Abstract
Purpose:
HIV pre-exposure prophylaxis (PrEP) is highly efficacious, and yet most individuals indicated for it are not currently using it. To provide guidance for health policymakers, researchers, and community advocates, we developed county-level PrEP use estimates and assessed locality and policy associations.
Methods:
Using data from a national aggregator, we applied a validated crosswalk procedure to generate county-level estimates of PrEP users in 2018. A multilevel Poisson regression explored associations between PrEP use and (1) state policy variables of Medicaid expansion and state Drug Assistance Programs (PrEP-DAP) and (2) county-level characteristics from the US Census Bureau. Outcomes were PrEP per population (prevalence) and PrEP-to-need ratio (PnR), defined as the ratio of PrEP users per new HIV diagnosis. Higher levels of PrEP prevalence or PnR indicate more PrEP users relative to the total population or estimated need, respectively.
Results:
Our 2018 county-level dataset included a total of 188,546 PrEP users in the US. Nationally, PrEP prevalence was 70.3/100,000 population and PnR was 4.9. In an adjusted model, counties with a 5% higher proportion of Black residents had 5% lower PnR (rate ratio (RR): 0.95, 95% confidence interval (CI) 0.93, 0.96). Similarly, counties with higher concentration of residents uninsured or living in poverty had lower PnR. Relative to states without Medicaid expansion or PrEP-DAP, states with only one of those programs had 25% higher PrEP prevalence (RR: 1.25, 95% CI 1.09, 1.45), and states with both programs had 99% higher PrEP prevalence (RR: 1.99, 95% CI 1.60, 2.48). There was a significant linear trend across the three policy groups, and similar findings for the relation between PnR and the policy groups.
Conclusions:
In a dataset comprising approximately 80% of PrEP users in the US, we found that Medicaid expansion and PrEP-DAP were associated with higher PrEP use in states that adopted those policies, after controlling for potential confounders. Future research should identify which components of PrEP support programs have the most success, and how to best promote PrEP among groups most impacted by the epidemic. States should support the admirable health decisions of their residents to get on PrEP by implementing policies that facilitate access.
Keywords: Pre-exposure prophylaxis, HIV, prevention
INTRODUCTION
After a thirty-year period in which condoms were the primary tool of public health to directly prevent HIV transmission, a new biomedical prevention era began in 2012 when the United States Food and Drug Administration (FDA) approved antiretroviral medication for HIV pre-exposure prophylaxis (PrEP). In the eight years since approval, the medication has been shown across numerous studies to be remarkably effective.(1, 2) Mathematical modeling indicates that bringing PrEP to scale among populations at high risk for transmission could produce population-level effects,(3) and observed data support this. In New South Wales, Australia, PrEP scale-up among men who have sex with men (MSM) has occurred alongside a substantial decrease in new HIV diagnoses.(4) In the United States, there has been substantial progress in PrEP uptake, with 172,479 individuals estimated to have used PrEP in 2017.(5, 6) Yet there remains substantial room for improvement, with the US Centers for Disease Control and Prevention (CDC) estimating that 1.2 million individuals (95% confidence interval 0.6 million – 1.8 million) are indicated for PrEP.(7)
Ensuring that groups facing the highest risk of HIV transmission are among those using PrEP will both maximize impact on HIV transmission and contribute to health equity. Unfortunately, many of the groups disproportionately impacted by the epidemic are less likely to take PrEP. A survey in 20 cities in the United States found that fewer Black MSM (26%) reported using PrEP than white MSM (42%).(8) Similarly, individuals with lower incomes were less likely to take PrEP. Moreover, individuals facing higher copayments or with less insurance coverage are less likely to stay on PrEP after initiation.(9) State and local efforts have been launched to increase PrEP uptake. Several states have adopted state PrEP Drug Assistance Programs for patients (PrEP-DAP). These programs aid in at least one of two areas: financing of medication costs or financing of ancillary costs such as physician visits and laboratory test costs.(10, 11) Some PrEP-DAP programs also assist patients in finding a local provider, by answering questions about PrEP, or by offering wrap-around services to address other patient needs. Tracking progress made in increasing PrEP use nationally, as well as continual monitoring of existing and emerging disparities in PrEP use, will be essential to inform efforts to enhance PrEP uptake and to target interventions and health policy appropriately.
Previously, we described a national dataset of PrEP users with state-level data from the fourth quarter of 2017.(5, 6) For the present analysis, we created county-level estimates of PrEP use in 2018, derived from a validated crosswalk procedure.(12) The objectives of this study are to characterize PrEP scale-up in the United States by (1) describing the distribution of PrEP use by county-level characteristics and (2) assessing the association between presence of PrEP-DAP and Medicaid Expansion policies on PrEP use through a multilevel analysis. To accomplish this, we use the traditional metric of use per population (PrEP prevalence) and a metric we previously proposed, PrEP use per new HIV diagnosis (PrEP-to-need ratio, PnR).
METHODS
Data
PrEP use data
Details regarding the dataset and the algorithm used to determine PrEP use have been previously published.(5, 6) Briefly, deidentified data come from a national claims data aggregator (Symphony Health) that creates person-level prescription datasets by combing records from clinics, pharmacies, and payers. The dataset includes commercial payers, Medicare Part D, Medicaid, cash, and assistance programs, but does not include closed healthcare systems such as Kaiser Permanente and groups that choose not to share their data with the data aggregator. Geographic data from the aggregator are available at the ZIP3 geographic level (the first 3 digits of a ZIP code). To determine the number of PrEP users per county, we used a US census-based ‘cross-walk’ algorithm that uses population weighting to convert ZIP3 data to county-level assignment.(12) The dataset does not include information regarding race/ethnicity of individuals at the geographic unit of analysis for this manuscript.
The algorithm to determine PrEP use, which was applied to the claims aggregator dataset, has been previously validated and presented.(5, 6) It uses diagnosis codes and medical data to assign TDF/FTC drug exposure periods to be considered as PrEP. Other classifications not assigned as PrEP included postexposure prophylaxis, HIV treatment, chronic hepatitis B management, and unclassified prescriptions due to insufficient data. A TDF/FTC period prevalence was defined as instances of exposure greater than or equal to one day of TDF/FTC in a given year. Estimates involving persons in each table were rounded to the nearest whole person.
A supplemental analysis (Supplement 1) was performed to clarify PrEP prescription missingness in the aggregator dataset. More details on this analysis method can be found in a previous publication.(12) In brief, we show how unclassified prescriptions (TDF/FTC monotherapy prescriptions without medical record data) and classified PrEP prescriptions (TDF/FTC prescriptions with medical records that indicate PrEP) were combined to arrive at a total estimate that was used for all primary analyses in this study. A sensitivity analysis presents a range of national upweighted estimates, seeking to account for the fact that not all pharmacies provided prescription data to the claims aggregator. These upweighted national estimates were obtained by assigning PrEP use increases in proportion to possible levels of missing pharmacy data, with a range from 0–30%. For instance, the ‘10% missingness’ estimate was calculated with (total observed PrEP users)/(1–0.10). We used a ‘best estimate’ national figure of 20% based on a previous validation result.(5) For all other analyses (Tables 1–4, Figure 1, Supplements 2 and 3), we used the ‘0% missingness’ data because claims aggregator data are known to be not missing at random at the sub-national level. For instance, Kaiser Permanente is a closed health system that did not share data and has a sizable number of users concentrated in the West region.(13)
Table 1.
PrEP users, prevalence, and PnR by demographics and state policies in the United States, 2018
| PrEP users 1, 2, 3 N (%) | PrEP users per 100,000 population (prevalence)° | New HIV Diagnoses per 100,000 population° | PrEP-to-Need Ratio (PnR)4 | |
|---|---|---|---|---|
| Totalº | 188546 (100) | 70.3 | 14.2 | 4.9 |
| Demographics of PrEP Usersº | ||||
| Sex | ||||
| Males | 177433 (94) | 135.3 | 23.5 | 5.7 |
| Females | 11932 (6) | 8.7 | 5.3 | 1.6 |
| Age groups | ||||
| 3.3 | ||||
| 5.6 | ||||
| 6.1 | ||||
| 5.4 | ||||
| 55 years and older | 14112 (7) | 15.9 | 4.3 | 3.7 |
| Urbanicity† | ||||
| Large central metro | 100292 (53) | 5.3 | ||
| Large fringe metro | 41680 (22) | 5.3 | ||
| Medium metro | 26218 (14) | 4.1 | ||
| Small metro | 8835 (5) | 3.8 | ||
| Micropolitan | 7119 (4) | 4.4 | ||
| Noncore | 4402 (2) | 27.6 | 5.4 | 5.1 |
| Policy | ||||
| PrEP-DAP States (NASTAD)* | ||||
| 3.9 | ||||
| Yes | 101869 (54) | 100.6 | 15.8 | 6.4 |
| Medicaid Expansion States** | ||||
| 3.1 | ||||
| Yes | 132933 (71) | 80.3 | 12.2 | 6.6 |
| Census region | ||||
| 6.4 | ||||
| 8.5 | ||||
| 3.0 | ||||
| West | 46406 (25) | 73.7 | 11.5 | 6.4 |
Including District of Columbia
aged 13 and above
Source: NASTAD20
Source: Kaiser Family Foundation 21
National Center for Health Statistics (NHCS) urban-rural classification scheme19
PrEP users rounded to nearest integer
Numbers may not sum to total number of PrEP users due to rounding and/or estimation method. For more information on estimation see Sullivan et al.12
Column percents may not total 100 due to rounding
Calculated as PrEP users/New HIV diagnoses. Higher PnR indicates more PrEP users relative to a denominator, which can be interpreted as a higher degree of PrEP scale-up.
Table 4.
Regressions of factors associated with PrEP-to-Need Ratio (PnR) in the United States, 2018
| Effect | Model 1: County | Model 2: County, state | Model 3: County, state, region | ||||||
|---|---|---|---|---|---|---|---|---|---|
| RR | 95% CI LB | 95% CI UB | RR | 95% CI LB | 95% CI UB | RR | 95% CI LB | 95% CI UB | |
| Black concentration 5% † | 0.94 | 0.93 | 0.96 | 0.94 | 0.93 | 0.96 | 0.95 | 0.93 | 0.96 |
| Latinx/Hispamc concentration 5% † | 0.99 | 0.98 | 1.01 | 0.99 | 0.98 | 1.00 | 0.99 | 0.98 | 1.00 |
| Percent poverty 5% † | 0.96 | 0.93 | 0.99 | 0.96 | 0.93 | 0.99 | 0.96 | 0.93 | 0.99 |
| Percent bachelor’s degree or higher 5% † | 1.06 | 1.04 | 1.07 | 1.05 | 1.04 | 1.07 | 1.05 | 1.04 | 1.07 |
| Percent uninsured 5% † | 0.94 | 0.90 | 0.99 | 0.95 | 0.90 | 1.00‡ | 0.95 | 0.90 | 1.00‡ |
| Urbanidty | 0.99 | 0.97 | 1.01 | 0.99 | 0.97 | 1.01 | 0.99 | 0.97 | 1.01 |
| PrEP-DAP OR Medicaid expansion versus none | 1.22 | 0.94 | 1.58 | 1.09 | 0.86 | 1.37 | |||
| PrEP-DAP AND Medicaid expansion versus none | 1.87 | 1.27 | 2.74 | 1.49 | 1.04 | 2.12 | |||
| Census region: Northeast versus Midwest | 1.41 | 1.02 | 1.96 | ||||||
| Census Region: South versus Midwest | 0.72 | 0.55 | 0.95 | ||||||
| Census Region West versus Midwest | 1.01 | 0.75 | 137 | ||||||
Generalized linear mixed effects model with a Poisson distribution, log link, random intercepts for county and state, variance components covariance structure. Model outcome is PrEP use, with an offset of new HIV diagnoses in 2017. Counties with no new HIV diagnoses in that year were excluded from this analysis.
LB, lower bound: UB, upper bound.
Change in the rate ratio given each 5% higher proportion of the covanate, for example, 5% higher uninsured.
Confidence interval does not include 1.0.
Figure 1.
Annual PrEP-to-need ratio (PnR) in the United States by county characteristic quartile, 2012–2018.
Denominator data
Population size data, used as the denominator for PrEP prevalence, were obtained from the US Census Bureau’s 5-year American Community Survey accessed through the National Historical Geographic Information System (NHGIS).(14) HIV diagnosis data, used as the denominator for PnR, were acquired from the US Centers for Disease Control and Prevention and accessed at AIDSVu.org.(15) We used the most recently available HIV diagnosis data, from 2017. New diagnosis data for counties that had 1–4 new HIV diagnoses was unavailable due to data suppression rules that are in place to protect privacy. To attain estimates for these counties, we first subtracted the total reported at the county-level from the state total (comprised of both county-reported diagnoses and county-suppressed diagnoses). We then assigned the difference proportionally to suppressed counties in proportion to their population size, a method we have previously described.(5, 6) We restricted population size and HIV diagnosis data to persons greater than age 13 because we anticipated few persons would be sexually active and prescribed PrEP before this age.
Covariate data
Analyses used a range of socio-demographic covariates selected based on a demonstrated association with PrEP initiation or retention in care(8, 9, 16, 17) and on the variable being publicly available nationally at the county-level. We have previously used these variables in national analyses, with more detail on each metric available elsewhere.(6, 18) In brief, individual-level data regarding the sex and age groups of PrEP users were obtained from the aggregator dataset. Urbanicity and region were categorical variables, assigned based on 2013 National Center for Health Statistics Urban–Rural Classification Scheme (19) and the US Census grouping of states by regions, respectively. County-level characteristics were obtained from the US Census Bureau’s 5-year American Community Survey (ACS), a method we have previously described.(18) The ACS dataset includes continuous sociodemographic variables at the county-level for percent of residents living in poverty, percent uninsured (not covered by either government and private insurance), concentration Black, concentration Latinx/Hispanic, percent with a bachelor’s degree or higher, and median household income.
Primary predictors of PrEP use
The presence of state PrEP-DAP was based on National Alliance of State and Territorial AIDS Directors (NASTAD) classification, with eligible programs being those that assist with at least one of the following: medication costs, copay costs, or clinical visit and laboratory testing costs and that were active by 2017; no states implemented PrEP-DAP programs in 2018.(20) State Medicaid expansion data were obtained from Kaiser Family Foundation (21) and used the definition of Medicaid expansion under the Affordable Care Act, in which people with annual incomes below 138% of the federal poverty level were made eligible for Medicaid. Eligible Medicaid expansion programs were implemented by 2016 or earlier; no states implemented Medicaid expansion in 2017–2018. States that began implementing PrEP-DAP programs or Medicaid expansion in 2019–2020 were classified as not providing the service for this analysis, because study outcomes represent PrEP use in 2018. Medicaid expansion may be important to PrEP use because states that have implemented it have seen increased health insurance coverage and improved population health outcomes across a wide variety of domains.(22) For multivariate analyses, we created a variable combining Medicaid expansion and PrEP-DAP due to collinearity between the two: 17 states had did not have Medicaid expansion or PrEP-DAP, 25 states had Medicaid expansion only, 1 state had PrEP-DAP only, and 7 states had Medicaid expansion and PrEP-DAP.
Analysis
Table 1 presents descriptive data regarding PrEP prevalence (PrEP users/100,000 population aged >13) and PrEP-to-need ratio (PnR, PrEP users/new HIV diagnosis aged >13) by urbanicity level, region and policy group. Available demographic data on individual PrEP users are presented by sex and by age group. Table 2 displays PrEP prevalence and PnR by county-level covariates. To account for large differences in population size across the 3,140 US counties (county population size range: 69 – 8,478,240) we created cut points so that each sociodemographic category in Table 2 represents approximately one quartile of the total US population.
Table 2.
PrEP users, prevalence, and PnR by population quartile of county covariates in the United
| PrEP users 1, 2, 3 N (%) | PrEP users per 100,000 p opulation (prevalence)° | New HIV Diagnoses per 100,000 population° | PrEP-to-Need Ratio(PnR)4 | |
|---|---|---|---|---|
| Black Concentration | ||||
| [0.0%, 3.2%) | 27380 (15) | 41.1 | 5.4 | 7.6 |
| [3.2%, 8.2%) | 48261 (26) | 71.2 | 11.0 | 6.5 |
| [8.2%, 18.5%) | 53657 (28) | 80.9 | 14.7 | 5.5 |
| [18.5%, 100.0%] | 59248 (31) | 87.8 | 25.7 | 3.4 |
| Latinx/Hispanic Concentration | ||||
| [0.0%, 5.0%) | 27043 (14) | 40.5 | 8.6 | 4.7 |
| [5.0%, 10.6%) | 41517 (22) | 61.6 | 13.7 | 4.5 |
| [10.6%, 25.1%) | 59466 (32) | 88.2 | 14.5 | 6.1 |
| [25.1%, 100.0%] | 60521 (32) | 90.8 | 20.1 | 4.5 |
| Percent bachelor’s degree or higher | ||||
| [0.0%, 22.7%) | 24468 (13) | 36.5 | 10.3 | 3.6 |
| (22.7%, 30.8%] | 36534 (19) | 55.4 | 16.4 | 3.4 |
| (30.8%, 37.3%] | 57602 (31) | 85.0 | 16.6 | 5.1 |
| [37.3%, 100.0%] | 69942 (37) | 103.3 | 13.7 | 7.6 |
| Median Household income | ||||
| [$0, $48968) | 15.4 | 3.0 | ||
| [$48968, $57647) | 16.0 | 4.0 | ||
| [$57647, $67964) | 14.5 | 5.2 | ||
| [$67964 or more] | 64265 (34) | 95.5 | 11.1 | 8.6 |
| Percent Poverty | ||||
| [0.0%, 10.9%) | 41897 (22) | 62.6 | 8.3 | 7.6 |
| [10.9%, 14.6%) | 45308 (24) | 67.9 | 12.0 | 5.7 |
| [14.6%, 17.1%) | 47785 (25) | 71.2 | 17.1 | 4.2 |
| [17.1%, 100.0%] | 53557 (28) | 79.3 | 19.5 | 4.1 |
| Percent Uninsured | ||||
| [0.0%, 7.0%) | 57261 (30) | 85.4 | 8.3 | 10.2 |
| [7.0%, 10.2%) | 42098 (22) | 62.8 | 10.5 | 6.0 |
| [10.2%, 12.9%) | 44186 (23) | 65.9 | 16.5 | 4.0 |
| [12.9%, 100.0%] | 45001 (24) | 67.1 | 21.6 | 3.1 |
aged 13 and above
PrEP users rounded to nearest integer
Numbers may not sum to total number of PrEP users due to rounding
Column percents may not total 100 due to rounding
Calculated as PrEP users/New HIV diagnoses
In Tables 3 and 4 we performed multilevel regressions on 2018 PrEP use data at the county level. Like descriptive analyses, we used two offset (denominator) variables in the regressions: per population (PrEP prevalence outcome, Table 3) and per new HIV diagnosis (PnR outcome, Table 4). In order to control for state and county-level covariates, generalized linear mixed effects models were used with Poisson distribution, log link, random intercepts for state and county, and a variance components covariance matrix for each random effect. For each outcome of interest, separate models assessed only county-level independent variables, county- and state-level variables together, and county-, state-, and region-level variables together. A pre-specified contrast assessed a linear trend across Medicaid expansion/PrEP-DAP presence. Rate ratios were computed by exponentiating model parameter estimates. Model fit was assessed by residual plots and deviance/df goodness of fit test. In Figure 1, Joinpoint regressions, a method we have detailed previously,(5) were performed to visualize trends of increase in PrEP prevalence and PnR from 2012 to 2018 for variables that were significant in the multivariable regression. For the PnR model shown in Table 4, some counties identified no new HIV diagnoses and were therefore excluded from the regression. To assess the impact of this, we conducted sensitivity analyses by using imputed values for these counties of 0.25, 0.5, and 1 new HIV case (Supplement 3).
Table 3.
Regressions of factors associated with PrEP Prevalence in the United States, 2018
Note: Generalized linear mixed effects model with a poisson distribution, log link, random intercepts for county and state, variance components covariance structure. Model outcome is PrEP use, with an offset of population size. RR - rate ratio; 95% CI - 95% confidence interval; LB - Lower Bound; UB - Upper bound.
Change in the rate ratio given each 5% higher proportion of the covariate, e.g. 5% higher uninsured
RESULTS
National Distribution of PrEP Use
In 2018, our dataset included a total of 188,546 PrEP users nationally. A sensitivity analysis that sought to account for missing data identified a national best estimate of all PrEP users to be 235,683, with range 188,546–269,351 (Supplement 1). The national prevalence of PrEP use was 70.3/100,000 population, and the national PnR was 4.9 (Table 1). Higher PrEP prevalence indicates more PrEP use relative to the population size and higher PnR indicates more PrEP use relative to new HIV diagnoses (a proxy for need). States with PrEP-DAP had nearly double the PrEP prevalence (100.6/100,000) and PnR (6.4) relative to states lacking such programs (51.9/100,000 and 3.9). Similarly, states that had expanded
Medicaid had higher PrEP prevalence and PnR (80.3/100,000 and 6.6) than states that had not (54.2/100,000 and 3.1). Urban areas had higher PrEP prevalence but only slightly higher PnR (Large central metropolitan areas 121.6/100,000 and 5.3) relative to nonurban areas (Noncore 27.6/100,000 and 5.1). The Northeast had the highest PrEP prevalence and PnR (106.3/100,000 and 8.5). The South had the lowest PnR (3.0), despite having similar PrEP prevalence to the Midwest. PrEP prevalence and PnR for each state are provided in Supplementary Table 2.
For males, PrEP prevalence and PnR (135.3/100,000 and 5.7) were higher than for females (8.7/100,000 and 1.6). PrEP prevalence and PnR were lowest in those under age 25 (51.5/100,000 and 3.3) and over age 54 (15.9/100,00 and 3.7) (Table 1).
PrEP Use by County Characteristics
PrEP prevalence and PnR were discordant when considering county-level race/ethnicity (Table 2). Counties in the quartile group with the highest concentration of Black residents had higher PrEP prevalence but lower PnR (87.8/100,000 and 3.4) than counties in the group with lowest concentration of Black residents (41.1/100,000 and 7.6). Similarly, counties in the quartile group with the highest concentration of Latinx residents had higher PrEP prevalence but moderately lower PnR (90.8/100,000 and 4.5) than counties in the group with the lowest concentration of Latinx residents (40.5/100,000 and 4.7).
For county-level education, median income, and insurance, PrEP prevalence and PnR were concordant (Table 2). Counties in the quartile with highest Bachelor’s degree attainment (103.3/100,000 and 7.6) or highest median income (95.5/100,000 and 8.6) had higher PrEP prevalence and PnR than those in the lowest (Bachelors attainment: 36.5/100,000 and 3.6; median income: 46.6/100,000 and 3.0). Counties in the quartile with highest proportion of uninsured residents had PrEP prevalence and PnR (67.1/100,000 and 3.1) lower than counties with the lowest proportion of uninsured residents (85.4/100,000 and 10.2).
Multilevel Analysis of the Association Between PrEP Use and County Characteristics and State-level Policies
A multilevel regression with an outcome of PrEP prevalence assessed region, state-level, and county-level covariate association with PrEP use per population (Table 3). Counties with higher concentrations of residents who were uninsured, Black, or Latinx, had moderately higher PrEP prevalence. This remained true after controlling for other factors such as urbanicity. Counties with higher education, as determined by proportion of residents holding a Bachelor’s degree, had higher levels of PrEP prevalence.
Relative to states without Medicaid expansion or PrEP DAP, states with either Medicaid expansion or PrEP DAP had 25% higher PrEP prevalence (rate ratio (RR): 1.25, 95% confidence interval (CI) 1.09, 1.45), and states with both Medicaid expansion and PrEP-DAP had 99% higher PrEP prevalence (RR: 1.99, 95%CI 1.60, 2.48). There was a significant linear trend across the policy groups (p<0.0001).
A multilevel regression model with an outcome of PnR (using new HIV diagnoses as the denominator) can be seen in Table 4. Among counties, each 5% higher proportion of Black residents was associated with a 5% lower PnR (RR: 0.95, 95%CI 0.93, 0.96). Similar levels of decrease in PnR were observed for counties with higher levels of uninsured residents or poverty. Conversely, a 5% increase in PnR was observed for counties with higher proportions of residents with a Bachelor’s degree (RR: 1.05, 95%CI 1.04, 1.07). Policy effects for PnR were similar in direction but attenuated relative to effects for PrEP prevalence, with a significant linear trend across policy groups (p=0.033).
A sensitivity analysis for the PnR model was conducted to assess the impact of including in the model counties that had PrEP prescriptions but no new HIV diagnoses in 2017. Under diverse scenarios, assuming either ¼ up to 1 new HIV diagnosis per missing county, the only substantial change regarded the variable of urbanicity. The association between urbanicity and PnR varied, which is expected because rural counties with small population are those most likely to have no new HIV diagnoses (Supplementary Table 3).
Annual Changes in PrEP Use
Joinpoint analyses estimated the annual percent changes for county-level variables that were significant in the multilevel PnR analysis (Figure 1). The quartile of counties with highest concentration of Black residents had a lower PnR in 2012 compared to quartiles with less concentration of Black residents. The disparity only attenuated slightly over time, and if current trends continue, would not be expected to change in the near future. Similar patterns were observed for quartiles of counties with higher concentration uninsured or with lower concentration of Bachelor’s degree holders: initial differences in PrEP use followed by trends indicating that disparities are not expected to be mitigated in the future unless there is a change in the trends. It is noteworthy that, although the relative level of disparities has not substantially changed over time, the absolute disparity in PrEP uptake has increased substantially over time.
DISCUSSION
In a dataset comprised of most PrEP users in the United States, we found that state-level policies supportive of PrEP access were associated with higher PrEP use. Adopting Medicaid expansion or PrEP-DAP alone were associated with 25% higher PrEP prevalence, but adopting both Medicaid expansion and PrEP-DAP was associated with 99% higher PrEP prevalence. Having both policies in place nearly doubled the observed number of PrEP users per population. This relation was observed in descriptive and multilevel analyses, and for outcomes of PrEP use per population and PrEP use per new HIV diagnoses. The association between policies supportive of PrEP access and PrEP use is encouraging, but the potential for residual confounding makes it challenging to ascertain causation. There is plausibility for a potential causal effect, because Medicaid expansion impacts the ability of persons to obtain insurance coverage, and non-expansion has previously been identified as a potential threat to achieving targets for PrEP scale-up.(23) Similarly, insurance coverage has been associated with higher PrEP use in a number of studies, with one finding four times higher PrEP use among insured patients.(24)
PrEP-DAP programs have previously been identified as potentially helping to promote PrEP uptake,(25) but to date no evidence has been presented regarding the impact of state PrEP-DAP programs. Studies of Drug Assistance Programs for HIV treatment (HIV-DAP) have found positive impacts when states allocate additional funds for the programs to expand their services,(26) and have identified key features of the most successful programs such as less restrictive eligibility criteria.(27) PrEP-DAP programs are implemented differently in different states, but can generally provide assistance with some combination of the following: mitigating the cost of the medication, mitigating the cost of seeking care (required clinical visits, lab testing), assisting with care seeking (finding a provider, visit scheduling, referrals), and wrap-around services that can include referral to other needed services such as mental health, substance abuse, or housing assistance. Our finding that such programs are associated with increased PrEP provision is encouraging and indicates that further research into these and other public policies to support PrEP is merited. For HIV treatment, the provision of such services increases retention in care and viral suppression outcomes.(26–28) Finding the optimal combination of PrEP-DAP services may be an important component of PrEP scale-up.
Counties with higher concentration of Black residents, higher levels of non-insured residents, or with lower levels of education had lower levels of PrEP relative to their need for PrEP as determined by their HIV epidemic burden. Given annual percent changes in PrEP and sustained disparities over time that do not appear to be shrinking, we do not expect to see these disparities mitigated in the absence of substantial intervention. These findings are in line with previous data finding that counties with higher concentration of Black residents or uninsured residents were less likely to have a clinic that prescribes PrEP.(29) At the individual level, both female sex and younger age groups had lower PrEP prevalence and PnR, in-line with our findings from 2017.(6) To address these disparities, well-resourced interventions and campaigns, grounded in evidence-based practice, will likely be required.
Differences in PrEP prevalence versus PrEP-to-need indicate the relevance of considering alternate metrics to the standard offset of population used in prevalence assessments. For instance, PrEP prevalence is four times higher in urban areas than in rural areas, yet there is little difference in PnR, indicating that higher levels of use in urban areas are in-line with higher need. PrEP prevalence in adjusted models was positively associated with higher concentrations of Black residents, indicating that counties with higher proportions of Black residents have higher uptake per person. Yet, when epidemic need was considered, these areas were observed to have lower PrEP use than was merited by their epidemic burden. Because interventions should scale in relation to their need, it is important to consider metrics that adjust for need such as PnR. Conversely, the more traditional metric of prevalence may be useful when determining the impact of interventions that do not target disparities but instead seek overall gains, such as public assistance programs like PrEP-DAP. Future assessments of the scale-up of health interventions should continue to explore metrics that consider differential levels of need across geography and population subgroups.
This analysis has several limitations. Associations are ecological and there is unmeasured confounding, and therefore causal relations cannot be assumed. We sought to control for potential confounders by conducting a multilevel analysis, but there are numerous potential confounders for which there are not national data available, and if confounding is at the state level even perfect individual-level data would not address this concern. The dataset does not include race/ethnicity at the individual level, a factor that we are seeking to address with future data analyses and public data releases. The dataset is missing data systematically, leading to bias. Non-inclusion of closed healthcare systems such as Kaiser and the Veteran Health Administration may particularly disadvantage the West region, in which Kaiser has a particularly large subscriber base.(13) The study also has a number of strengths. It is one of the first national policy analyses to use county-level PrEP data. It used both standard prevalence and need-adjusted metrics to explore health disparities and outcomes, with sensitivity analyses employed to challenge assumptions of the model where possible. To our knowledge, it is the first to assess the impact of structural factors on PrEP use. In numerous areas of health, policies are key determinants of outcomes. Despite structural factors being a key part of most behavioral health theories, the impact of these factors is rarely assessed; this study serves as an important first step for consideration of policies and PrEP use.
In assessing county-level data, we discovered disparities in how PrEP is being accessed and used. The analysis indicates the promise of state-level policies supportive of PrEP scale-up. Future research should seek to identify specific components of PrEP support policies that yield the best outcomes and conduct natural experiments to further adjust for potential residual confounding. If we are to end the HIV epidemic, it will require policy-level solutions that support individuals who want to proactively protect their health with the best available HIV prevention intervention. States should strongly consider facilitating the admirable health decisions of their residents to get on PrEP by expanding Medicaid and developing and optimizing PrEP-DAP programs.
Supplementary Material
FUNDING
This work was supported in part by the National Institute of Mental Health (R01MH114692) and by the National Institute of Allergy and Infectious Diseases (R01AI143875). Support also provided by the Adolescent Medicine Trials Network for HIV/AIDS Interventions (ATN, protocol 159) from the National Institutes of Health (U19HD089881), and by the Emory Center for AIDS Research (P30AI050409). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
CONFLICT OF INTEREST
Aaron Siegler and Patrick Sullivan are Investigators on a grant from Gilead Foundation, paid to their institution.
REFERENCES
- 1.Grant RM, Anderson PL, McMahan V, Liu A, Amico KR, Mehrotra M, et al. Uptake of pre-exposure prophylaxis, sexual practices, and HIV incidence in men and transgender women who have sex with men: a cohort study. The Lancet infectious diseases. 2014;14(9):820–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Marcus JL, Hurley LB, Hare CB, Nguyen DP, Phengrasamy T, Silverberg MJ, et al. Preexposure Prophylaxis for HIV Prevention in a Large Integrated Health Care System: Adherence, Renal Safety, and Discontinuation. J Acquir Immune Defic Syndr. 2016;73(5):540–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Jenness SM, Goodreau SM, Rosenberg E, Beylerian EN, Hoover KW, Smith DK, et al. Impact of the Centers for Disease Control’s HIV Preexposure Prophylaxis Guidelines for Men Who Have Sex With Men in the United States. The Journal of infectious diseases. 2016;214(12):1800–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Grulich AE, Guy R, Amin J, Jin F, Selvey C, Holden J, et al. Population-level effectiveness of rapid, targeted, high-coverage roll-out of HIV pre-exposure prophylaxis in men who have sex with men: the EPIC-NSW prospective cohort study. The lancet HIV. 2018;5(11):e629–e37. [DOI] [PubMed] [Google Scholar]
- 5.Sullivan PS, Mera Giler R, Mouhanna F, Pembleton E, Guest J, Jones J, et al. Trends in active prescriptions of emtricitabine/tenofovir disoproxil fumarate for pre-exposure prophylaxis against HIV infections, United States, 2012–2016. Annals of epidemiology. 2018;28(12):833–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Siegler AJ, Mouhanna F, Giler RM, Weiss K, Pembleton E, Guest J, et al. The prevalence of pre-exposure prophylaxis use and the pre-exposure prophylaxis-to-need ratio in the fourth quarter of 2017, United States. Annals of epidemiology. 2018;28(12):841–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Smith DK, Van Handel M, Wolitski RJ, Stryker JE, Hall HI, Prejean J, et al. Vital Signs: Estimated Percentages and Numbers of Adults with Indications for Preexposure Prophylaxis to Prevent HIV Acquisition–United States, 2015. MMWR Morbidity and mortality weekly report. 2015;64(46):1291–5. [DOI] [PubMed] [Google Scholar]
- 8.Finlayson T, Cha S, Xia M, Trujillo L, Denson D, Prejean J, et al. Changes in HIV Preexposure Prophylaxis Awareness and Use Among Men Who Have Sex with Men - 20 Urban Areas, 2014 and 2017. MMWR Morbidity and mortality weekly report. 2019;68(27):597–603. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Coy KC, Hazen RJ, Kirkham HS, Delpino A, Siegler AJ. Persistence on HIV preexposure prophylaxis medication over a 2-year period among a national sample of 7148 PrEP users, United States, 2015 to 2017. J Int AIDS Soc. 2019;22(2):e25252. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Golden MR, Lindquist S, Dombrowski JC. Public Health–Seattle & King County and Washington State Department of Health Preexposure Prophylaxis Implementation Guidelines, 2015. Sexually transmitted diseases. 2016;43(4):264–5. [DOI] [PubMed] [Google Scholar]
- 11.Sullivan PS, Mena L, Elopre L, Siegler AJ. Implementation Strategies to Increase PrEP Uptake in the South. Current HIV/AIDS reports. 2019;16(4):259–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Sullivan PS, Mouhanna F, Mera R, Pembleton E, Castel AD, Jaggi C, et al. Methods for county-level estimation of pre-exposure prophylaxis coverage and application to the U.S. Ending the HIV Epidemic jurisdictions. Annals of epidemiology. 2020. [DOI] [PubMed] [Google Scholar]
- 13.Kaiser Permanente. Who we are: Fast Facts 2019. [Available from: https://about.kaiserpermanente.org/who-we-are/fast-facts.
- 14.Manson S, Schroeder J, Van Riper D, Ruggles S. IPUMS National Historical Geographic Information System: Version 12.0 [Database]. Minneapolis: University of Minnesota. 2017;39. [Google Scholar]
- 15.Centers for Disease Control and Prevention. HIV Surveillance Report, 2017. 2019.
- 16.Liu AY, Cohen SE, Vittinghoff E, Anderson PL, Doblecki-Lewis S, Bacon O, et al. Preexposure Prophylaxis for HIV Infection Integrated With Municipal- and Community-Based Sexual Health Services. JAMA Intern Med. 2016;176(1):75–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Chan PA, Mena L, Patel R, Oldenburg CE, Beauchamps L, Perez-Brumer AG, et al. Retention in care outcomes for HIV pre-exposure prophylaxis implementation programmes among men who have sex with men in three US cities. J Int AIDS Soc. 2016;19(1):20903. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Siegler AJ, Bratcher A, Weiss KM. Geographic Access to Preexposure Prophylaxis Clinics Among Men Who Have Sex With Men in the United States. Am J Public Health. 2019;109(9):1216–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.National Center for Health Statistics. Urban Rural Classification Scheme for Counties [Available from: https://www.cdc.gov/nchs/data_access/urban_rural.htm.
- 20.NASTAD: National Alliance of State and Territorial AIDS Directors. PrEP Assistance Programs 2019. [Available from: https://www.nastad.org/prepcost-resources/prep-assistance-programs.
- 21.Kaiser Family Foundation. Current Status of State Medicaid Expansion Decisions 2019. [Available from: https://www.kff.org/health-reform/slide/current-status-of-the-medicaid-expansion-decision/.
- 22.Antonisse L, Garfield R, Rudowitz R, Artiga S. The effects of Medicaid expansion under the ACA: updated findings from a literature review. 2018.
- 23.Holtgrave DR, Greenwald R. A SWOT Analysis of the Updated National HIV/AIDS Strategy for the U.S., 2015–2020. AIDS Behav. 2016;20(1):1–6. [DOI] [PubMed] [Google Scholar]
- 24.Patel RR, Mena L, Nunn A, McBride T, Harrison LC, Oldenburg CE, et al. Impact of insurance coverage on utilization of pre-exposure prophylaxis for HIV prevention. PLoS One. 2017;12(5):e0178737. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Marcus JL, Volk JE, Pinder J, Liu AY, Bacon O, Hare CB, et al. Successful Implementation of HIV Preexposure Prophylaxis: Lessons Learned From Three Clinical Settings. Current HIV/AIDS reports. 2016;13(2):116–24. [DOI] [PubMed] [Google Scholar]
- 26.Hanna DB, Buchacz K, Gebo KA, Hessol NA, Horberg MA, Jacobson LP, et al. Association between U.S. state AIDS Drug Assistance Program (ADAP) features and HIV antiretroviral therapy initiation, 2001–2009. PLoS One. 2013;8(11):e78952. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Snider JT, Goldman DP, Rosenblatt L, Seekins D, Juday T, Sanchez Y, et al. The Impact of State AIDS Drug Assistance Policies on Clinical and Economic Outcomes of People With HIV. Med Care Res Rev. 2016;73(3):329–48. [DOI] [PubMed] [Google Scholar]
- 28.Mugavero MJ, Amico KR, Horn T, Thompson MA. The state of engagement in HIV care in the United States: from cascade to continuum to control. Clinical infectious diseases : an official publication of the Infectious Diseases Society of America. 2013;57(8):1164–71. [DOI] [PubMed] [Google Scholar]
- 29.Siegler AJ, Bratcher A, Weiss KM, Mouhanna F, Ahlschlager L, Sullivan PS. Location location location: an exploration of disparities in access to publicly listed pre-exposure prophylaxis clinics in the United States. Annals of epidemiology. 2018;28(12):858–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.