The standardized HIV prevalence rates in this report may vary across sites or demographic groups. Therefore, statistical tests were performed to determine whether it was valid to summarize these rates by combining sites. We used logistic regression to evaluate the variation in trends among sites and demographic groups for MSM tested in STD clinics, heterosexual patients tested in STD clinics, and IDUs tested in drug treatment programs. If the prevalence rates were uniform across sites and subgroups, then summary indicators of the trends in HIV prevalence rates were valid. For each risk group, we evaluated the variation in trends among the three principal racial/ethnic groups (blacks, Hispanics, and whites), three age groups, sex (for heterosexual men and women and for IDUs), and test site. In addition, for IDUs,
we evaluated variation among the four U.S. geographic regions.
We first evaluated variation in trends among sites within each racial/ethnic group. Additionally, for heterosexual patients and IDUs, we evaluated variation in trends between men and women. For each such group, we included only sites with a total of at least 40 HIV tests and at least 8 HIV-positive test results from 1993–1997. Because we ran multiple tests across each site and multiple tests within sites, we chose a significance level of p < 0.01 instead of a significance level of p < 0.05. We fit two logistic regression models to the proportion of persons with HIV infection. The first model included main effects for year of test (YOT) (continuous), site, age group, and sex (for heterosexual patients and IDUs). The second model included additional terms representing the interaction between YOT and site. We used a likelihood ratio test to determine whether there was a significant interaction between site and YOT.
If this test result was significant, we concluded that the time trend was not
the same at all sites for this racial/ethnic group.
Similarly, we used likelihood ratio tests to determine whether there was a significant interaction between age group and YOT and (for heterosexual patients and IDUs) whether there was a significant interaction between sex and YOT. In addition, for each racial/ethnic group and each site, we fit logistic regression models with main effects for YOT, age group, and (for heterosexual patients and IDUs) sex to determine whether there was a significant trend at any site.
Thus, for MSM tested at STD clinics, each racial/ethnic group included in the initial model had main effects for YOT, site, age group, and the interaction of YOT with site. For heterosexual patients tested at STD clinics, each race/ethnicity group in the initial model had main effects for YOT, site, age group, sex, and the interaction of YOT with site. For IDUs tested at DTCs, each race/ethnicity group in the initial model had main effects for YOT, region, site, age group, sex, and the interaction of YOT with site.
- MSM at STD Clinics
Black MSM. Of the 23 STD clinics included in this report, 21 had sufficient data to analyze individual clinic trends for black MSM. The results of this analysis showed no significant variation among clinics, indicating that it was valid to combine all clinics for the trends analysis. Although prevalence was relatively stable in 5 of the 21 clinics (2 in Atlanta and 1 each in Houston, Los Angeles, and Miami) that met the criteria for separate analysis, there was a decreasing trend for black MSM when all 21 clinics were combined. The tests for age group by YOT interaction were not significant; thus, the trends over time were similar for the three age groups.
Hispanic MSM. The criteria for analysis by individual clinic for Hispanic MSM were met by 11 STD clinics. The test for clinic by YOT interaction was not significant, indicating that the combined clinic trends analysis was valid. There was an overall decreasing trend when the clinics were combined, primarily as a result of significant downward trends in two clinics (New York City and San Francisco). Prevalence was relatively stable in the other nine clinics. Because of small numbers in many of these clinics, there was not sufficient power to determine significant decreases.
However, the age group by YOT interaction was significant. Further analysis showed that two clinics (Houston and San Francisco) were the determinants of this interaction. In both these clinics, there were years during which no persons in the age group “under 25 years” tested HIV-positive. Trends were similar for the other clinics across the three age groups.
White MSM. For white MSM, the trends varied among the 12 clinics with sufficient data for analysis by individual clinic (p = 0.005), primarily because of 1 clinic each in Houston (where rates were stable) and Phoenix (where rates tended to increase). In general, for each of the other 10 clinics, prevalence tended to decrease over time. For the combined analysis, there was an overall significant downward trend when Houston and Phoenix were excluded, as well as when these two clinics were included. The trends across the three age groups did not vary.
- Heterosexual Patients at STD Clinics
For the entire sample of heterosexual patients, the sex by YOT interaction was not significant, indicating that trends in HIV prevalence were the same for heterosexual men and women at the clinics. This finding eliminates the need to analyze separate trends for these men and women. In addition, the age group by YOT interaction was not significant. This finding makes it possible to estimate a valid combined indicator of the trend across the age groups.
Patients. Of the 23 STD clinics included in this report, 22 met the criteria for analysis by individual clinic for black heterosexual men and women. The site by YOT interaction was not significant. Although trends were relatively stable for black heterosexual patients in the analyses by individual clinic, the one exception was a Houston clinic in which HIV prevalence declined significantly (p = 0.006).
Patients. The criteria for analysis by individual clinic for Hispanic heterosexual patients were met by 18 STD clinics. The site by YOT interaction was not significant. Although HIV prevalence declined significantly at one clinic in Los Angeles (p = 0.008), rates were stable for the other clinics.
Patients. Of the 23 STD clinics included in this report, only 10 met the criteria for analysis by individual clinic for white heterosexual patients. For these 10 clinics, the site by YOT interaction was marginally significant (p = 0.08), indicating that trends among clinics are generally consistent. With the exception of a significant downward trend in one Los Angeles clinic (p = 0.004), none of the clinics showed a significant decline when analyzed separately. However, when the 10 clinics were combined, there was a significant downward trend.
- IDUs Entering Drug Treatment Programs
Across the entire sample of IDUs entering drug treatment programs, the interaction of sex and YOT was not significant, indicating that the trends in HIV prevalence were the same for male IDUs and female IDUs. This finding eliminated the need to analyze trends separately for men and women. In addition, the interaction of age group and YOT was not significant. This finding made it possible to estimate a valid overall indicator of the trend across all age groups. The interaction of region and YOT was significant, indicating that trends differed by region. Further analyses showed no significant interactions between site and YOT within regions, indicating that it was valid to combine DTCs for the regional trends analyses.
Northeast. Eight of the 22 DTCs included in this report were located in the Northeast region. Of these, data were sufficient for analyses by individual clinic in seven DTCs for black IDUs, eight DTCs for Hispanic IDUs, and seven DTCs for white IDUs. Overall prevalence for these clinics decreased in the Northeast for each race/ethnicity group from 1993–1997. There were significant downward trends for blacks in one Newark DTC; for Hispanics in one Newark DTC, three New York City DTCs, and the Puerto Rico DTC; and for whites in one Newark DTC and one New York City DTC.
South. Of the six participating DTCs located in the South, the criteria for inclusion in the individual analyses were met for blacks at four DTCs, for Hispanics at three DTCs, and for whites at three DTCs. When these DTCs were combined, there was a significant downward trend. However, analyses by race/ethnicity group showed that the decrease was not an overall result; the only significant decline was among Hispanics. Rates were relatively stable for these DTCs with the exception of the one in San Juan, where there was a significant downward trend (p = 0.001).
Midwest. Only two of the DTCs included in the report were located in the Midwest. One DTC had sufficient data for individual analysis for white IDUs; sufficient data were available from both DTCs for blacks and for Hispanics. When the DTCs were combined, HIV prevalence rates were stable, and no significant trends were found in the individual analyses.
West. Six participating DTCs were in the West. The criteria for analysis by individual clinic were met for blacks at one DTC, for Hispanics at two DTCs, and for whites at five DTCs. Prevalence rates were stable, and no significant trends were found in the individual analyses.
Go to Appendix III