The US Ending the HIV Epidemic (EHE) launched in 2019 and is led by the US Department of Health and Human Services. EHE sought to reduce new HIV infections and scale up HIV prevention and treatment strategies to improve health outcomes for populations and communities most affected by HIV [1]. A total of 57 jurisdictions across 26 states, as well as Washington, DC, and San Juan, Puerto Rico, have been prioritized in the US EHE initiative [2]. Among these 26 states, 48 counties across 19 states were prioritized due to significant HIV burden, and 7 additional states were prioritized due to substantial occurrence of HIV in rural areas [2].

Pre-exposure prophylaxis (PrEP) is highly effective in preventing HIV transmission and remains the centerpiece of HIV prevention efforts in the United States, yet the scale-up of PrEP has led to stark disparities by race and ethnicity, limiting its impact [3]. PrEP is well tolerated, available in inexpensive generic forms, and prevents over 99% of HIV transmission when taken as directed. In numerous areas, including the United States [4] and New South Wales [5], scale-up of PrEP has been associated with reductions in new HIV transmission at the population level. Despite the spectacular promise of PrEP, less than half of persons in the United States indicated for PrEP were taking it in 2024 [6]. From 2012 to 2017, PrEP use increased by an average of 56% annually [7], followed by a more gradual annual growth of 18% from 2018 to 2021 [3]. The PrEP-to-need ratio (PnR) is a measure of PrEP use that compares the number of PrEP users in a group or geographic area (numerator) to the number of HIV diagnoses (denominator) [8]. This metric has been used to measure PrEP uptake among groups that would most benefit from it in numerous studies and has been used in other areas such as the US Centers for Disease Control and Prevention’s (CDC) recent examination of mpox vaccine provision [9]. PnR assessments have revealed lower PrEP use among Black and Latinx persons, younger persons, women, and persons in the US South relative to epidemic burden [3,10]. Additionally, modeling suggests that the greatest overall HIV epidemic reduction occurs in scenarios with equal PrEP provision relative to HIV burden across racial groups (PnR=1) [11]. Thus, increasing uptake of PrEP in areas with greatest need, as defined by higher HIV incidence, maximizes the number of new infections averted for a given level of PrEP use. National disparities in PrEP provision by race, however, have increased over time [12]. There is an urgent need to identify policies and programs that improve PrEP uptake and PnR.

Despite substantial scale-up of PrEP attributable to past efforts and policies, legal and program environments in which PrEP-eligible people live are highly heterogeneous, and evidence regarding policy impact is lacking. The United States has made significant efforts to scale up HIV PrEP through EHE Programs, with over US $1.5 billion invested in HIV prevention from 2020 to 2024 [13]. EHE funding has been granted to the EHE priority jurisdictions through the Health Resources and Services Administration and CDC mechanisms that allowed state and local health departments and community health centers to implement HIV prevention and treatment initiatives. Diverse programs have been implemented with this funding—ranging from social media messaging to medical records improvements, and from clinician training to telemedicine PrEP prescribing [14-16]—yet, no research program has sought to determine the comparative effectiveness and efficiency of these PrEP interventions on a population basis to guide future investments. Outside of HIV EHE efforts, PrEP uptake may also be influenced by state-level policy decisions and the availability of CDC grant programs such as grants that were designed to support PrEP demonstration projects. Both Medicaid expansion and the use of Pre-Exposure Prophylaxis Drug Assistance Programs (PrEP DAP) offer promising approaches that are associated with increased PrEP use that is also more aligned with epidemic need [17]. It is critical to identify the program elements that have contributed most to successful outcomes within each PrEP service.

National prescription, medical, and diagnosis datasets using data from commercial entities are robust and capture a majority of all US prescriptions with line-level, deidentified individual data [17]. These comprehensive datasets enable accurate, nationwide data on PrEP prescribing, which can be used to develop valuable insights on HIV prevention efforts when analyzed in tandem with data regarding PrEP policies and programs provision. Moreover, such aggregated datasets allow for data to be rapidly assessed regarding the uptake of newer PrEP formulations and modalities (eg, long-acting injectable [LAI] PrEP medications), providing information regarding the use and care characteristics of newer PrEP modalities relative to previously approved modalities.

To support public health jurisdictions in reducing new HIV transmission, the consistent availability of national HIV PrEP data is essential. Such data can be used to describe patterns of PrEP use and to understand which programs and policies are effective at facilitating PrEP uptake, particularly among groups that will most benefit from PrEP use. The PREMISE (PrEP Epidemiology, Modeling, and Surveillance) research program aims to (1) provide context regarding the ways in which PrEP is being brought to scale in the United States, such as the use characteristics of different PrEP modalities; (2) describe the impact of health policies and programs on PrEP use and PnR in the United States; and (3) assess the characteristics of these policies and programs, their impact on population health, and their cost-effectiveness (Figure 1).

The study population consists of patients who had claims between 2016 and 2024 for antiretroviral (ARV) medications approved by the Food and Drug Administration (FDA) for PrEP. Data on PrEP use are available from all 50 states, Washington, DC, and Puerto Rico. Objective 1 will use descriptive and exploratory assessments to contextualize PrEP scale-up. Objective 2 will compare trends in jurisdictions implementing a particular program versus a control group of jurisdictions not implementing that program. Selection of control jurisdictions will vary depending on the level of analysis: for instance, state-level analyses may include all states, whereas analyses at the county level may include counties that are matched by relevant characteristics such as HIV prevalence or county size. Data will be analyzed retrospectively from January 1, 2016, to December 31, 2024, and we will incorporate 4 future annual data updates for the years 2025 to 2028. This study period was selected because the PrEP dataset was considered by the data provider to be internally consistent in the way it was received from pharmacies over the time period, and allows for sufficient time before the launch of EHE programs in 2019. Objective 3 will model program effects for all EHE jurisdictions on future PrEP use based on an up to 20-year time horizon, expanding our team’s existing models for 6 EHE jurisdictions [18,19].

Our analyses will use varied analytic approaches to answer our multiple research questions. To assess changes in PrEP provision over time (objective 1), we will use descriptive and modeling studies. To address gaps in understanding of what programs and policies will best facilitate PrEP scale-up and improved efficiency in the form of PnR (objective 2), we will use quasi-experimental approaches. Many policies and access programs influence PrEP access and uptake among states and over time. Therefore, we will use natural experiments to improve understanding of the impact on PrEP use by collecting information from jurisdictions regarding which programs were conducted and the dates of program operation. We will collect data regarding program characteristics and financial investment to improve our ability to assess causal relationships between different policies or programs and PrEP outcomes. Critical to this design is that, unlike randomized trials, it allows for assessment of health program outcomes with data from programs that were brought to scale outside of research settings by health jurisdictions. Such data will more accurately represent how health interventions are actualized in the United States. Objective 3 will use a mathematical modeling approach that is detailed below. For this paper, we followed the RECORD (Reporting of Studies Conducted Using Observational Routinely Collected Data) checklist (Multimedia Appendix 1) of items to be reported in observational studies using routinely collected health data [20].

Our inquiry will be grounded in a legal implementation framework that we have previously proposed to facilitate the development of health policies to optimally address infectious diseases (Figure 2) [21]. In this framework, now adapted to this research program, we posit law as a causal factor influencing HIV transmission. Here, we define law broadly, as sets of rules and regulations that are enforced by the government [22]. As seen in the model, law is enacted through institutions and can influence health positively through constructs such as insurance coverage of an otherwise unaffordable service, or negatively through adverse changes in coverage programs such as Medicaid. Other mechanisms are indirect; for example, wraparound services may assist clients in gaining stable housing, which may in turn facilitate medication adherence. Given this theoretical framework, changes in laws affecting PrEP provision have the potential to have substantial and long-lasting effects on vulnerability to HIV acquisition, including the provision of health services relative to needs for those services. Laws can be either a public health tool (eg, PrEP DAP programs) or a public health threat (eg, judicial rulings that may prevent the US Preventive Service Taskforce recommendations on PrEP from being binding for insurance coverage without cost sharing). Our research acknowledges the more proximal causes of persons taking or adhering to PrEP (eg, an individual’s perceived benefits of taking PrEP) and also focuses on programs and policies that shape and provide context for these decisions. We study programs and policies because (1) they are amenable to change, (2) they are measurable, and (3) there is plausibility that policies and programs can influence PrEP uptake. Relevant data sources will include legal coding and surveys to quantify programs and policies, census data to describe neighborhood factors, and PrEP claims databases to describe use and sociodemographic factors (eg, race, ethnicity, sex, and age). Due to the inherent limitations of our data sources, we cannot measure all components of the framework; instead, we will leverage it to inform the selection of data that are available.

This program of research will use three main data sources (see Table 1 for a summary of the variables of interest and covariates): (1) data from surveys of health jurisdictions and legal data extracted by the National Alliance of State and Territorial AIDS Directors (NASTAD), (2) a national PrEP database from a third-party data provider (prescription, procedure, and diagnosis claims datasets), and (3) US Census data. Other data sources as they become available will be included in analyses.

We will conduct a survey of health jurisdictions to understand which types of programs were provided in each jurisdiction over time and the degree of resources committed to implementing each program to obtain program data; these data will inform objective 2 analyses. Table 2 details the primary program types that we will assess, representing an array of program settings and types ranging from clinician training to electronic medical records changes to direct financial coverage of PrEP services. We will also assess community change mechanisms such as messaging campaigns and community-based organization-led PrEP outreach. Each of these mechanisms might feasibly impact PrEP use, so generating evidence on the relative impact of each program will be critical to guide future investments. The full text of the jurisdictional survey can be seen in Multimedia Appendix 2. Using lists of health jurisdiction members of NASTAD, we sent the survey to program staff representatives for each of the 50 states and the 8 jurisdictions that receive EHE funding independently from their state. Our instructions clarify that completing the survey may require the knowledge of multiple jurisdictional team members, and the survey is set up to facilitate this with a link that can be accessed by multiple respondents. Data collection launched in May 2025, accompanied by a webinar offered to all health jurisdictions to introduce the survey. We plan to complete data collection in the fall of 2025. To maximize the survey response rate, we will conduct outreach activities, including periodic email reminders, hosting open question and answer sessions, and individual support calls to engage jurisdictions to assist their survey completion.

We will also gather information on the passage and implementation of policies regarding PrEP at the state level through a systematic review of legislative records in existing state public health statutes and the legal database Westlaw. Data will be extracted, if available, regarding program types, program start or end dates, funding agencies, jurisdictions implementing the program, the program implementers, and the program financial expenditures. To improve data quality, we will optimize interrater reliability through clear definitions of domains, use raters who are familiar with the content area, and provide thorough training for raters through practice-based learning sessions. Validation of the dataset will be accomplished through conducting parallel extraction of independent coders to determine intercoder agreement for a proportion of the dataset.

We have obtained a national PrEP database through a contract with a third-party data provider—a health care company that aggregates national health data from clinics, providers, and pharmacies. This database includes prescription claim data, procedure claim data, and diagnosis claim data. PrEP medications can be billed through either pharmacy or medical benefits, depending on how the medication is dispensed (oral tablet or injection, pharmacy, hospital or clinic) and the requirements of the insurance plan. The diagnosis claim dataset contains medical conditions diagnosed using either the ICD-9-CM (International Classification of Diseases, Ninth Edition, Clinical Modification) or ICD-10-CM (International Classification of Diseases, Tenth Edition, Clinical Modification). For each deidentified person included in the database, the following variables are available: prescribed medication, date of each fill, number of pills or injections, sex, age, payer (Medicaid and private health insurance), and ZIP3 (the first 3 digits of a ZIP code). We will seek more specific user locations through both data crosswalks to the county level and providers’ practice location addresses. On June 18, 2025, lenacapavir was approved by the FDA as a twice-yearly LAI PrEP [23], and during the research period, we expect additional medications to be approved by the FDA for PrEP. Therefore, our data provider contract includes access to prescription data for lenacapavir and other potential future modalities.

We will gather publicly available data about jurisdictions (eg, characteristics and demographic information) from the US Census Bureau website. Using the Census’s search and filtering functions, we will extract relevant data at the appropriate geographic level (eg, state or county). These data will include distributions of educational attainment, poverty status, insurance coverage, and race and ethnicity composition, each of which will be used to examine jurisdictional differences and their associations with PrEP uptake.

Other potential sources of data may be incorporated into analyses as they become available over time.

This study involves secondary data analysis of fully deidentified prescription claims, procedure claims, and diagnosis code data, as well as a survey of US health jurisdictions to gather information on PrEP-related funding and program implementation. There was no human subjects enrollment or interaction, and no collection or use of any personal identifiable information. As such, this study does not constitute human subjects research or clinical investigation as defined in the federal regulations and was not subjected to institutional review board (IRB) review [55]. Investigators completed a worksheet provided by the Emory IRB, documenting that IRB review was not required. The study claims data are accessed by Emory under a contract and data use agreement with a third-party provider that specifies appropriate storage and data security procedures. As is normative for this type of claims data, the contract does not allow for external sharing of line-level data by Emory. However, other investigators can access the data directly through the data provider, ICON plc.

From 2016 to 2024, there were more than 20 million claims of medications approved by the FDA for PrEP: FTC/TDF, FTC/TAF, and CAB (see Figure 4). From these prescriptions, we excluded about 6 million claims for coinciding periods that included triple ARV drug regimens (exclusion criterion 1). We then merged the diagnosis claims into the dataset to allow for further determination. We confirmed 11 million PrEP prescription claims that were connected to individuals who had no HIV diagnosis, no PEP diagnosis, and had never used a triple ARV regimen (Group 1 certified PrEP claims). For the remaining 3 million claims, we used clinical guidance for possible other uses of the drugs to determine whether a prescription was for PrEP (Group 2 certified PrEP claims) or for an alternative use. We determined 700,000 of these claims to be not for PrEP use: 6000 claims for PrEP medications that occurred within 3 days of a PEP diagnosis (exclusion criterion 2), 447,000 claims from individuals with 2 or more HIV diagnosis codes (exclusion criterion 3), 247,000 claims from individuals with only one HIV diagnosis, and 3 or more claims for triple ARV regimens (exclusion criterion 3). About 25,000 claims for PrEP medications occurred prior to FDA medication approval, 500 claims for individuals under the age of 12 years at the time of prescription, and 26,000 claims did not reflect standard PrEP drug strengths (exclusion criterion 4). This led to a final prescription dataset with over 1 million cumulative patients having initiated PrEP and over 13 million separate PrEP prescriptions.

For medical benefits claims, there were about 34,000 procedure claims for PrEP provided in a clinical setting (see Figure 5). From these procedure claims, we excluded about 600 claims for coinciding periods that included triple ARV drug regimens (exclusion criterion 1). Using the same diagnosis claims dataset as described above, we confirmed 25,000 PrEP procedure claims for individuals who had no HIV diagnosis, had no PEP diagnosis, and had never used a triple ARV regimen (Group 1 certified PrEP claims). For the remaining 9000 claims, using the same clinical guidance as described above to determine whether a procedure was for PrEP (Group 2 certified PrEP claims), we determined 2200 of these claims to be not for PrEP use: 1670 claims from individuals with 2 or more HIV diagnosis codes (exclusion criterion 3), with the remaining claims from individuals with only 1 HIV diagnosis and 3 or more claims for triple ARV regimens (exclusion criterion 3), or PrEP medications that occurred within 3 days of a PEP diagnosis (exclusion criterion 2). About 9000 procedure claims for PrEP were duplicates. Under exclusion criterion 4, there was 1 claim for an individual under the age of 12 years at the time of procedure and about 100 claims for LAI PrEP with the drug name “cabotegravir” listed but an incorrect HCPCS level two J code. This led to a final procedure dataset with over 6000 cumulative patients having initiated PrEP through medical benefits with about 22,500 procedure claims for LAI PrEP and about 400 procedure claims for oral PrEP.

By developing an algorithm to determine PrEP use from other prescribing of ARV medication, this early analysis provides outcome data for all subsequent components of the PREMISE project. For objective 1, these data will be used to inform our descriptive and longitudinal analyses of changes in PrEP use over time. To accomplish objective 2, the data will be used as an outcome variable, allowing for the determination of the impact of PrEP policies and programs. To guide objective 3, these data will be used to inform the modeling of the future impacts and cost-effectiveness of different programs and policies.

Extraction of PrEP prescription and procedure claims was completed as part of objective 1 in 2025. Following this critical component, the planned analyses listed under objectives 1-3 in Table 3 will be conducted. This timeline provides detailed information about the planned analyses for the grant, although we may add additional analyses as new products (eg, long-acting lenacapavir) come to market or new programs, policies, or their cessation occur over time.

To reach EHE targets, it is crucial to have consistent national HIV PrEP data to describe patterns of use and to understand which programs and policies are effective at facilitating PrEP uptake. In PREMISE, we propose to build on our history of developing public PrEP data resources and innovative metrics of PrEP use. The research program comes from a collaborative and interdisciplinary group that will use national PrEP data, legal coding, and policy coding to inform the US HIV epidemic response. Prior studies internationally assessing PrEP policies have largely focused on whether PrEP is legally accessible and covered by insurance or public programs [56,57], which may not fully reflect the effectiveness of PrEP implementation. There is a need to understand the impact of policies and programs that support and promote PrEP as a public health good.

The aggregated medical claims dataset that provides PrEP use outcome data for all study aims contains a majority of claims in the United States but is subject to a number of limitations. First, it does not include prescriptions from closed health care systems, including the Veterans Affairs. Data must therefore be interpreted with caution for areas with high numbers of persons enrolled in closed health care systems, such as Kaiser Permanente users. Second, PrEP use in clinical trials and other informal channels may not be captured by the claims dataset. A limitation of the quasi-experimental approach is the lack of randomization to determine the effects of programs and policies, a limitation inherent to such assessments. We will minimize the impact of this limitation by using analytic designs that carefully adjust for secular trends and will also explore in our findings other considerations for establishing a causal relationship, including effect size, dose-response, and plausibility of the causal mechanism (eg, some programs might be anticipated to have a higher impact on care initiation than retention in care). Another limitation of the quasi-experimental approach is that the COVID-19 pandemic impacted PrEP use during the epidemic period [58-60]; this could influence the outcomes of analyses. To address this challenge, we will conduct a series of visual and quantitative sensitivity analyses for each program or policy that considers the pandemic period and its impact on results. There is a dependency in identifying programs with positive effects and modeling of these effects. We anticipate that some of the funds invested in EHE will have a quantifiable impact on PrEP use and care retention, but if we find no impact from any programs, we will conduct modeling informed by other data, such as those from clinical trials and observational studies. Another limitation is that funding for EHE programs could be discontinued, but based on our power calculations, we still will have sufficient data to perform the program and policy analyses, and evidence of success may become critical to encouraging funding reauthorization. Another potential challenge is the possibility of new medication approvals; we have discussed this with our aggregated data provider, and the contract will include all current and future PrEP medication data. Lastly, while the general structure of the algorithm may be applicable to other datasets, it may not be fully transferable due to potential variations in data structure, coding practices, and available variables across different data providers.

HIV PrEP has the potential to contribute substantially to ending the HIV epidemic, and increasing PrEP coverage to persons at substantial risk of transmission is critical to deliver on the promise of the intervention. Diverse programs have been launched to support increased PrEP use and retention in care, and systematic data are needed to understand the impact of these programs in real-world implementation settings such as health departments. The analyses proposed in this protocol will fill this gap, providing critical information regarding program impact. We anticipate that such data will facilitate increased investment in prevention efforts by providing an evidence base to understand the relative and absolute impacts of PrEP promotion interventions.