Conclusion. A direct estimate of HIV incidence in this
population is 0.77 new infections per 1000 persons/year.

221