The overall design of SHIMS was based on the direct measurement of HIV incidence at the population level before and after the scale-up of the national combination HIV prevention strategy. To identify eligible participants for the first incidence cohort, a nationally representative, household-based cross-sectional survey was conducted from December 2010 to June 2011, with key epidemiologic and bio-demographic information collected from Swaziland adults, 18 to 49 years of age. At ages 50+ years, recent (incident) HIV infections are very infrequent.

The household sample size was calculated at 14,927 to permit 80% power to detect a 45% reduction in HIV incidence in men from pre- to post- national scale up of the national combination HIV prevention strategy, assuming a baseline HIV incidence of 2.0% among men. We assumed a design effect of 1.25 (based on previous similar DHS surveys), a retention rate of 90%, a household vacancy rate of 13% and a household refusal rate of 5%; and that 10% of men would not be contactable, 23% would be HIV infected and 5% would refuse to participate in the survey. The households were selected using a two-stage cluster sampling procedure, similar in design to that used by the SDHS. In the first stage, a systematic random sample of 575 of the country’s 2076 enumeration areas (EAs) was drawn. In the second stage, a full listing of all households was conducted within each of the 575 EAs. This household listing provided the sampling frame from which a random sample of 26 households was drawn from each EA. Field teams contacted the selected households, and conducted a census of household residents with the head of household (or other available responsible adult household member).

Survey eligibility criteria included residing in or sleeping in the household the night before, reporting an age between 18-49 years inclusive, and the ability to provide informed consent in either SiSwati or English. Further information contributing to the SHIMS sampling methodology and weighting are provided elsewhere [6].

Interviewers administered questionnaires in SiSwati or English to survey participants on demographic, clinical and behavioral topics, including self-reported male circumcision status. HIV testing, including pre- and post-test counseling, was conducted in a private location in or just outside of the home. Blood samples were collected by venipuncture and rapid HIV testing was performed using Determine™ HIV-1/2Ag/Ab Combo (Alere, Japan). Determine reactive samples were confirmed with Uni-Gold™ HIV Test (Trinity Biotech, Ireland), following the country’s serial testing algorithm. Internal and external quality assurance procedures were implemented throughout the survey. Indeterminate test results were resolved by a two HIV EIA testing algorithm (Genscreen HIV-1/2 Version 2 Assay, Bio-Rad Laboratories, California, USA; and Vironostika HIV Uni-Form II plus O, bioMerieux Inc., France), and final HIV status based on the EIA results were returned to participants in the context of high quality post-test counseling including referral to prevention and, if indicated, HIV care services (pre-ART, CD4+ testing).

Survey weights were applied to the data, accounting for differential probability of selection in the cluster sampling procedure and survey non-response rates. Given N’s (with the exception of Table 1) are estimated for SHIMS-sized population. HIV prevalence was based on testing positive in the survey; self-reported HIV status was not used. Illustrations of a circumcised and uncircumcised penis were provided to assist participants in correctly identifying their own (men) or their partner’s (women) circumcision status. Analyses were performed with SAS version 9.2 (SAS Institute, Cary, NC).

All study participants provided written informed consent prior to the collection of data and blood samples. Ethical approval was obtained from the Swaziland Scientific and Ethics Committee, Columbia University Institutional Review Board (IRB) and U.S. Centers for Disease Control and Prevention IRB before initiation of field work.

The 14,891 selected households were approached, 13,335 were found occupied, and of these, 12,571 households participated in the baseline cross-sectional survey, resulting in a household response rate of 94.3% (Table 1). Urban and rural household response rates did not differ substantially. Among participating households, 13,508 women and 11,048 men aged 18-49 were identified as eligible. Despite up to 9 repeated visits to obtain individual interviews, some eligible person were either not contacted or when contacted refused to participate or make themselves available for interview. Eighty-one (81) % of women (11,042) and 65% of men (7,130) took part in the survey, having agreed to both the HIV test and a face-to-face interview. Urban men had a higher response rate than their rural counterparts (68% versus 63%). No significant differences in response rates were observed by age category or region (data not shown).

Over half (56%) of the population was under age 30 and most (70%) lived in rural areas of the country (Table 2), a finding which mirrors the population distribution reflected in the most recent National Census (2007) [7]. Over a quarter (29%) of adults reported ever attending only primary school, 50% reported attending secondary school, and 14% reported attending tertiary education. Seven percent of participants reported never attending school. Education-related differentials between surveyed men and women were minimal.

Nearly one-half of women and nearly two-thirds of men age 18-49 reported not being married, and more unmarried men than women reported never having had sex (16 % versus 6%). More women (22%) than men (11%) reported having a marital partner who either lived in a different household or whose living/sleeping arrangements were unknown. Seven (7) percent of women reported being pregnant.

Overall, 22% of participants reported having no sexual partners in the prior six months, 27% of men and 18% of women. Two-thirds (67%) reported having one partner in the past 6 months, 54% of men and 78% of women; while 10% of participants reported having two or more partners, 18% of men and 4% of women. One in 6 (17%) of men reported that they were circumcised.

HIV prevalence was 32% among adults aged 18-49 years (Table 3), 24% among men and 39% among women. In both men and women, HIV prevalence increased steeply with age and peaked before age 40 (Figure 1). Peak HIV prevalence was lower for men (47%) than women (54%) and occurred at an older age (35-39 versus 30-34). The difference in HIV prevalence by sex was most pronounced among young adults: HIV prevalence among <25 year olds was five times higher among women than men (26% versus 5%), a pattern commonly observed in sub-Saharan Africa associated with earlier age at first sex and exposure to HIV in women.

The data from the current study allow a comparison of HIV prevalence rates and trends with data from the 2006-2007 Swaziland DHS study. The two surveys were very similar, with the SDHS based on a representative sample of 275 EAs and a response rate of 94.8% [1], similar to the SHIMS response rate of 94.4%. The 2006-2007 SDHS HIV prevalence, recalculated for 18-49 year olds was 31%, indicating that by this aggregate measure, the HIV epidemic in Swaziland has shown little change between the between 2007 and 2011. Figure 2 describes inter-survey changes in the age-pattern of HIV prevalence in men and women. While overall HIV prevalence (males and females combined, 18-49) was nearly the same in both surveys (31-32%), important changes in HIV prevalence are evidenced in age and sex decompositions. First, in both men and women, from the 2006 SDHS to SHIMS 2011, there have been declines in HIV prevalence at each age up through age 25-29 for women and 30-34 for men. The change is especially evident in men. Next, in both sexes, peak prevalence is higher and occurs at older ages in SHIMS 2011 compared to 2006-2007 SDHS. The pattern is especially evident in women with peak prevalence increasing from 49% at 25-29 years to 54% at 30-34 years old. Lastly, HIV prevalence remains higher at all ages after the peak in SHIMS 2011 compared to 2006-2007 SDHS; this pattern again is more pronounced in women.

HIV prevalence was similar in urban and rural areas, and across the four regions of Swaziland, ranging from 31% in Shiselweni and Hhohho to 34% in Manzini (Table 3). HIV prevalence varied substantially by education attainment, with higher HIV prevalence observed among both men and women with less education. In men, HIV prevalence among non-school attenders was more than double (41% versus 16%) that of those who attended a tertiary institution (college, university).

Unmarried men and women who reported never having had sex had relatively low HIV prevalence rates (4% women, 1% men). Infections in men and women reporting having had no previous sexual activity is likely due to underreporting associated with prevailing social unacceptability of early sexual debut, especially among girls. Another possible explanation is that coerced sex and rape was not understood and reported as sexual activity. The prevalence of HIV was substantially higher in sexually active unmarried women (43%) compared to sexually active unmarried men (20%). HIV prevalence among women did not vary by pregnancy status (38% for pregnant; 39% for non-pregnant). Among married men and women, HIV prevalence varied little by whether living with partner or not. In both women and men, higher reported numbers of sexual partners (last 6 months) were associated with higher HIV prevalence. In men, HIV prevalence was markedly higher among those reporting one partner in the last six months compared to those reporting no partners (31% versus 9%). There was however little difference in HIV prevalence between men reporting one versus two or more partners (31% versus 28%). Uncircumcised men had a 26 percent HIV prevalence rate compared to 16 percent among circumcised men, although this can be explained in part by the preponderance of circumcised men being younger (i.e. less cumulative exposure to HIV risk).

Overall, 71% of adults in Swaziland reported having received HIV testing services in the past (Table 4). More women (84%) than men (55%) reported prior HIV testing, regardless of current serostatus. Among all HIV-infected adults, 62% were already aware of their positive serostatus while 38% were unaware, either because they had not tested previously (17%) or because they did not receive a positive result in an earlier test (21%).

HIV-infected men were far more likely (50%) than infected women (31%) to be unaware of their positive serostatus, but among those who were aware of their positive status, a larger proportion of men (60%) than women (46%) reported ART use (Figure 3). This prevailing pattern of service utilization results in nearly twice as many HIV infected women (35%) than men (19%) knowing their HIV positive status but not using ART. Age-related differentials in knowledge of HIV serostatus and HIV service utilization are striking. Nearly 4 in 5 (78%) HIV infected men under age 25 were not aware of their serostatus, compared to about one-third (38%) of infected men age 45+ years. Comparable figures for women were 46% and 23%, respectively. Of note, knowledge of HIV serostatus and ART use among the HIV-infected did not vary substantially between urban and rural areas in Swaziland (data not shown).