HIV remains a significant global health challenge, demanding innovative prevention approaches and more effective use of existing strategies to reduce incidence among priority populations such as gay, bisexual, and other men who have sex with men (GBMSM).¹ In the United States (US), HIV Partner Services (HIV-PS) is an established public health strategy², designed to facilitate treatment referrals for individuals diagnosed with HIV and link their sexual partners to diagnosis, treatment and preventive services.^3,4 While HIV-PS has demonstrated public health benefits, including timely diagnosis and treatment for infected individuals, transmission chain disruption through contact tracing, and targeted linkage to preventative measures like HIV pre-exposure prophylaxis (PrEP)⁵, its full potential remains unrealized. Enhancing the effectiveness of HIV-PS in identifying undiagnosed HIV infections and reaching those at increased risk in GBMSM populations is a critical priority for public health. With further optimization, HIV-PS could serve as an efficient strategy to achieve these goals.^6–8

Traditional HIV-PS primarily targets newly diagnosed individuals, aiming to quickly link them to medical care, while also identifying their sexual partners for rapid testing and care to prevent onward transmission.⁹ However, this approach often overlooks sexually active individuals with a previous HIV diagnosis (PWDH) who are inconsistently engaged in care. This group, identified as contributing one of the highest proportions of onward HIV transmissions^10–12, can transmit the virus during periods when detectable viral loads and unprotected sex with at-risk partners coincide.^13,14 Although partners of PWDH are often perceived to be at lower risk due to expected behavioral changes and antiretroviral therapy (ART) use^15–17, variability in the adoption and persistence of these behaviors, coupled with inconsistent ART adherence and engagement in care, leaves partners vulnerable to HIV acquisition.^13,14,18

Additionally, the changing dynamics of sexual partnerships in today’s digitally connected era have reshaped sexual networks, presenting challenges for traditional HIV-PS, which targets only named partners of index cases.^4,19 This approach potentially misses opportunities to engage additional individuals indirectly connected to the index case through their named sexual partners (i.e., sexual partners of the partners named by index cases). An implementation question therefore arises as to whether targeting additional waves of partners would yield substantial reductions in HIV incidence compared to focusing solely on immediate named partners.

These gaps likely contribute to the limited population-level impact of HIV-PS on reducing HIV incidence among US GBMSM.²⁰ Research indicates that HIV-PS identifies less than 5% of new HIV diagnoses in the US²¹, and recent modeling suggests that removing HIV-PS, as currently implemented, would only minimally affect cumulative incidence among GBMSM.⁸ While maximally optimizing steps within the HIV-PS cascade theoretically reduced incidence, realistic improvements only yielded modest results.⁸ The model emphasized that the limited number of newly diagnosed index cases and their partners accessing HIV transmission-interrupting services (ART and PrEP) through traditional HIV-PS is central to this limited impact. However, the study did not evaluate the potential benefits of including previously diagnosed individuals as index cases, or engaging partners of the index case’s partners in these services.

In this study, we used a stochastic network model of HIV transmission among GBMSM in Atlanta to assess the impact of expanding HIV-PS by incorporating out-of-care PWDH (as an index-pool expansion strategy) and partners of the index case’s named partners (as a partner-pool expansion strategy) on HIV incidence among GBMSM. We aimed to quantify how modifications to both upstream (working with index patients) and downstream (working with partners) HIV-PS cascade steps collectively influence HIV incidence reduction among GBMSM.

In our calibrated base (or reference) model representing existing HIV-PS, and focused solely on newly diagnosed GBMSM, we estimated 8196 (95% SI: 7951, 8509) new HIV infections among the 100,000 GBMSM in the network over 10 years, with a final year (year-10) HIV IR of 0.82 (95% SI: 0.75, 0.89) per 100 person-years (Supplemental Table 1-A). Notably, our findings highlight that substantial reductions in HIV incidence were only achieved when HIV-PS efforts included ART reengagement for previously diagnosed GBMSM who were out-of-care. Incorporating these individuals into the HIV-PS index-case pool had minimal impact on cumulative HIV incidence or year-10 IR in most single cascade step improvement scenarios (maximum PIA achieved = 2%), unless ART reengagement was also improved within the HIV-PS context (Table 2). In scenarios where ART reengagement was enhanced, the percentage of infections averted increased substantially, from 8% with a 10% increase in ART reengagement probability for PD-indexes to a maximum of 44% with 100% reengagement. Similarly, year-10 IR also decreased substantially in these scenarios, from 0.82 (base scenario) to 0.40 (95% SI: 0.36, 0.45) with 100% ART reengagement. In contrast, direct expansion of the HIV-PS partner pool through incorporation of Wave-2 partners did not meaningfully alter cumulative HIV incidence or year-10 IR in any of the single step improvement scenarios (maximum PIA achieved = 0.4%, Supplemental Table 3-A).

Results presented in Supplemental Tables 2-A through 2-C illustrate why single step changes in the HIV-PS cascade, except those involving ART reengagement for PD-indexes, resulted in minimal reductions in HIV incidence. The inclusion of PD-indexes in HIV-PS significantly expanded the pool of eligible and engaged index-cases, increasing it over ten-fold from a median of 663 eligible and 441 identified index-cases (Supplemental Table 1-C) to 7066 eligible and 4537 identified index-cases (Supplemental Table 2-B). This extension indirectly raised the pool of partners eligible for partner services from 781 (base model) to 7565 (intervention model). Despite this increase, only a small fraction of this enlarged partner pool engaged in downstream HIV transmission-disrupting services through HIV-PS. Of 7565 eligible partners in the intervention model, only 803 were identified (Supplemental Table 2-B). Among these, 351 had a prior HIV diagnosis, while 452 had a prior HIV-negative or unknown status. Supplemental Table 2-C shows that only a subset of this group progressed further: specifically, out of 452 HIV-negative/unknown partners, only 178 received HIV screening via partner services, resulting in 32 new HIV diagnoses. Of these 32 newly diagnosed partners, only 13 initiated ART through HIV-PS. Limited ART reengagement and PrEP initiation through HIV-PS meant that none of the 351 identified partners with prior HIV-positive status who might have been out of care, or the 146 screened partners who tested HIV-negative who might not have been on PrEP, initiated these transmission-disrupting strategies. This shift from a large pool of eligible individuals to a smaller group available for transmission-disrupting services, and the further reduction to a much smaller group directly receiving these services within the HIV-PS framework explains why expanding the index case pool without addressing engagement in HIV disrupting services (particularly ART reengagement) did not lead to substantial population-level impacts.

Table 3 presents the results of optimized scale-up scenarios where upstream and downstream HIV-PS cascade steps are jointly maximized. Compared to the reference model focusing solely on newly diagnosed GBMSM for HIV-PS, maximizing all cascade steps resulted in a 9.3% (95% SI: 4.8%, 13.0%) reduction in cumulative HIV incidence (PIA). Including Wave-2 partners of newly diagnosed index cases in HIV-PS did not appreciably decrease HIV incidence beyond that obtained in the base/reference model (maximum PIA: 11.3% [95% SI: 6.8%, 15.4%]). In the intervention model including out-of-care previously diagnosed GBMSM in HIV-PS, maximizing all cascade steps showed more significant improvements in HIV incidence reduction. Without improvements to ART reengagement for PD-indexes, maximal optimization of HIV-PS cascade steps yielded 34.5% PIA (95% SI: 30.8%, 37.4%). Maximizing ART reengagement for PD-indexes within the HIV-PS framework nearly doubled this impact, achieving a maximum PIA of 63.1% (95% SI: 60.7%, 65.2%) when all cascade steps were maximized. Trends in year-10 IR reflected similar improvements, with the lowest IR (0.26 [95% SI: 0.22, 0.30]) observed in the intervention model incorporating previously diagnosed individuals in HIV-PS and maximizing ART reengagement for all eligible within partner services (Figure 1C).

Table 3 also shows the estimated NNT to achieve reductions in HIV incidence under each model and scenario. In the reference model, maximal optimization of all HIV-PS cascade steps had the lowest NNT at 9 (95% SI: 6, 18), indicating that engaging nine index-cases in HIV-PS would avert 1 new HIV infection if all HIV-PS cascade steps are maximally improved. NNT varied widely across the intervention models, ranging from 19 to 229 with joint maximal optimization of all HIV-PS cascade steps when out-of-care PWDH were included in HIV-PS without improving ART reengagement for PD-indexes. However, NNT fell to 4 (95% SI: 4, 5) when reengagement in ART for PD-indexes was also maximized.

Figure 2 shows the relationship between improved index case and partner participation in HIV partner services and the percentage of new HIV infections averted (PIA) over 10 years. Improved participation had minimal impact on PIA when involving only index cases and their immediate partners (2A) or including wave-2 partners (2D). However, PIA increased significantly when previously diagnosed individuals were included (2B), reaching up to 60% when ART reengagement for these individuals was also maximized (2C).

In the US, HIV partner services is instrumental for linking persons diagnosed with HIV to care while facilitating testing, prevention, and care services for their sexual partners,^4,6,19,21 However, the population-level impact of HIV-PS on HIV incidence may remain limited without major enhancements.⁸ In this study, we evaluated the potential benefits of enhancing the HIV-PS framework among GBMSM by including out-of-care individuals with diagnosed HIV (PWDH) as a priority population in HIV-PS and extending partner elicitation beyond index case’s (“Wave-1”) named partners. While extending partner elicitation had minimal impact on the partner pool, incorporating PWDH who were out-of-care substantially increased both index-case and partner pools, thus broadening the participant base eligible for downstream transmission-disrupting services. However, translating this participant pool expansion into significant HIV incidence reduction required further enhancements to ART and PrEP delivery within HIV-PS. Calls to prioritize reengagement of out-of-care PWDH in HIV care and prevention services, including HIV-PS, are not new and have been made over the past decade.^33–35 While prior studies have been largely descriptive or qualitative, our study quantifies the potential impact of incorporating out-of-care PWDH into HIV-PS, particularly when paired with targeted improvements in ART reengagement. By doing so, we provide actionable insights for operationalizing these enhancements within HIV-PS systems.

Targeted intervention approaches maximize benefits by directing resources to those most in need for HIV-transmission disrupting services.^36,37 In our study, prioritizing out-of-care PWDH alongside newly diagnosed HIV cases for HIV-PS significantly expanded the pool of eligible index-case and partner pools available for these services, suggesting this enhancement may be an effective approach to reaching a wider segment of the target population with HIV transmission-disrupting services compared to current HIV-PS strategies. Additionally, it fosters a collaborative approach to HIV control by emphasizing the importance of sustained engagement with PWDH to achieve meaningful population-level reductions in HIV incidence.¹¹

CDC’s annual HIV testing data indicating that about half of all positive CDC-funded HIV tests in the US are among previously diagnosed individuals support the feasibility of prioritizing PWDH in HIV-PS.³⁸ Retesting PWDH, which typically occurs during transitions in medical care, could serve as a trigger for routine linkage to HIV-PS for these individuals, similar to current practice for newly diagnosed cases. Similarly, focusing on individuals with new STI diagnoses is a practical strategy, as they are already engaged with healthcare and may have higher transmission risk behaviors. By integrating HIV-PS into STI care services, health departments could streamline reengagement efforts, reduce barriers, and enhance outreach to individuals at heightened risk of onward transmission. Other practical approaches include pharmacy-based strategies, where lapsed ART prescriptions help identify and reengage out-of-care PWDH.³⁹ Emergency departments and urgent care settings, where many disengaged individuals seek care, also offer key opportunities through automated HIV-PS and ART reengagement prompts to facilitate timely linkage. These practical approaches circumvent challenges related to data accuracy and timeliness experienced by prior data-led efforts to reengage out-of-care PWDH^40–42, as it leverages individuals already in contact with the healthcare system. Other actionable steps include expanding access to low-barrier HIV care, offering incentives for out-of-care PWDH to reengage, and potentially reserving slots for out-of-care PWDH at local public health programs.^43,44 Such measures could enhance participation and reengagement in HIV care.

While expanding reach to a broader segment of the target population holds advantages, enhancing the delivery of and engagement in HIV diagnosis, treatment and prevention within the HIV-PS framework is crucial to achieving significant population-level reductions in HIV incidence, especially in the US South. Our previous study highlighted the stark drop-off from the number of GBMSM engaged in upstream HIV-PS cascade steps to those benefiting from downstream services such as ART and PrEP, which directly mitigate HIV incidence, as a key limitation of traditional HIV-PS impact.⁸ This study underscores that expanding the pool of HIV-PS participants available for these downstream services through targeted program expansions, such as prioritizing PWDH in HIV-PS, would similarly fail to tangibly impact HIV incidence if not coupled with enhanced engagement in these services within HIV-PS, particularly ART reengagement for PWDH. Addressing PWDH reengagement in ART within HIV-PS was projected to substantially increase the effectiveness of HIV incidence reduction efforts and enhance the overall efficiency of HIV-PS as a tool for HIV control. Achieving increased inclusion of PWDH into local HIV prevention efforts may involve using all instances of HIV retesting as a trigger for initiating partner services, collaborating with community organizations to encourage participation in partners services for PWDH, and utilizing existing surveillance systems to re-engage PWDH to both ART and HIV-PS for partner elicitation and linkage to relevant services. Introducing secure telemedicine in HIV-PS also holds promise for supporting re-engagement among PWDH, particularly those facing healthcare system fatigue or barriers to engaging with in-person services, potentially improving their HIV-PS participation rates.

Several limitations affect the interpretation of our findings. First, the parameters used in the stochastic model were derived from pre-COVID-19 surveillance data and empirical research, neglecting the impact of the COVID-19 pandemic on both sexual behaviors among GBMSM and HIV-related healthcare disruptions.^45,46 However, current evidence suggests that any such impacts may be minimal, as behavioral and service parameters have largely returned to pre-pandemic levels.⁴⁷ Second, our model assumed consistent behavioral responses over time and did not account for potential changes in risk behaviors following HIV diagnosis.^14–16,48 Addressing these nuances is crucial for accurately predicting outcomes post-HIV-PS participation among GBMSM, yet empirical data in this rapidly evolving landscape, marked by increasing PrEP and ART use, remain limited. Third, our study focused solely on Black, White, and Hispanic racial groups, excluding other races and ethnicities, which limits the generalizability of our findings beyond these racial groups. Finally, our study focused on Atlanta, a metropolitan area with unique population demographics, HIV prevalence, and HIV-PS system. Extrapolating our findings to cities, whether with more or less established HIV-PS systems, necessitates careful consideration of local contexts, population dynamics, and healthcare infrastructure variations.