Current Work

This effort builds off a NEEMA/CAMP 1.0 project that estimated the county-level population of MSM and has proven to be very useful in DSTDP and DHAP work to estimate STI/HIV rates and disparities, as well as sizes of populations indicated for HIV pre-exposure prophylaxis. Aside from MSM, sub-national estimates of the sizes of sexually active adults and adolescents do not exist, but would help “complete” our local understanding of disease burden for STI, particularly given low availability of sex-of-sex-partner data aside from syphilis, and HIV. Furthermore, such estimates would facilitate relative comparisons of syphilis and HIV burden between MSM and other populations. This project will create analogous denominators for sexually-active men who have sex with women, women who have sex with men and adolescents at the county level, by race/ethnicity, and re-evaluate our previous approach to estimating MSM denominators.

Project Objectives or Major Research Questions

This project will seek to develop state and county-level estimates for the population of men who have sex with women (exclusively), women who have sex with men, and sexually-active adolescent sexual minority males (ASMM), overall and by race/ethnicity.  These denominator estimates will be usable to generate estimates of disease rates in the relevant populations.

This project follows up on our previous efforts to estimate the population attributable fraction of CT/GC on HIV incidence among MSM and the possible impact of screening and treatment of CT/GC on HIV incidence. For this analysis, we will estimate the cost-effectiveness of different CT/GC screening scenarios with an outcome of HIV-related QALYs.

Our approach will be to use an agent-based modeling framework (EpiModelHIV) to model the impact of different levels of CT/GC screening coverage on HIV incidence. The model will be used to obtain the necessary inputs for the cost-effectiveness analyses:

  •     Number of new HIV infections per year
  •     Number of new CT/GC infections per year
  •     Number of CT/GC infections treated
  •     Number of CT/GC screenings
Screening Scenarios

Anatomical site of screening: Rectal CT/GC infection is estimated to confer higher risk of HIV acquisition compared to urethral infection. Our current model parameterization assumes that rectal and urethral infections confer the same increased risk of onward HIV transmission. All CT/GC screening events are assumed to include screening for urethral infections, with only a subset including rectal screening. We will estimate the cost-effectiveness of increasing the proportion of screening events that include both urethral and rectal screening.

Screening by serostatus: CT/GC screening coverage is higher among men living with HIV compared to those who are HIV-. We will examine the cost-effectiveness of increasing CT/GC screening in each of these populations.

Behavioral indications for screening: CDC guidelines recommend more frequent CT/GC screening for men who are at high risk for sexually transmitted infections. We will estimate the cost-effectiveness of increasing the frequency of CT/GC screening among men at high risk (based on number of sexual partners in the past 6 months) compared to increasing the coverage of CT/GC screening among all sexually active MSM.

Project Goals: The federal initiative to End the HIV Epidemic in America includes responding to each newly diagnosed case of HIV with interventions targeted at the index case and sexual contacts, who may be either infected or uninfected. In this project, we will build on our robust stochastic network model for transmission of HIV to estimate the population-level impact of identifying, HIV screening, and linking their sexual partners to ART (if infected) or PrEP (if uninfected). This study would directly support efforts to address the fourth pillar of the EHE plan, “Respond,” that seeks to rapidly detect new clusters of HIV infections.

General Methodology: This project would build directly off of NEEMA 1.0 efforts completed by investigators in the EpiModel Research Lab supported by CAMP.1–3 The goal of this extension study is to explore the impact of a comprehensive HIV prevention and treatment intervention. The model will be for a partner notification intervention that identifies sexual partners of men who have sex with men (MSM) with newly diagnosed infection, or MSM who reengage in care with a detectable viral load, for HIV testing. Partners who screen positive will be linked to HIV treatment, while partners who screen negative will be linked to HIV PrEP. Following our approach to modeling bacterial STI expedited partner therapy,4 we will vary both the proportion of all partners (since the last negative test or in a given timeframe) reached as well as the proportion of reached partners who engage in ART or PrEP once identified. We will also evaluate the synergistic effects of these partner-targeted interventions along with existing individual-level interventions (e.g., increasing HIV screening frequency overall).

Project goals: In 2018, 24% of new HIV infections in the US were attributed to heterosexual contact and 19% of new infection were among women. A disproportionate number of these new infections were clustered in the Southeastern US and among women of color. Reducing the number of incident infections in this population will be necessary to reduce racial disparities in HIV and to reach EHE goals. This project will estimate the race-and-sex-specific changes in viral suppression and PrEP uptake required to reduce HIV incidence by 90% over ten years. This study will directly address the “Treatment” and “Prevention” pillars of the EHE and quantify their relative impact in terms of sex-and-race-specific reductions in incidence.

Methodology: For this project we will expand upon NEEMA 1.0 efforts and an NIH funded research project to integrate two agent-based stochastic network models of HIV transmission, one among MSM and one among heterosexuals. The goal of this integration is to model the impact of increasing viral suppression and PrEP uptake among heterosexuals while taking account of the larger population that defines the epidemic context of incident transmissions.  The model will include three race groups (Non-Hispanic Black, Hispanic/Latinx, and other), and heterosexual, MSM and bisexual male populations. The goal will be to determine the level of viral suppression and PrEP coverage among heterosexuals of each race groups that will be needed in order to reduce incidence by 90% within each group. We will vary retention on ART for diagnosed individuals and PrEP initiation for those that meet the CDC guidelines for PrEP. For this study behavior within the MSM population will be held constant in order to focus on the independent contributions of changes in the continuum of care for heterosexuals on incidence among heterosexuals.

Data sources: The data for the heterosexual population will come from the NSFG.  If possible, we will obtain the region indicators for the NSFG sample and focus our analysis initially on the Southeastern region. In either case we will reweight the NSFG to reflect the demographic distribution of the Southeastern region as defined by the CDC.  Parameters for the MSM with be calculated from ARTNet, a national study of MSM behavior used to parameterize agent-based models during FY5 of CAMP 1.0. We will use results from the prior NEEMA 1.0 denominators study in conjunction with reported female partners by MSM in the ARTNet study and males reports of both male and female partners in the NSFG to estimate the size of the bisexual male bridge population.

Note that this concept is still under development by academic leads and CDC partners; what follows is an excerpt from recent correspondence to CDC and contains a series of options (not included here for the sake of brevity and simplicity) for consideration by CDC which is awaiting feedback.

National estimates of prevalence of injection drug use (IDU) are currently limited to self-report among respondents to household surveys. Several indirect indications of IDU behavior suggest prevalence is increasing, but this is difficult to measure directly due to inadequacies of the survey data. Indirect indications of increasing IDU behavior include increased drug overdose mortality rates, increases in observed and estimated acute HCV infections, geographically isolated increases in IDU-related HIV infections, and increases in other infections that are associated with IDU, e.g. infectious endocarditis and osteomyelitis. We propose to use estimates of indirect indications of IDU behavior in combination with estimated conditional probabilities of having these outcomes among persons who inject drugs (PWID) to estimate PWID population size.

The proposed method is similar to the method currently used to estimate the multiplier for acute hepatitis infections (Klevens et al., AJPH, 2014). For acute HCV, a starting number of reported cases to surveillance is inflated by available estimates of the conditional probabilities of being symptomatic given infection, seeking care given symptoms, and being reported to surveillance given care-seeking behavior (together these 3 quantities estimate the number of infections not reported to surveillance). Here, we propose an analogous multiplier method that begins with surveillance data for indirect IDU indications, and adjust these with survey data and estimates from literature review to estimate PWID population size. In contrast to the case of acute HCV, because the starting surveillance quantity is a signal of IDU, rather than a direct PWID number, several multiplier estimates may be used to triangulate on the true population quantity.

Proposals of how this might work for three indirect indicators of IDU – drug overdose deaths, substance abuse treatment, and skin and soft tissue infections are outlined separately. One important consideration going forward is data availability. Ideally, we would need data for each strategy considered, stratified by basic demographic characteristics such as sex, age, and race. We would ideally also work this exercise through for multiple indicators so they can be compared. Reasonable ranges for each input will be used for modeling to enable estimated ranges for PWID population size. Given limitations in data quality, we anticipate some inputs will need to be creatively estimated and are seeking CDC’s feedback on both the proposed method and possible data sources for this method.

During NEEMA 1.0, we developed teen-SPARC, an Excel tool designed to help public health jurisdictions and other entities explore the potential impact of behavior change on reducing sexually transmitted infections (STIs) among the sexually active high-school-attending adolescent population. This tool has been parameterized by default using national data, but is designed to be tailored to other jurisdictions, particularly those that participate in YRBSS. Recent dissemination activities include a well-attended webinar and a case study paper under review demonstrating the use of this tool and its possible applications.

We propose additional activities to enhance the uptake and use of teen-SPARC. In Year 1, we will work to build the teen-SPARC user community in multiple ways.  We will begin by adding user registration and visitation/download metrics to the teen-SPARC website that allow for us to monitor usage. We will reach out to all webinar attendees to determine ongoing usage and identify means to support that usage.  We will develop a user listserv that provides a discussion forum for users. We will present at multiple conferences, through workshops and/or scientific presentation sessions. Throughout, we will query current and potential users about features to improve or add to teen-SPARC.  We will summarize progress on all steps in an end-of-year report to CDC. All of these steps will feed into our efforts to expand and enhance teen-SPARC in Y2.

This effort builds off a NEEMA/CAMP 1.0 project that estimated the county-level population of MSM and has proven to be very useful in DSTDP and DHAP work to estimate STI/HIV rates and disparities, as well as sizes of populations indicated for HIV pre-exposure prophylaxis. Aside from MSM, sub-national estimates of the sizes of sexually active adults and adolescents do not exist, but would help “complete” our local understanding of disease burden for STI, particularly given low availability of sex-of-sex-partner data aside from syphilis, and HIV. Furthermore, such estimates would facilitate relative comparisons of syphilis and HIV burden between MSM and other populations. This project will create analogous denominators for sexually-active men who have sex with women, women who have sex with men and adolescents at the county level, by race/ethnicity, and re-evaluate our previous approach to estimating MSM denominators.

Project Objectives or Major Research Questions

This project will seek to develop state and county-level estimates for the population of men who have sex with women (exclusively), women who have sex with men, and sexually-active adolescent sexual minority males (ASMM), overall and by race/ethnicity.  These denominator estimates will be usable to generate estimates of disease rates in the relevant populations.

This project follows up on our previous efforts to estimate the population attributable fraction of CT/GC on HIV incidence among MSM and the possible impact of screening and treatment of CT/GC on HIV incidence. For this analysis, we will estimate the cost-effectiveness of different CT/GC screening scenarios with an outcome of HIV-related QALYs.

Our approach will be to use an agent-based modeling framework (EpiModelHIV) to model the impact of different levels of CT/GC screening coverage on HIV incidence. The model will be used to obtain the necessary inputs for the cost-effectiveness analyses:

  •     Number of new HIV infections per year
  •     Number of new CT/GC infections per year
  •     Number of CT/GC infections treated
  •     Number of CT/GC screenings
Screening Scenarios

Anatomical site of screening: Rectal CT/GC infection is estimated to confer higher risk of HIV acquisition compared to urethral infection. Our current model parameterization assumes that rectal and urethral infections confer the same increased risk of onward HIV transmission. All CT/GC screening events are assumed to include screening for urethral infections, with only a subset including rectal screening. We will estimate the cost-effectiveness of increasing the proportion of screening events that include both urethral and rectal screening.

Screening by serostatus: CT/GC screening coverage is higher among men living with HIV compared to those who are HIV-. We will examine the cost-effectiveness of increasing CT/GC screening in each of these populations.

Behavioral indications for screening: CDC guidelines recommend more frequent CT/GC screening for men who are at high risk for sexually transmitted infections. We will estimate the cost-effectiveness of increasing the frequency of CT/GC screening among men at high risk (based on number of sexual partners in the past 6 months) compared to increasing the coverage of CT/GC screening among all sexually active MSM.

Note that this concept is still under development by academic leads and CDC partners; what follows is an excerpt from recent correspondence to CDC and contains a series of options (not included here for the sake of brevity and simplicity) for consideration by CDC which is awaiting feedback.

National estimates of prevalence of injection drug use (IDU) are currently limited to self-report among respondents to household surveys. Several indirect indications of IDU behavior suggest prevalence is increasing, but this is difficult to measure directly due to inadequacies of the survey data. Indirect indications of increasing IDU behavior include increased drug overdose mortality rates, increases in observed and estimated acute HCV infections, geographically isolated increases in IDU-related HIV infections, and increases in other infections that are associated with IDU, e.g. infectious endocarditis and osteomyelitis. We propose to use estimates of indirect indications of IDU behavior in combination with estimated conditional probabilities of having these outcomes among persons who inject drugs (PWID) to estimate PWID population size.

The proposed method is similar to the method currently used to estimate the multiplier for acute hepatitis infections (Klevens et al., AJPH, 2014). For acute HCV, a starting number of reported cases to surveillance is inflated by available estimates of the conditional probabilities of being symptomatic given infection, seeking care given symptoms, and being reported to surveillance given care-seeking behavior (together these 3 quantities estimate the number of infections not reported to surveillance). Here, we propose an analogous multiplier method that begins with surveillance data for indirect IDU indications, and adjust these with survey data and estimates from literature review to estimate PWID population size. In contrast to the case of acute HCV, because the starting surveillance quantity is a signal of IDU, rather than a direct PWID number, several multiplier estimates may be used to triangulate on the true population quantity.

Proposals of how this might work for three indirect indicators of IDU – drug overdose deaths, substance abuse treatment, and skin and soft tissue infections are outlined separately. One important consideration going forward is data availability. Ideally, we would need data for each strategy considered, stratified by basic demographic characteristics such as sex, age, and race. We would ideally also work this exercise through for multiple indicators so they can be compared. Reasonable ranges for each input will be used for modeling to enable estimated ranges for PWID population size. Given limitations in data quality, we anticipate some inputs will need to be creatively estimated and are seeking CDC’s feedback on both the proposed method and possible data sources for this method.

The goal of this modeling project is to build upon a model developed as part of NEEMA/CAMP 1.0 which assessed the impact of changes in risk behaviors among adolescents by race. This analysis demonstrated significant reductions in risk behaviors (proportion of adolescents ever having sex, numbers of partners, those reporting being currently sexually active) among Black and Hispanic adolescents. On the other hand, it demonstrated decreased condom use among White adolescents. The initial project focused on STI levels (gonorrhea and chlamydia) whereas this follow-on project will focus on pregnancies expected over ten years in the absence of those changes vs. with the changes implemented.  The analyses will focus on heterosexually active adolescents. We will project outcomes specific to race/ethnicity groups, and work with DASH partners to estimate the cost or savings associated with each outcome.

Major questions or issues include: How many pregnancies have been averted by the behavioral changes documented among adolescents over a 10-year period? What is the estimated cost saved at the societal level by these averted cases?

During NEEMA 1.0, we developed teen-SPARC, an Excel tool designed to help public health jurisdictions and other entities explore the potential impact of behavior change on reducing sexually transmitted infections (STIs) among the sexually active high-school-attending adolescent population. This tool has been parameterized by default using national data, but is designed to be tailored to other jurisdictions, particularly those that participate in YRBSS. Recent dissemination activities include a well-attended webinar and a case study paper under review demonstrating the use of this tool and its possible applications.

We propose additional activities to enhance the uptake and use of teen-SPARC. In Year 1, we will work to build the teen-SPARC user community in multiple ways.  We will begin by adding user registration and visitation/download metrics to the teen-SPARC website that allow for us to monitor usage. We will reach out to all webinar attendees to determine ongoing usage and identify means to support that usage.  We will develop a user listserv that provides a discussion forum for users. We will present at multiple conferences, through workshops and/or scientific presentation sessions. Throughout, we will query current and potential users about features to improve or add to teen-SPARC.  We will summarize progress on all steps in an end-of-year report to CDC. All of these steps will feed into our efforts to expand and enhance teen-SPARC in Y2.

The goal of this modeling project is to build upon a model developed as part of NEEMA/CAMP 1.0 which assessed the impact of changes in risk behaviors among adolescents by race. This analysis demonstrated significant reductions in risk behaviors (proportion of adolescents ever having sex, numbers of partners, those reporting being currently sexually active) among Black and Hispanic adolescents. On the other hand, it demonstrated decreased condom use among White adolescents. The initial project focused on STI levels (gonorrhea and chlamydia) whereas this follow-on project will focus on pregnancies expected over ten years in the absence of those changes vs. with the changes implemented.  The analyses will focus on heterosexually active adolescents. We will project outcomes specific to race/ethnicity groups, and work with DASH partners to estimate the cost or savings associated with each outcome.

Major questions or issues include: How many pregnancies have been averted by the behavioral changes documented among adolescents over a 10-year period? What is the estimated cost saved at the societal level by these averted cases?

Widespread hepatitis A virus (HAV) outbreaks are ongoing in 30 states due to person to person transmission, resulting in over 27,000 reported cases, and 275 deaths. Key risk groups for these outbreaks are drug using populations, homeless individuals, and men who have sex with men (MSM), and local health departments urgently require information on target vaccination levels to prevent outbreaks and achieve herd immunity. Previous studies estimating target vaccine coverage required for preventing HAV outbreaks have been limited to endemic transmission and universal vaccination as opposed to specific at-risk populations. No studies have examined target vaccine coverage among homeless or drug using populations, with only one study for MSM using Australia data from the 1990s (Regan et al. Epidemiol Infect 2016). Recent outbreaks in San Diego and Louisville, collaboratively worked upon by the local health departments, CDC, and the CAMP 2.0 UCSD team, offer a singular data and modeling opportunity for addressing this question among homeless and/or drug using populations.
Year 1 Objective: Estimate the population immunity threshold required to stop a hepatitis A virus (HAV) outbreak associated with person to person transmission among individuals at high risk in San Diego, CA and Louisville, KY.
Year 1 method:  Estimates of the basic reproduction number (R0) and herd immunity threshold among individuals at high risk (defined as a singular homeless or drug using population) in each geographical setting will be assessed using two methods: 1) using a dynamic deterministic compartmental transmission model of HAV among homeless or drug using populations calibrated to surveillance data, and 2) using statistical estimation methods based on surveillance data. Two separate papers will be generated (one for each geographical setting) which will explore: R0, herd immunity threshold, and the estimated impact of vaccination efforts on the numbers of HAV infections, hospitalizations, and deaths prevented among homeless or drug using populations.

Eliminating hepatitis B virus (HBV) requires reducing transmission. While the Advisory Committee on Immunization Practices currently recommends vaccination against hepatitis B for all infants and adults with risk factors for infection, complete series coverage among adults remains low.  In a Vaccine Safety Datalink study of more than 88,000 adult hepatitis B (HepB) vaccine recipients, only 54% received 3 doses within 1 year, indicating many adults who start the current vaccine series remain at risk for HBV infection.  Traditionally, vaccination against hepatitis B infection is a 3-dose series given over the course of six months.  Recently, a 2-dose HepB vaccine that can be completed in one month was licensed for use in adults.  This 2-dose regimen could result in higher vaccine completion rates (compared with the 3-dose series), especially among individuals in high risk settings.

In previous NEEMA work, we assessed the cost-effectiveness of using a 2-dose vaccine vs a 3-dose vaccine in various high groups that have an increased risk of HBV infection.  While that project quantified the costs and benefits of using a 2-dose vaccine among high-risk individuals, it did not incorporate the cost and benefits of providing that vaccine in setting in which high-risk persons of seek services.  This project will build on that previous work to examine the cost-effectiveness of a 2-dose vs 3-dose vaccine in various settings that serve persons in those high-risk groups.

Project Objectives or Major Research Questions

The objective of this study is to assess cost-effectiveness of using a 2-dose versus 3-dose HepB vaccine in a variety of settings that serve persons at high risk for HBV infection.  We propose evaluating the cost-effectiveness of the different vaccination strategies in the following selected settings:

  • STD clinics
  • TB clinics
  • Substance use clinics
  • Syringe service programs
  • Jail (median length of stay in LA County is 29 days)
  • HBV screening and vaccination outreach events in communities with large immigrant populations

Note that this concept is still under development by academic leads and CDC partners; what follows is an excerpt from recent correspondence to CDC and contains a series of options (not included here for the sake of brevity and simplicity) for consideration by CDC which is awaiting feedback.

National estimates of prevalence of injection drug use (IDU) are currently limited to self-report among respondents to household surveys. Several indirect indications of IDU behavior suggest prevalence is increasing, but this is difficult to measure directly due to inadequacies of the survey data. Indirect indications of increasing IDU behavior include increased drug overdose mortality rates, increases in observed and estimated acute HCV infections, geographically isolated increases in IDU-related HIV infections, and increases in other infections that are associated with IDU, e.g. infectious endocarditis and osteomyelitis. We propose to use estimates of indirect indications of IDU behavior in combination with estimated conditional probabilities of having these outcomes among persons who inject drugs (PWID) to estimate PWID population size.

The proposed method is similar to the method currently used to estimate the multiplier for acute hepatitis infections (Klevens et al., AJPH, 2014). For acute HCV, a starting number of reported cases to surveillance is inflated by available estimates of the conditional probabilities of being symptomatic given infection, seeking care given symptoms, and being reported to surveillance given care-seeking behavior (together these 3 quantities estimate the number of infections not reported to surveillance). Here, we propose an analogous multiplier method that begins with surveillance data for indirect IDU indications, and adjust these with survey data and estimates from literature review to estimate PWID population size. In contrast to the case of acute HCV, because the starting surveillance quantity is a signal of IDU, rather than a direct PWID number, several multiplier estimates may be used to triangulate on the true population quantity.

Proposals of how this might work for three indirect indicators of IDU – drug overdose deaths, substance abuse treatment, and skin and soft tissue infections are outlined separately. One important consideration going forward is data availability. Ideally, we would need data for each strategy considered, stratified by basic demographic characteristics such as sex, age, and race. We would ideally also work this exercise through for multiple indicators so they can be compared. Reasonable ranges for each input will be used for modeling to enable estimated ranges for PWID population size. Given limitations in data quality, we anticipate some inputs will need to be creatively estimated and are seeking CDC’s feedback on both the proposed method and possible data sources for this method.

While TB transmission rates have declined nationally in the US in the last two decades – owing largely to TB control efforts – ongoing transmission of TB, which can lead to sporadic outbreaks in various vulnerable populations and communities across the US (such as those experiencing homelessness, incarceration, poverty, lack of access to care, and are indigenous/rural/marginalized), remains an important priority for TB control. In this regard, preventing large outbreaks (cluster size of 10 and above) are an important component in curbing ongoing transmission. The epidemiological impact of preventing these outbreaks (through surveillance, outbreak investigation, and collection and analysis of molecular data) and the value of improving the methods of outbreak detection through incorporation of higher resolution data (e.g., whole genome sequencing and incorporation of ancillary epidemiological data) has not been quantitatively described. Using (i) epidemiological data on TB outbreaks that were detected in the US between 2014-2017 (including the frequency and size of outbreaks and characteristics of the corresponding underlying populations); (ii) novel stochastic models of TB outbreaks in the US (which will be used to simulate outbreaks of different frequency, size, and population characteristics); and to the extent they are available, (iii) estimates of the US TB outbreak control response (including the components of a typical response and the corresponding resources required), we propose to estimate the epidemiological impact and cost-effectiveness of TB outbreak surveillance and control in the US. Note that the cost-effectiveness analysis will be dependent on available data to inform the corresponding cost estimates and is planned for Year 2 of this project; we will discuss the availability and reliability of cost estimates with CDC partners before finalizing our plan for Year 2.

Project Objectives or Major Research Questions

We propose to collaborate closely with DTBE, particularly with Surveillance, Epidemiology, and Outbreak Investigations Branch (SEOIB) and the Data Management, Statistics, and Evaluation Branch (DMSEB), to develop TB outbreak models in various vulnerable populations in the US, by incorporating data on TB outbreaks, and outbreak response. We propose to use these models to estimate: (a) the projected number of TB cases averted (and the proportion of recent transmission reduced) from preventing outbreaks of large sizes; (b) number of TB cases averted in key counties/jurisdictions or sub-populations (e.g., homeless, incarcerated, rural, indigenous etc.); (c) the added value of using data from whole-genome sequencing linked to epidemiological investigation, incremental to conventional genotyping; and if feasible (as discussed above), (d) corresponding estimates of cost and cost-effectiveness of using the whole-genome sequencing/phylogenetic approach (incremental to conventional genotyping) [provided cost data for both the conventional and the new approach are available]. Importantly, we will also include a comparison of existing outbreak response efforts to one of no outbreak response, by way of motivating sustained and/or increased support for existing outbreak response activities, as these activities likely play a substantial role in mitigating the effects of existing outbreaks.