Rates of Prevalent HIV Infection, Prevalent Diagnoses, and New Diagnoses Among Men Who Have Sex With Men in US States, Metropolitan Statistical Areas, and Counties, 2012-2013

Introduction

The United States’ human immunodeficiency virus (HIV) epidemic has long been characterized by a concentration of infection among men who have sex with men (MSM). MSM accounted for an estimated 53% of people living with a diagnosis of HIV in 2013 and 67% of new HIV diagnoses in 2014, despite representing a minority of men in the United States [1, 2]. Furthermore, individuals of color and those living in the Southern United States also comprise a disproportionate share of new infections, new diagnoses, those living with an HIV diagnosis, and deaths among persons living with an HIV diagnosis [1, 3, 4]. The intersection of these groups, black MSM in the South, represents the most affected subgroup in the United States, for whom extremely high levels of HIV prevalence and incidence have been recorded across a variety of surveillance and research designs [5-9]. The 2015 update to the United States National HIV/AIDS Strategy (NHAS) places priority on MSM, particularly black MSM, and individuals living in the South, and establishes indicators to monitor progress in reducing the extent of HIV disparities in these populations, to monitor the state of the US epidemic, and to guide the allocation of prevention efforts [10]. The implementation of the NHAS will require intensive local action, informed by local estimates of the extent of HIV infection [11].

To best understand the burden of existing and new HIV infections among MSM at subnational levels, high-quality data sources for both infection numerators and population-size denominators are needed, yet historically have been incomplete. In the United States, HIV infection is reportable; since 2004, name-based diagnoses of HIV infection have been reported by all 50 states to the Centers for Disease Control and Prevention (CDC). CDC routinely releases HIV surveillance data in reports and on HIV Atlas [1, 5, 12]. These data include estimated state-, county-, and metropolitan statistical area (MSA)-level counts of MSM newly diagnosed with HIV infection in each year and all MSM currently living with diagnosed HIV infection. AIDSVu.org is a website that uses these data to develop highly detailed maps and other visualizations of the HIV epidemic and allows downloads of county-specific CDC data on HIV prevalence [13]. There is no direct surveillance measurement of the total number of MSM living with HIV infection (ie, including those not yet diagnosed), but CDC has used existing surveillance data to model the number of MSM living with HIV infection at the state level [14].

Unlike for other groupings of persons (eg, by race, sex, or age) for which CDC publishes both counts and rates of HIV diagnoses and prevalence, surveillance-based estimates for MSM are available only as counts. This limits our understanding of not only the total burden of HIV infection among MSM but also the relative burden across geographic areas in which numbers of MSM may vary. CDC led an effort to create a national estimated rate of the MSM living with diagnosed HIV infection using surveillance data from 37 states and an MSM population-size denominator based on a synthesis of data sources, primarily the National Health and Nutrition Examination Survey (NHANES) [2]. CDC also funds an ongoing supplemental surveillance project in 20 of the largest US cities that directly measures the rate of prevalent HIV infection in a nonrepresentative sample of MSM [6-8, 15, 16]. These data are valuable for understanding the impact of HIV infection among MSM in those cities, but other areas need similar estimates to better understand their own community’s HIV epidemic and to target resources appropriately. This is particularly critical as part of the nationwide call for organizations to incorporate the goals and indicators from the NHAS into their program plans [10, 11]. The rate of new HIV diagnoses is a central indicator for success in the NHAS goals of reducing new HIV infections (Goal 1, Indicator 2) and reducing HIV-related health disparities for MSM (Goal 4, Indicator 9) [10].

To generate these rates at the state, county, and MSA levels, we needed new data on the number of MSM in each of those areas. We have recently published a method that allocates the national proportion of MSM to all US states, counties, and MSAs, using additional NHANES results and American Community Survey (ACS) data [17, 18]. Here, we apply these new denominators to the publicly available HIV case surveillance data to obtain rates of HIV diagnoses and prevalence among MSM at national, state, county, and MSA levels.

Methods

We estimated and examined a variety of HIV infection indicators for MSM at multiple levels by combining publicly available disease numerator data with denominators for MSM, as follows.

Numerator Data Sources and Methods

Estimated MSM diagnosis and infection count data were extracted from public CDC and AIDSVu.org data sources to inform estimates at the national level and all 50 US states plus the District of Columbia (Table 1) [1, 12, 13]. County-level data on prevalence of persons living with diagnosed HIV infection (diagnosed prevalence) among MSM came from AIDSVu.org [13]. MSA-level data on diagnosed HIV prevalence and new HIV diagnoses among MSM came from a CDC surveillance report [5].

To protect the privacy of persons living with HIV infection in smaller communities, standard publication practices for HIV surveillance data suppress county data where there are fewer than 5 cases; this includes stratified counts, such as for MSM. In addition, CDC adheres to agreements with each individual state regarding additional restrictions on the public release of HIV surveillance data at the county level. Although for the 2012-2013 HIV case surveillance data most states (n=32) allowed CDC to release data about MSM from all counties that met the 5+ case rule, many either only allowed release of these data from large mostly urban counties (n=10) or did not allow the data to be released at all (n=8).

Denominator Data Sources and Methods

We used the method by Grey et al [17] to estimate the number of MSM in the United States in 2012 and 2013. In brief, we began with 2 published estimates. The first was the result of a meta-analysis-based estimate of the percentage of US men who have had sex with another man in the past 5 years [2]. The second, from an analysis of NHANES data, was the percentages of US men who had sex with another man in the past year, at each of 4 levels of urbanicity: large central metropolitan areas, large fringe metropolitan areas, medium or small metropolitan areas, and nonmetropolitan areas [18]. We next evaluated the percentage of same-sex male couple households among all households in each county, from the ACS [19]. We then calculated the ratio of these percentages to the percentage among all counties at the same level of urbanicity. By multiplying these ratios by the percentages reported by Oster et al [18], we assigned each county a new, locally tailored estimated percentage of men who had sex with a man in the past year among adult men, a method similar to that published by Lieb et al [20]. Finally, we multiplied each county’s estimate by the number of adult men in the county, according to the ACS. In order to estimate the number of men who had sex with a man in the past 5 years, generally considered more indicative of the sexually active MSM population, we scaled the single-year population sizes to equal the estimate given by Purcell et al [2] of 3.9% of the US adult male population. To obtain estimates at the state and MSA levels, we aggregated MSM population size estimates from their composite counties.

In addition to the numerator-based suppression previously mentioned, standard publication practices for HIV surveillance data also suppress the display of county-level rates where there are fewer than 100 persons in the population, including in any single group stratum, such as MSM. Ultimately, of the 1521 counties with available and unsuppressed data on MSM living with an HIV diagnosis, an additional 253 (16%) had fewer than 100 MSM according to our estimation method. The remaining 1268 counties, which contained 84% of the US adult male population and 89% of all MSM living with an HIV diagnosis in 2012, contributed to the county-level analysis.

HIV Burden Indicators

At the state level, we computed the rates of HIV prevalence in 2012 (diagnosed and undiagnosed), diagnosed prevalence in 2012, and new diagnoses in 2013. Prevalence rates were among all MSM; new diagnosis rates were among all MSM and among only MSM not previously diagnosed with HIV (total MSM in 2013 minus total MSM living with a diagnosis at year-end of 2012). State-level estimates were aggregated to yield national estimates. At the county level, we computed the rates of diagnosed HIV prevalence among all MSM in 2012. At the MSA level, we computed the rates of diagnosed HIV prevalence in 2012 and new diagnosis in 2013 among all MSM.

Sensitivity Analyses

To explore the potential effect of regional underreporting or within-urbanicity stratum underrepresentation of male-male cohabitation on the ACS-based Grey estimator, we also produced HIV prevalence and new diagnosis rates using MSM denominators according to more simplified models that assumed estimates from Oster et al and Purcell et al (see Supplement for details) [2].

Results

In the United States in 2012, an estimated 15.0% of MSM were living with HIV infection (diagnosed and undiagnosed) and 11.1% were living with an HIV diagnosis (Tables 2-3, Figure 1). Using the comparison method in Purcell et al [2], the diagnosed HIV prevalence rate among MSM was 57.5 times greater than among other US men. States in the South had the highest rates of diagnosed HIV prevalence among MSM, with all rates of ≥15% located in the South (Table 2, Figure 1). Among states, the rates of diagnosed HIV prevalence among MSM living in Louisiana, Mississippi, and South Carolina were all approximately twice the national rate. The rates of new HIV diagnoses among MSM for 2013 at the state level strongly followed the same pattern as HIV prevalence (Table 2, Figure 2). Only southern states had new diagnosis rates per MSM and per MSM without an HIV diagnosis of ≥1.00/100 MSM, with 2 (Louisiana and Mississippi) having rates per MSM without an HIV diagnosis of ≥2.00/100 MSM.

Plotting the diagnosed HIV prevalence case counts versus rates among MSM in 2012 (Figure 3), 4 groups are notable: populous states with ≥15,000 cases and diagnosed prevalence between 10% and 15% (California, Florida, Illinois, New York, Texas), southern states with diagnosed prevalence of ≥15% and <15,000 cases, and states with diagnosed prevalence of <15% and <15,000 cases. Georgia is uniquely high in both the rate and case count of MSM living with diagnosed HIV infection.

All but 4 high-prevalence MSAs are located in the South, and 6 southern MSAs have diagnosed HIV prevalence rates among MSM for 2013 of ≥25% (Table 4, complete MSA data provided in supplement). The map of county-level diagnosed HIV prevalence rates among MSM in 2012 shows a similar pattern of high-prevalence urban areas (Figure 4). This map also shows that several rural counties in the South have diagnosed HIV prevalence rates of ≥20% and ≥30% (Figure 4).

The new MSM population size estimation by Grey et al also appears to produce similar state-level HIV rates to previous approaches and, in the instances of certain MSAs, more plausible results (see Supplement for full sensitivity analysis results).

‎View this figure

‎View this figure

‎View this figure

‎View this figure

Discussion

Despite incredible achievements in prevention, HIV diagnosis, and antiretroviral therapies over the 35 years of the epidemic, the impact of HIV in US communities of MSM remains staggering. We examined prevalence of HIV at the state and county levels, and found that, although there are important local variations in HIV prevalence, the prevalence of HIV among MSM is consistently orders of magnitude higher than among other Americans. Whereas prevalence of HIV is routinely expressed in cases per 100,000 persons for other groups, for MSM, we report cases per 100 persons [5]. The levels of prevalence among MSM correspond to international benchmarks for epidemicity or hyperendemicity, with all 50 states at least tripling the prevalence criterion for generalized epidemics among MSM (>1% prevalence) and 6 states (all in the South) meeting the UNAIDS criterion for hyperendemicity (>15% prevalence) [21]. The county-level findings demonstrate that although most HIV infections among MSM are concentrated among urban-dwelling MSM, rural areas still represent important places of HIV burden that may be underserved by HIV prevention and care services [5, 22]. Furthermore, rates of new diagnoses among MSM provide insight into the ongoing trajectory of the epidemic, providing a national, regional, and local view of an epidemic force of infection that requires urgent response and prioritization [23].

Our analysis builds on previous reports of the importance of local variations in social and structural epidemic determinants [24], and underscores previous identification of the US South as the area of greatest concentration of HIV infection [10]. In terms of state-specific HIV prevalence rates among MSM, 9 of the top 10 are in the South. We also have new evidence that the HIV epidemic in the South is continuing to grow disproportionately to other regions. New HIV diagnoses rates among MSM for 2013 show that the top 13 states were all in the South, with 10 states having new diagnoses rates of >1% per year. These concentrations of new diagnoses in the South might represent relative increases in new transmissions or could represent previous successes in reducing HIV transmissions among MSM outside the South [25-27]. Our analysis furthers previous surveillance reports and analyses by recognizing that there are differences in the distribution of MSM by state and by county; having high-quality estimates of MSM denominators allows for more reliable comparisons in prevalence and diagnosis rates across geographic areas. Denominators also allow for a more direct quantification of the health disparity among MSM than do rates expressed against a denominator of all men.

Our estimates of HIV prevalence differ importantly from previous estimates of HIV prevalence in the United States. The study by Purcell et al, based on data from 37 states, reported a diagnosed HIV prevalence among MSM in 2007 of about 8% [2, 28]. Our estimate for 2012, including data from all states and the District of Columbia, is that diagnosed HIV prevalence was 11% and that, including estimated undiagnosed infections, 15% of US MSM live with HIV infection. Previous data from the National HIV Behavioral Surveillance System (NHBS) have shown higher HIV prevalence (18% in 2011) among MSM recruited in high-risk venues in the 20 largest cities in the United States [29, 30]; these differences likely reflect selection bias for men in high-risk venues, and the higher risk of HIV infection in urban areas. Our findings have relative advantages over NHBS-derived estimates because of population-based sources of numerator and denominator data and inclusion of areas outside the largest US cities, including rural areas.

Evaluating and monitoring data in smaller geographic areas is critical to making effective local responses and to measuring their impact. Indeed, the first step of the first goal of the NHAS is to assess the communities in which HIV is more concentrated[11]. Our data are also of relevance to the updated National Strategy [31]. In terms of Goal 3, reducing HIV-related disparities, our new data illustrate the extent of the health disparity among MSM—a staggering 57-fold disparity in diagnosed HIV prevalence rates among MSM compared with other US men. Because our method of calculating rates of new HIV diagnoses among MSM can be updated annually, it can also serve as a source of data to evaluate Indicator 9, measures of health disparities among MSM. The updated NHAS uses a denominator of all men to calculate the rate of new HIV diagnoses among MSM nationally, which was the best available data at the time. However, at the local level, similar metrics will have varying amounts of bias, depending on the relative concentration of MSM in a particular jurisdiction. Thus, for local planning and evaluation, rates based on local MSM denominators should be used. To most comprehensively address this purpose, it will require developing similar race-specific estimates of MSM at the state and county levels. Such data are not yet available but would be important for the ongoing evaluation of an epidemic with profound disparities by race and ethnicity.

Limitations

These results have important limitations. Because all residential locations informing HIV case numerators were determined at the time of diagnosis, and the denominators were based on recent ACS data, postdiagnosis migration and undercounting of nonresidents may contribute to mismeasurement. Furthermore, there are differences in the time frames for assessment of male-male sex for numerators and denominators. HIV surveillance classifies males as MSM if the individual reports any sex with a man since 1977, whereas MSM denominators were defined as having male-male sex in the previous 5 years [1, 17]. This shorter time frame is more informative for assessing public-health-actionable HIV burden, compared with a lifetime definition of male-male sex. However, to the extent that some men living with HIV were classified as MSM at the time of diagnoses but may not have engaged in sexual activity in the previous 5 years, some inflation of prevalence may be observed. There is also a potential age mismatch between the HIV case surveillance-based numerator data and the MSM population size denominators. Publicly available HIV surveillance data reports MSM cases for all persons 13 years of age and older, whereas MSM population size methodologies are for MSM 18 years of age and older. This potentially inflates MSM prevalence. However, persons aged 13-19 years compose only 0.96% of all US HIV diagnoses and diagnosed HIV prevalence is low among MSM aged 13-17 years, so the extent of this bias is likely minimal [1, 32].

Conclusions

Surveillance data have been described as the conscience of the HIV epidemic [33], and the new insights provided here on the rates of HIV prevalence and new diagnoses for US MSM constitute a call of conscience for heightened responses and improved monitoring of HIV epidemics among MSM, especially in the South. Across the United States, MSM are affected by HIV at rates that are orders of magnitude higher than for other Americans. This health disparity is even more pronounced in the South. There is a need for increased resources for HIV prevention, treatment, and care for MSM. In the South, this must include expansion of access to health care through Medicaid expansion under the Affordable Care Act; increased access to comprehensive HIV prevention services, including for preexposure prophylaxis (PrEP); and increased resources for programs to support immediate referrals for antiretroviral therapy for those who are newly diagnosed with HIV. Despite the disproportionate impact of HIV in the South, PrEP uptake among MSM is lower in the South than in other geographic areas [34, 35]. In terms of monitoring, we believe that our analysis illustrates the power of having denominators available to characterize health outcomes of sexual minority groups, and we join the Institute of Medicine’s call [36] to collect data on sexual orientation and gender identity in federal data collections and electronic health records, as well as to consider collecting such data in the United States Census. The health disparities in HIV for US MSM illustrated in this report are intolerable, and we call for urgent action to meet the treatment and care needs of those MSM living with HIV and to support all available evidence-based approaches to prevent new infections.