The Role of HIV Partner Services in the Modern Biomedical HIV Prevention Era: A Network Modeling Study : Sexually Transmitted Diseases

Secondary Logo

Journal Logo

Original Studies

The Role of HIV Partner Services in the Modern Biomedical HIV Prevention Era: A Network Modeling Study

Jenness, Samuel M. PhD, MPH; Le Guillou, Adrien MD, MPH; Lyles, Cynthia PhD; Bernstein, Kyle T. PhD, ScM; Krupinsky, Kathryn MPH; Enns, Eva A. PhD§; Sullivan, Patrick S. DVM, PhD; Delaney, Kevin P. PhD, MPH

Author Information
doi: 10.1097/OLQ.0000000000001711

HIV partner services have the potential to prevent HIV by intervening with the exposed partners of persons newly diagnosed with HIV (index patients). Recommended partner interventions include HIV screening to diagnose asymptomatic infection, with subsequent linkage to HIV services depending on the test result.1 Traditionally in the United States, HIV partner services activities have focused on linkage to antiretroviral therapy (ART) for partners who are newly diagnosed,2 potentially lowering community HIV viral load (VL) and reducing rates of subsequent HIV transmission (i.e., “treatment as prevention”). However, some jurisdictions have used HIV partner services as a pathway to other prevention services, such as HIV preexposure prophylaxis (PrEP) and sexually transmitted infection (STI) screening.3

HIV partner services consistently increase the coverage of these HIV services,4 but the degree to which this program can contribute to population-level HIV incidence reduction is unclear. Achieving this population impact requires reducing the time to ART or PrEP uptake compared with what is achieved through other, individual-level interventions. For example, 2018 national estimates for the average time from HIV infection to diagnosis and from diagnosis to VL suppression are 40 and 4 months, respectively.5 Partner services must identify a large fraction of HIV-infected partners currently unsuppressed and link them to ART to reduce these clinical delays and thereby lower population-based HIV transmission rates. Current partner services metrics suggest challenges in achieving this.1

Gay, bisexual, and other men who have sex with men (MSM) have been the largest group of users of HIV partner services in the United States because of their disproportionate HIV incidence.1 Although levels of ART and PrEP coverage have increased among MSM, gaps between actual and recommended use persist, particularly in disadvantaged MSM subpopulations.6 Optimization of partner services for these MSM may be an approach to improving the uptake of these biomedical tools, particularly in jurisdictions (such as Atlanta) where the current partner services infrastructure is weak and the incidence of HIV is high.7,8

In this study, we used a stochastic network model of HIV transmission dynamics among MSM in Atlanta to evaluate the role of HIV partner services in this current HIV biomedical prevention era. Our aims were to quantify how both traditional (HIV screening and ART linkage) and newer (PrEP linkage) components of HIV partner services function together to achieve HIV prevention goals.


Study Design

We used a stochastic network model of HIV, Neisseria gonorrhoeae (NG), and Chlamydia trachomatis (CT) transmission dynamics for US MSM, built with the EpiModel software platform.9 Our model simulates disease dynamics over dynamic sexual partnership networks using temporal exponential random graph models (TERGMs).10 This approach is necessary to realistically represent and track sexual partnership histories. This current study builds on our previous studies of partner-based strategies for HIV/STIs.11 Full methodological details are provided in the Appendix (, and model code is available on GitHub (

Our model represented main, casual, and one-time sexual partnerships for Black, Hispanic or Latino (Hispanic), and White/other-race (White) MSM, aged 15 to 65 years, in the Atlanta metropolitan statistical area. The starting network size in model simulations was 10,000 MSM (approximately one-tenth of the estimated Atlanta MSM population size of 102,00012), which could increase or decrease over time based on arrival (sexual debut) and departure (mortality or sexual cessation; Appendix S5,

Core HIV/STI Model

The model consisted of 5 core components: (1) statistical network models (TERGMs) to represent dynamic sexual partnership formation and dissolution; (2) statistical generalized linear models to predict behavior within partnerships; (3) simulation of pathogen transmission across active partnerships; (4) simulation of HIV/STI disease progression, recovery, and natural history; and (5) simulation of HIV/STI prevention and clinical interventions (including partner services). Transmissions of NG and CT were simulated along the same sexual network as HIV, but with disease recovery. Full details on the STI components are found in Appendices S9–12 (

Empirical data from the ARTnet study were used to fit the statistical models for networks and behavior (components 1 and 2; Appendix S2, ARTnet was a Web-based egocentric network study on MSM in the United States, which was conducted in 2017 to 2019.13 We weighted model parameters by census-based race/ethnicity and age distributions for Atlanta MSM to account for the ARTnet sampling design. Predictors in TERGMs for partnership formation included partnership type (main, casual, and one-time), demographic heterogeneity in network degree, and assortative mixing by demographics and sexual position (Appendix S3, Partnership dissolution was modeled for main and casual partnerships, with rates stratified by partnership type and age. Generalized linear models were fit to predict the frequency of sexual acts per partnership per week and the probability of condom use per act as a function of race/ethnicity, age, HIV status, and partnership type and duration (Appendix S4,

Our base model included both individual-based and partner services-based (described hereinafter) pathways to HIV screening, treatment, and prevention services (Appendix S7, For the individual-based pathway, in each weekly time step, MSM could be screened for HIV based on rates calibrated to reproduce estimates from the Georgia Department of Public Health surveillance of the proportion of MSM with HIV who were diagnosed.14 Men who have sex with men who screened positive entered the HIV care cascade, with opportunities to initiate antiretroviral therapy (ART), which reduced VL and disease progression. Men who have sex with men on ART could cycle off and on ART based on rates also calibrated to Georgia Department of Public Health data.15 Lower VL due to sustained ART use was associated with a reduced probability of HIV transmission. Other factors modifying the HIV transmission probabilities included PrEP use, condom use, sexual position, circumcision, and current NG or CT infection (Appendix S8,

Men who have sex with men who screened HIV negative entered the HIV prevention cascade, consisting of initiation, adherence, and persistence in PrEP care for daily oral tenofovir/emtricitabine.16 Men who have sex with men who tested HIV negative and met indications for PrEP based on Centers for Disease Control and Prevention (CDC) guidelines were eligible to start. Indicated MSM then started PrEP based on an initiation probability generating a coverage level of 15%.6 Heterogeneous PrEP adherence was modeled, with 78.4% meeting a high-adherence level that resulted in a 99% reduction in HIV risk.17 Preexposure prophylaxis discontinuation was based on secondary estimates of the proportion of MSM who were retained in PrEP care at 6 months (57%).18 Preexposure prophylaxis care included quarterly HIV and STI screening.

HIV Partner Services

Table 1 summarizes the design elements and parameters for partner services in our model. The 4 included steps of the partner services cascade were the following: (1) identification of newly diagnosed MSM (“index patients”) through routine HIV screening and their participation in partner services; (2) elicitation and identification of recent (past year) sexual partners of index patients; (3) provision of HIV screening to those partners; and (4) linkage of partners to HIV care or prevention services, depending on partner HIV status. Hereinafter, we call the first 2 steps “upstream” and the latter 2 steps “downstream.” The base (calibrated) model scenario represents partner services as currently delivered in Atlanta, whereas the counterfactual (experimental) scenarios modify metrics for each step.

TABLE 1 - Partner Services Cascade Parameters and Sources
Parameter Description Base Model Parameter Sources
Missing = No Complete Case
Index network degree Minimum number of cumulative partners in window before triggering partner services 1 1 Standard Operating Protocol*
Index initiation Probability that an index patient will be initially offered and accept partner services 66.6% 66.6% 2018 CDC Partner Services Report
Index partnership window Number of weeks in the past that a partnership of each type qualifies for partner elicitation Standard Operating Protocol*
 Main partners 52 52
 Casual partners 52 52
 One-time partners 52 52
Partner identification
 Partners/index ratio Number of partners identified per index patient 0.432 0.432 2018 CDC Partner Services Report
 Casual to main scalar Odds ratio for the probability identification of casual partners to that of main partners 0.50 0.50
 One-time to main scalar Odds ratio for the probability identification of 1-time partners to that of main partners 0.25 0.25
Partner HIV screening Probability of completing HIV screening for partners identified and reached 39.4% 84.0% 2018 CDC Partner Services Report
Partner PrEP linkage Probability of a partner who screened HIV− and with current indications starting PrEP 0 0 Assumed based on Standard Operating Protocol for ATL*
Partner care linkage Probability of a partner who newly screened HIV+ and not currently in care will start ART 38.7% 96.0% CDC Partner Services Report
Partner ART reengagement Probability of a partner who previously screened HIV+ and previously in ART care will restart ART 0 0 Assumed based on Standard Operating Protocol for ATL*
*Communication with CDC coauthors (K.T.B. and C.L.) familiar with standard partner services protocol.

All sexually active index patients (minimum cumulative network degree = 1) who had recent partners (eligible partner time frame = past 52 weeks) were eligible for partner services. In the base scenario, 67% of index patients accepted partner services (Table 1). Recent partners of index patients were identified with probabilities stratified by partnership type, highest for main partners and lowest for one-time partners. Identified partners could self-report as either HIV positive or unknown status; the latter were offered HIV screening. Partners who newly screened HIV positive or self-reported HIV positive and were still ART naive were eligible for HIV care linkage, resulting in immediate ART initiation. Partners who screened HIV negative were eligible for immediate PrEP linkage if they met CDC indications. Partners who self-reported as HIV positive but who had fallen out of care were eligible for immediate ART reengagement. Our base scenarios assumed no PrEP linkage or ART reengagement, as these partner service activities are uncommon in Atlanta.

Experimental scenarios evaluated improvements to each partner services cascade step. Our first set of scenarios evaluated the impact of improving each step, alone and jointly. Counterfactual scenarios outcomes were compared with a “missing = no” base scenario (Table 1) that calculated CDC partner services metrics assuming that missing data represented a failure. Sensitivity analyses explored the alternative “complete case” assumption that excluded missing data (potentially overestimating success). Second, to understand the optimal impact of partner services, we simulated scenarios with all metrics at their maximum values. Third, we estimated the impact of more realistic improvements for the upstream steps, in which 90% of index patients and 25% or 50% of their partners were reached.

Calibration, Simulation, and Analysis

We calibrated the model with a Bayesian approach that defined prior distributions for uncertain parameters and fit the model to surveillance data on diagnosed HIV, NG, and CT incidence for the target population (Appendix S13, After calibration, we simulated the model 500 times for each scenario and summarized the distribution of results with medians and 95% simulation intervals (SIs). The time horizon for all scenarios was 10 years.

Our primary outcomes were HIV incidence per 100 person-years at risk in the last year of the time horizon and cumulative incidence across the horizon. We calculated the number and percent of infections averted relative to the cumulative incidence in the base scenario. To estimate the efficiency of some experimental scenarios, we calculate the number needed to treat (NNT) as the number of index patients needing to start partner services to avert one new HIV infection. To understand the HIV prevention mechanisms, we calculated cross-sectional HIV care, prevention, and partner services cascade metrics.


Table 2 summarizes our base and primary experimental scenarios. In the base scenario, the point HIV incidence among Atlanta MSM was 1.16 per 100 person-years at risk (95% SI, 0.91–1.44) at year 10, with 1090 cumulative infections for 10 years. Improvements to each individual partner services step, while holding the values of other steps constant, had negligible impact on HIV incidence, with less than 1% of infections averted even when steps were individually set to their optimal (maximum) value.

TABLE 2 - Impact of Changes to Individual Partner Services Cascade Step Changes on Point and Cumulative HIV Incidence
Scenario Epidemiological Outcomes
Incidence Rate at Year 10 Cumulative Incidence Over 10 y No. Infections Averted Over 10 y Percent Infections Averted Over 10 y
Median (95% SI) Median (95% SI) Median (95% SI) Median (95% SI)
Base scenario
 No changes 1.16 (0.91 to 1.44) 1090 (1006 to 1193)
Index initiation probability
 70% 1.15 (0.88 to 1.44) 1090 (999 to 1192) 0 (−102 to 91) 0.0 (−9.4 to 8.4)
 100% 1.16 (0.89 to 1.42) 1092 (992 to 1200) −2 (−110 to 98) −0.2 (−10.1 to 9.0)
Partner identification probability
 Base ×1.5 1.16 (0.90 to 1.47) 1093 (1002 to 1185) −3 (−95 to 88) −0.3 (−8.7 to 8.1)
 100% (all eligible partners) 1.14 (0.88 to 1.45) 1079 (987 to 1182) 11 (−92 to 103) 1.0 (−8.4 to 9.5)
Partner HIV screening probability
 Base ×2.0 1.15 (0.91 to 1.42) 1087 (982 to 1187) 3 (−97 to 108) 0.3 (−8.9 to 9.9)
 100% (all eligible partners) 1.14 (0.87 to 1.45) 1092 (988 to 1189) −2 (−99 to 102) −0.2 (−9.1 to 9.4)
Partner PrEP linkage probability
 50% 1.15 (0.91 to 1.45) 1092 (988 to 1187) −2 (−97 to 102) −0.1 (−8.9 to 9.4)
 100% (all eligible partners) 1.15 (0.90 to 1.43) 1087 (985 to 1185) 3 (−95 to 105) 0.3 (−8.7 to 9.6)
Partner ART linkage probability
 Base ×1.5 1.15 (0.90 to 1.46) 1092 (989 to 1191) −2 (−101 to 101) −0.2 (−9.3 to 9.3)
 100% (all eligible partners) 1.15 (0.89 to 1.43) 1091 (997 to 1194) −1 (−104 to 93) −0.1 (−9.5 to 8.5)
Partner ART reengagement probability
 50% 1.15 (0.89 to 1.44) 1088 (988 to 1177) 2 (−87 to 102) 0.2 (−8.0 to 9.4)
 100% (all eligible partners) 1.13 (0.88 to 1.42) 1082 (986 to 1183) 8 (−93 to 104) 0.7 (−8.5 to 9.5)

Supplemental Table 3 ( demonstrates why: individual step changes had limited impact on the HIV prevention and care cascade metrics of PrEP coverage and HIV VL suppression. Supplemental Tables 4 and 5 ( subsequently demonstrate why those cascade metrics did not change. In the base scenario, for 10 years, the number of index patients reached was 622, who reported on 2694 partners, of whom 273 were identified, of whom 92 HIV screened, of whom 19 screened HIV positive, of whom 6 were immediately linked to ART. It was only the MSM who reached this final step who had the potential to lower community VL, which could reduce the population transmission rate. When instead we assumed that all partners were identified, but everything else equal, ART linkage would increase to 63 partners; assuming optimal PrEP linkage, but everything else equal, 37 partners would start PrEP. These counts are too small to make a meaningful impact on PrEP coverage and VL suppression. Supplemental Table 1 ( shows that limited impact was not due to the choice of calibration scenario, with no difference in incidence in the complete case scenario. In fact, when we simulated a scenario without partner services, there was effectively no difference in incidence, suggesting a minimal impact of partner services as currently delivered.

Table 3 shows the projected impact under optimal conditions, in which both upstream partner services steps (index identification and partner identification) and downstream steps (PrEP, ART linkage, ART reengagement) were set to their maximum values. Supplemental Figure 1 ( visualizes the HIV incidence time series across these scenarios. Here again, when changes were limited to upstream steps, the impact was minimal (1.0% infections averted); however, as downstream services were added, the impact grew: 5.5% of infections averted when all partners were screened, 10.7% when all indicated partners were provided PrEP, 5.4% when newly diagnosed HIV-positive partners were linked to ART, 17.6% when previously diagnosed HIV positive out of ART care were relinked to ART, and 21.7% when all of these downstream services were optimized. Table 3 also provides the NNT, which ranges from 7.7 when upstream steps are optimized to 1.3 when all steps are optimized. The mechanisms for the HIV prevention impact are shown in Supplemental Table 6 ( PrEP coverage increased by 3 percentage points, and HIV VL suppression increased by 6 percentage points. The mechanisms for these changes are demonstrated in Supplemental Tables 7 and 8 ( in our optimal scenario, the number of partners identified was 4082, of whom 2477 were HIV screened, of whom 1454 started PrEP, 462 newly started ART, and 580 restarted ART.

TABLE 3 - Impact of Changes to Partner Services on Point and Cumulative HIV Incidence, at Theoretical Maxima of Cascade Steps
Scenario Epidemiological Outcomes
Incidence Rate at Year 10 Cumulative Incidence Over 10 y No. Infections Averted Over 10 y Percent Infections Averted Over 10 y NNT (No. Indexes Started)
Median (95% SI) Median (95% SI) Median (95% SI) Median (95% SI) Median (95% SI)
Base scenario
 No changes 1.16 (0.91 to 1.44) 1090 (1006 to 1193)
Counterfactual scenarios
 All index and all partners identified 1.14 (0.90 to 1.45) 1079 (983 to 1179) 11 (−89 to 107) 1.0 (−8.2 to 9.8) 7.71 (−182.00 to 362.05)
  + All partners HIV screened 1.10 (0.83 to 1.37) 1030 (941 to 1123) 60 (−33 to 149) 5.5 (−3.0 to 13.7) 5.93 (−53.91 to 100.25)
  + All eligible partners provided PrEP 1.02 (0.79 to 1.26) 973 (898 to 1058) 117 (32 to 192) 10.7 (2.9 to 17.6) 3.38 (1.61 to 12.77)
  + All eligible partners linked to ART 1.08 (0.86 to 1.38) 1031 (941 to 1131) 59 (−41 to 149) 5.4 (−3.8 to 13.7) 6.15 (−76.53 to 73.07)
  + All eligible partners relinked to ART 0.91 (0.70 to 1.15) 898 (822 to 973) 192 (117 to 268) 17.6 (10.7 to 24.6) 1.81 (1.01 to 3.63)
  + All eligible linked and relinked to ART 0.90 (0.69 to 1.16) 894 (822 to 972) 196 (118 to 268) 18.0 (10.8 to 24.6) 1.76 (1.04 to 3.45)
  + All partner services at maxima 0.87 (0.65 to 1.10) 853 (778 to 928) 237 (162 to 312) 21.7 (14.9 to 28.6) 1.32 (0.78 to 2.35)

Figure 1 visualizes the general relationship between the improvement in the 2 upstream steps (index and partner initiation probabilities) at current versus improved levels of downstream step service provision. The left panel shows that there is no impact on infections averted when scaling up one or both upstream identification steps without improvements in downstream steps. The right panel shows that when downstream services are first optimized, improving upstream steps yields a general linear increase in infections averted after some base provision of each. In summary, upstream steps were not impactful without improvement in downstream service provision, and downstream steps were maximized only with improvements to upstream steps.

Figure 1:
General relationship between improved partner services index initiation and partner identification probabilities on the 10-year percent of infections averted (PIA) compared with the base (calibrated) model scenario. Panel A (left) demonstrates the relationship under base downstream services (HIV screening, PrEP, and ART), whereas panel B (right) shows the relationship when all downstream services are set to their theoretical maxima. Improvements to index or partner engagement have no discernable impact on the PIA under base downstream services but are associated with an upward of 20% infections averted when downstream services are maximized.

Figure 2 and Supplemental Tables 9 to 15 ( estimate the impact of partner services under more realistic scenarios in which 90% of index patients were identified, and either 25% or 50% of partners were identified. Although the impact on HIV incidence under these scenarios was not as high as the optimal scenarios, notable HIV prevention benefits (2%–8% of infections averted) were achieved. Figure 2 shows the relationship between improvements in HIV screening and improvements in either PrEP linkage or ART reengagement. Like Table 3, there was an overall larger relative benefit of ART service engagement than PrEP engagement. This was further demonstrated in the scenarios in Supplemental Figure 2 (, in which we observed greater synergy between HIV screening and ART scale-up than PrEP scale-up.

Figure 2:
General relationship between improved partner HIV screening and either partner ART linkage/reengagement for HIV-infected partners (top panels) or PrEP initiation for HIV-negative partners (bottom panels) on the 10-year percent of infections averted (PIA) compared with the base (calibrated) model scenario, assuming 90% of index patients are identified and either 25% of their partners (left panels) or 50% of their partners (right panels) are identified. Higher levels of HIV screening are associated with greater HIV prevention, but ART engagement had a greater relative effect than PrEP initiation. The total benefit was greater when the partner identification was higher because of a larger number of identified partners who eligible for ART or PrEP.


In this study, we estimated the population impact of HIV partner services in the context of our modern HIV biomedical prevention era of increased availability of highly effective antiretroviral HIV prevention tools. We found that HIV partner services, as currently delivered in our target population, provided clear clinical benefits to newly diagnosed index patients and their sexual partners. However, the intervention had limited impact on HIV transmission due to the steep drop from upstream cascade steps (identifying index patients and their partners) to delivery of downstream transmission-disrupting services (ART and PrEP). Improvements only to upstream steps had a negligible impact because of the limited delivery of downstream services. To achieve population prevention benefit, therefore, substantial but achievable improvements in the upstream steps of partner services must be paired with similar improvements in downstream service scale-up.

The success of partner services has traditionally been measured in terms of process metrics of partners identified and screened.4,19 However, it is unclear how these metrics translate into HIV incidence reduction. Mathematical modeling can help bridge the gap to estimate these quantities. However, models require realistic representations of sexual networks and the partner identification and HIV screening processes across those networks. Prior partner services models have used more simplistic representations that represent partner services as general increases in HIV diagnosis rates.20 This could overestimate the impact of partner services by failing to track partnership dyads that inherently limit the reach of this program. Our model builds upon the robust network structure representation that avoids this problem.11

Our study highlights the important contribution that partner services could make toward HIV prevention, with an estimated 22% of HIV infections averted under optimal service levels. Achieving these gains, however, will require substantial further investments to increase the size and speed of HIV partner services. This program has been weakened by the COVID-19 pandemic, which reallocated many disease intervention specialists who conduct this work away from HIV and toward COVID contract tracing.21 In promising recent developments, after years of flat or declining funding for the public health workforce assigned to partner services, the US American Recue Plan Act of 2021 has included $1 billion to expand and enhance disease intervention specialist staff for HIV/STI partner services over the next 5 years.22 With this, some of the optimistic targets in our study may become feasible.

Our study also demonstrates the excellent efficiency (i.e., low NNT) of partner services, which translates into lower resources per infection averted. For example, we estimated an NNT of 1.3 index patients must be engaged to prevent one HIV infection under optimal delivery. This high efficiency for partner services was achieved by targeting HIV prevention services directly to areas of the sexual network with the highest HIV transmission potential. In comparison, we have estimated the NNT for individually targeted HIV PrEP in the same target population to be 20 to 30.23 A network-based cost-effectiveness analysis is suggested to compare investments in these 2 interventions because their costs and durability also differ.

The primary mechanism for partner services to impact HIV incidence is by reducing the time at each stage along the sequence from infection to diagnosis to ART initiation to VL suppression for sexual partners. The first step, HIV diagnosis, has been demonstrated to be a bottleneck in this sequence,5 delaying HIV VL suppression by years. Targeted HIV screening within partner services does fill a gap in individual-focused routine HIV screening efforts,24 including diagnosis during acute stage HIV infection when many HIV transmissions occur, but that does not necessarily translate to population prevention benefits. In our model, the accelerated HIV screening achieved through partner services by itself did not have a major impact, both because the current undiagnosed fraction was relatively low and because the numbers initiating ART after testing through partner services were small. Still, if HIV screening for partners were maximized, our model suggests a prevention benefit of up to 5.5% of infections.

Efforts to further integrate delivery of ART and PrEP within partner services would also improve the overall impact of this program. Health jurisdictions like Seattle have used partner services to identify and link partners to HIV PrEP during time periods that are, by definition of being a contact with a HIV-infected person, high risk.3 A network intervention in Chicago successfully used partner services to link high-risk young persons to PrEP care.25 For ART, we found that partner services efforts to identify partners who are known to be HIV positive but who have fallen out of HIV care could have a strong impact, consistent with prior research.26 Our study suggests that prioritizing ART reengagement for partners self-reported as HIV positive but not in care had the overall greatest population benefit in HIV prevention (18% of infections averted).


First, we represented HIV partner services as a single type that would be provided by disease intervention specialists through local health authorities. In reality, alternative versions of partner services, both formal and informal, are provided by clinicians diagnosing index patients and even index patients themselves.1 However, the empirical data to differentiate these versions are too sparse to support a model. Including these alternative versions could expand the potential reach of partner services, but likely not to a degree that would substantially alter our conclusions.27 Second, HIV surveillance data suggest a bimodal distribution of individual-based HIV screening, where some MSM remain undiagnosed until AIDS.28 We did not model this heterogeneity because of the complexity in representing the interaction of testing typologies with HIV progression rates. This resulted in a slight underestimate of the simulated HIV diagnostic delay in the model compared with recent national estimates,5 which could have a minor impact on the projected impact of partner services. Third, we assumed that partner services only influenced the population force of infection through modifications to PrEP and ART coverage. Partner services may also have the ability to reduce HIV acquisition or transmission rates among identified partners through behavioral counseling (e.g., to increase condom use)2 or through linkage to testing and treatment for bacterial STIs.29 However, the causal evidence for these other prevention mechanisms is weak. Fourth, although we represented the co-circulation of NG and CT in our model, we do not currently use an incident STI diagnosis as a trigger for HIV partner services. That was outside the current scope of this study, although it is planned for future analyses. Including STIs to identify index patients could increase the overall index patient pool that could subsequently be elicited for partners who could benefit from HIV prevention services.30 Finally, these projected results are in the context of an Atlanta MSM target population, which has relatively high HIV prevalence and incidence far above HIV elimination targets. Partner services may have a higher relative impact in low-prevalence settings where uptake of other interventions has become saturated.


HIV partner services in the United States are at a crossroads. The new Ending the Epidemic plan calls for meeting ambitious HIV prevention goals in the next decade through a combination of public health efforts and intervention approaches. In the biomedical HIV prevention era, however, traditional partner services are complimented by a variety of individual-, partner-, network-, and community-based approaches to increase the coverage of antiretroviral medication to both persons with HIV (ART) and those without HIV at ongoing risk for HIV transmission or acquisition (PrEP). Our current study not only demonstrates the optimal reach of HIV partner services in this context but also estimates its currently limited role in population health without substantial investments in the size and scope of this program. Recent increases in resources for partner services22 may help overcome these limitations. Our study can help guide the optimization of partner services to maximize its impact.


1. Centers for Disease Control and Prevention (CDC). Recommendations for partner services programs for HIV infection, syphilis, gonorrhea, and chlamydial infection. MMWR Recomm Rep 2008; 57:1–83; quiz CE1-4.
2. Magaziner S, Montgomery MC, Bertrand T, et al. Public health opportunities and challenges in the provision of partner notification services: The New England experience. BMC Health Serv Res 2018; 18:75.
3. Katz DA, Dombrowski JC, Barry M, et al. STD partner services to monitor and promote HIV pre-exposure prophylaxis use among men who have sex with men. J Acquir Immune Defic Syndr 2019; 80:533–541.
4. Udeagu C-CN, Beer L, Lu J-F, et al. Partner services among HIV-positive adults receiving medical Care in the United States: Medical Monitoring Project, 2013−2014. AIDS Patient Care STDS 2020; 34:99–101.
5. Crepaz N, Song R, Lyss SB, et al. Estimated time from HIV infection to diagnosis and diagnosis to first viral suppression during 2014–2018. AIDS 2021; 35:2181–2190.
6. Kelley CF, Kahle E, Siegler A, et al. Applying a PrEP continuum of care for men who have sex with men in Atlanta, Georgia. Clin Infect Dis 2015; 61:1590–1597.
7. Goodreau SM, Rosenberg ES, Jenness SM, et al. Sources of racial disparities in HIV prevalence in men who have sex with men in Atlanta, GA, USA: A modelling study. Lancet HIV 2017; 4:e311–e320.
8. Siegler AJ, Bratcher A, Weiss KM, et al. Location location location: An exploration of disparities in access to publicly listed pre-exposure prophylaxis clinics in the United States. Ann Epidemiol 2018; 28:858–864.
9. Jenness SM, Goodreau SM, Morris M. EpiModel: An R package for mathematical modeling of infectious disease over networks. J Stat Softw 2018; 84:8.
10. Krivitsky PN, Handcock MS. A separable model for dynamic networks. J R Stat Soc Series B Stat Methodol 2014; 76:29–46.
11. Weiss KM, Jones JS, Katz DA, et al. Epidemiological impact of expedited partner therapy for men who have sex with men: A modeling study. Sex Transm Dis 2019; 46:697–705.
12. Grey JA, Bernstein KT, Sullivan PS, et al. Estimating the population sizes of men who have sex with men in US states and counties using data from the American community survey. JMIR Public Health Surveill 2016; 2:e14.
13. Weiss KM, Goodreau SM, Morris M, et al. Egocentric sexual networks of men who have sex with men in the United States: Results from the ARTnet study. Epidemics 2020; 30:100386.
14. HIV/AIDS Epidemiology Section of the Georgia Department of Public Health. HIV Surveillance Summary 2018. Available at: Accessed May 1, 2022.
15. Georgia Department of Public Health, HIV/AIDS Epidemiology Section. HIV Surveillance Summary, Georgia 2017. 2019. Available at: Accessed April 20, 2020.
16. Jenness SM, Maloney KM, Smith DK, et al. Addressing gaps in HIV preexposure prophylaxis care to reduce racial disparities in HIV incidence in the United States. Am J Epidemiol 2019; 188:743–752.
17. Grant RM, Anderson PL, McMahan V, 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. Lancet Infect Dis 2014; 14:820–829.
18. Chan PA, Mena L, Patel R, 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; 19:20903.
19. Rao S, Song W, Mulatu MS, et al. HIV partner service delivery among Blacks or African Americans—United States, 2016. MMWR Morb Mortal Wkly Rep 2019; 68:81–86.
20. Nichols BE, Götz HM, van Gorp ECM, et al. Partner notification for reduction of HIV-1 transmission and related costs among men who have sex with men: A mathematical modeling study. PLoS One 2015; 10:e0142576.
21. Jenness SM, Le Guillou A, Chandra C, et al. Projected HIV and bacterial sexually transmitted infection incidence following COVID-19–related sexual distancing and clinical service interruption. J Infect Dis 2021; 223:1019–1028.
22. CDC DIS Workforce Development Funding. 2021. Available at: Published online August 26, 2021. Accessed May 18, 2022.
23. Jenness SM, Goodreau SM, Rosenberg E, 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. J Infect Dis 2016; 214:1800–1807.
24. Bernstein KT, Stephens SC, Moss N, et al. Partner services as targeted HIV screening—Changing the paradigm. Public Health Rep 2014; 129(Suppl 1):50–55.
25. Teixeira da Silva D, Bouris A, Ramachandran A, et al. Embedding a linkage to preexposure prophylaxis care intervention in social network strategy and partner notification services: Results from a pilot randomized controlled trial. J Acquir Immune Defic Syndr 2021; 86:191–199.
26. Udeagu CN, Shah S, Misra K, et al. The usefulness of HIV partner services in the age of treatment as prevention: A registry-based study. Lancet HIV 2020; 7:e482–e490.
27. Malave MC, Shah D, Sackoff JE, et al. Human immunodeficiency virus partner elicitation and notification in New York city: Public health does it better. Sex Transm Dis 2008; 35:869–876.
28. Centers for Disease Control and Prevention (CDC). HIV testing and risk behaviors among gay, bisexual, and other men who have sex with men—United States. MMWR Morb Mortal Wkly Rep 2013; 62:958–962.
29. Weiss KM, Jones JS, Anderson EJ, et al. Optimizing coverage vs frequency for sexually transmitted infection screening of men who have sex with men. Open Forum Infect Dis 2019; 6:ofz405. doi:10.1093/ofid/ofz405.
30. Golden MR, Kerani RP, Stenger M, et al. Uptake and population-level impact of expedited partner therapy (EPT) on Chlamydia trachomatis and Neisseria gonorrhoeae: The Washington State community-level randomized trial of EPT. PLoS Med 2015; 12:e1001777.

Supplemental Digital Content

Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American Sexually Transmitted Diseases Association.