Census data for the total population of Colorado was obtained from the Colorado State Demography website, which uses the most recent decennial US Census Bureau data and input from county and local agencies to estimate and forecast the population for intercensal years to forecast population at the state and county levels [8]. Numbers of adults aged 18-59 years were extracted from the population totals as was the distribution by sex and the geographic distribution of adults by county. We aggregated Colorado counties using the 2013 National Center for Health Statistics Urban-Rural Classification Scheme for Counties in the following groupings: large central metro/urban core, large fringe metro/suburban, medium/small metro, and nonmetro [9] (see Figure 1). Forecasted 2017 county population data were used to determine total county population, sex, and age stratifications [8].

We conducted a literature review to determine the most relevant and accurate percentages of adult populations that were likely to fall into the MSM and PWID categories. To reflect populations at increased risk of HIV and to remain consistent with the selection criteria used by Smith and colleagues in their first national PrEP estimates, for our final calculations, we selected references that included estimates of proportions of men who had had sex with other men in the past 12 months and proportions of adults who had injected drugs in the past 12 months [6]. We then applied the national and regional estimates from the literature to the Colorado adult male and overall adult populations. Once the estimated numbers of MSM and PWID in Colorado were calculated, we utilized state HIV surveillance data to subtract the number of individuals known to be living with HIV from each group to determine the potential number of individuals at increased risk of HIV acquisition [3].

Individuals with a history of both male-male sex and injection drug use were subtracted from the MSM estimates. The exercise was repeated at the county level by applying varying percentages of individuals estimated to have had male-male sex or injected drugs in the past 12 months by county type (urban-rural designation) and then subtracting the number of MSM or PWID living with HIV in each county type from the total.

Using CDC-derived percentages of individuals in the MSM and PWID categories with indications for PrEP, we calculated the number of individuals in these categories by applying percentages to Colorado as a whole and by summing estimates for each county type [6] (see Textbox 1 for the complete formula).

The number of adults aged 18-59 years living in Colorado in 2017 was 3,252,648. Of those, 50.58% (1,654,138/3,252,648) were men. Overall, 52.66% (1,713,125/3,252,648) of the population lived in counties classified as large central metro/urban core or large fringe metro/suburban[8] (see Table 1 for the distribution of adults by county urbanicity type).

Our literature review yielded 4 publications that characterized the percentage of the given populations of males that could be considered MSM, 3 of which included estimates for male-male sexual activity in the past 12 months [10-13]. The review yielded 2 publications that described population proportions of PWID, 1 of which specifically characterized the percentage of populations with a history of injection drug use in the past 12 months [11,14] (see Table 2 for details of the reviewed publications). The estimates most applicable to our analysis were described by Oster et al [11], who suggested that 2.5% of the male population nationally and 3% of the males in the Western United States had a history of sex with men in the past 12 months. At the county level, estimates for recent male-male sexual activity ranged from 1.1% of adult males in nonmetro counties to 4.4% of adult males in large central metro counties [11]. We compared these findings with estimates produced by Grey et al [10], who suggested that 2.4% of the adult male population nationally had a history of sex with men in the past 12 months and that 3.8% of men in Colorado had had sex with a man in the past 5 years. Male-male sexual activity in Colorado in the past 12 months was not described in the study by Grey et al. County-level estimates ranged from 1% to 6.8% in nonmetro and large central metro counties, respectively [11].

Oster et al [11] also estimated that 0.3% of the adult population nationally and 0.4% of the adult population in the Western United States had a history of injection drug use in the past 12 months. Estimates of recent injection drug use at the county urbanicity levels ranged from 0.3% in the large central metro counties to 0.5% in the nonmetro counties [11].

By applying the national and regional MSM and PWID percentages by Oster et al [11] to the Colorado population as a whole, regardless of county type, we determined that statewide, 41,353-49,624 males aged 18-59 years were likely to be MSM in the past 12 months, depending on the whether we applied the national MSM estimates (lower estimate) or Western US MSM estimates (higher estimate). Using the national and Western US estimates for PWID, we determined that 9758-13,011 individuals (males and females aged 18-59 years) were likely to have injected drugs in the past 12 months, also with the higher estimate derived from estimates for the Western United States rather than nationally. After accounting for individuals living with HIV in those 2 categories, we determined that 33,199-41,470 MSM and 9098-12,351 PWID were eligible for the PrEP indications analysis.

When we applied estimates of percentages of individuals with MSM and PWID behavior in the past 12 months by county urbanicity type to the number of adults ages 18-59 years living in each county type in Colorado, we determined that 36,354 males were estimated to have had male-male sex in the past 12 months and 10,143 individuals were likely to have injected drugs in the past 12 months. After accounting for MSM and PWID living with HIV in Colorado, we estimated that 28,200 MSM and 9483 PWID were eligible for the PrEP indications analysis.

Applying published estimates for the proportions of MSM and PWID with CDC-recommended indications for PrEP, we determined that the MSM population with indications for PrEP ranged from 8200 to 10,243 males statewide. We estimated that 1683-2285 PWID had indications for PrEP statewide. By applying the formula for PrEP indications using the county-level MSM and PWID estimates, we determined that 6965 MSM in Colorado were likely to have indications for PrEP and 1827 PWID were likely to have indications for PrEP (see Table 3 for the estimated numbers of MSM and PWID with indications for PrEP statewide and by county type). Majority of MSM with PrEP indications were located in a large central metro county (Denver) or in large fringe metro counties, while the number of PWID with PrEP indications was more evenly distributed throughout the state (see Figures 2 and 3).

Of the 3,252,648 individuals aged 18-59 years residing in Colorado in 2017, we determined that 8792-12,528 MSM and PWID were likely to have indications for PrEP as described in the 2015 CDC guidelines for PrEP use. Of those, approximately 81%-82% were MSM, and 65% lived in large central or large fringe metro areas. By target population, 70% of MSM lived in large central or large fringe metro areas and 50% of PWID lived in large central or large fringe metro areas, reflecting the more rural distribution of injection drug use. The county-level distribution of PrEP for MSM was similar to the distribution of HIV among MSM in Colorado, in which 70%-75% of the MSM living with HIV resided in the 5-county Denver Metro area [3].

As has been observed with national PrEP estimates, the number of individuals estimated to have an indication for PrEP in Colorado was close to the number of individuals living with HIV in the state [6]. Some authors have suggested that an alternate method for estimating PrEP need could employ HIV diagnoses as a reference point [15]. In 2016, approximately 8500 MSM, 1400 MSM-PWID, and 500 PWID were known to be living with HIV in Colorado, which are similar number overall to the estimates for PrEP we present here [3]. However, the number of non-MSM-PWID living with HIV in Colorado is significantly lower than the number of individuals we estimated were PWID with indications for PrEP.

A more nuanced approach to using HIV diagnosis data to estimate numbers of individuals with indications for PrEP at the local level was recently published by the CDC. This approach relies on the ratio of the percentage of PWID diagnosed with HIV to the percentage of MSM diagnosed with HIV in a given area. This ratio is further refined by applying race and ethnicity data [16]. This approach accounts for heterogeneity in transmission risk factors regionally. Interestingly, the published estimates for MSM with PrEP indications in Colorado were significantly higher than those we present here. This is partly due to the weighted proportion of HIV diagnoses in Colorado who are MSM, but more importantly, this more recent publication relied on much larger MSM estimates based on the report of sexual activity with men in the past 5 years as opposed to that in the past 12 months, as is in the PrEP guidelines [7]. This significant methodological difference is the key to understanding the variation in estimates. Given that sexual activity varies for people over time, especially in the light of advances in PrEP and understanding of treatment as prevention, having both estimates gives a broader picture of the potential for PrEP use in the state. This alternative methodological approach is limited in states with lower numbers of HIV diagnoses annually, where the transmission category distribution may vary significantly year to year. The recent CDC publication also does not describe estimates for numbers of individuals with PrEP indications at the county level, which are crucial for local resource allocation and planning, in particular as it relates to support for PrEP clinical services in underserved counties (see Figures 2 and 3). The method we present here enables jurisdictions to generate targeted PrEP estimates based on more timely local data, which can then be compared to nationally generated state-level estimates as they become available.

Our analysis is subject to several limitations. Most notably, we did not include estimates of heterosexual individuals with indications for PrEP. We initiated this process using similar methodology as that employed for MSM and PWID estimates but deemed the estimates likely to be inaccurate as the epidemiology of HIV in Colorado is heavily skewed toward MSM with a much lower percentage of individuals living with HIV in the state being heterosexual than is the case nationally. Similarly, we were not able to estimate the prevalence of transgender people in Colorado or subsequent numbers of transgender individuals with indications for PrEP. The inclusion of these 2 populations would make the analysis richer and more complete and will be the focus of future efforts at the state and local health department levels.

The analysis is also limited by our reliance on national and regional estimates of sexual behavior and injection drug use, which may or may not be accurate for Colorado. In particular, Colorado has been heavily affected by the opioid epidemic and may have a significantly higher number of PWID than presented here [17]. Also, both the behavioral estimates obtained from the papers by Grey et al and Oster et al as well as the PrEP indication estimates by Smith et al rely on data from the National Health and Nutrition Examination Survey, which is limited to individuals who are not institutionalized or homeless [6,10,11]. This exclusion is likely to lead to an underestimation of the prevalence of recent injection drug use, thereby leading to an underestimate of the number of PWID with indications for PrEP. A revised estimate of the prevalence of PWID that accounts for homeless and incarcerated populations would be of great benefit. Finally, as noted by Smith and colleagues, as sample sizes get smaller, estimation is more unstable [6]. Therefore, the estimates we have presented have been used specifically for program planning purposes and are limited in generalizability.

Although simple in its methodology, this exercise is a practical means to estimate the need for PrEP at the state and local levels. To our knowledge, this is the first instance of a state- and county-level application of national estimates. Additional methodologies using surveys and more precise population-level statistics have been employed in other jurisdictions as alternate approaches to estimating PrEP demand or PrEP targets [18-20]. However, to obtain preliminary estimates, especially for states with relatively smaller epidemics or for whom resources for population-level HIV prevention analyses are more limited, we offer this approach as a feasible option that provides immediate, actionable estimates that can be quickly revised as new estimates for key populations become available.

As Colorado and its individual metro areas develop strategic plans to end the HIV epidemic, measurable targets for PrEP uptake help direct efforts to the most relevant populations and regions [1]. While the estimates for PrEP indication vary when derived by county type compared to the statewide estimates we produced and compared to the most recent estimates for Colorado from CDC, taken as a whole, these estimates provide a range of numbers that can serve as targets for PrEP use, possibly in a stepwise manner starting with more conservative estimates and increasing our targets as we build demand and gain capacity for PrEP provision. At this time, our state health department is conducting an analysis of insurance claims data to determine the approximate number of PrEP prescriptions filled in 2017. This will serve as a starting point for measuring progress toward optimal PrEP uptake. Ongoing education and financial support for PrEP programs will be crucial to ensuring that this highly effective intervention reaches all individuals for whom it could be beneficial.