Given the deficiencies in HIV collection metrics, modernization and augmentation of the HIV surveillance system is necessary. In doing so, an HIV surveillance ecosystem (Fig. 1) can be built that accurately reflects the seropositive population, incorporates persons who were missed, provides community level access, and leverages technological advances to identify the most vulnerable individuals.

Fig. 1

Schematic of a modernized HIV surveillance system that leverages technological advances and community

Using Indirect Estimation to Assess Who is Missing

Traditional estimation methods to understand the undiagnosed population have limitations and challenges. Relying on individuals to be tested can give a sense of the prevalence of diagnosed HIV but may not provide an accurate estimate of the prevalence of people who are undiagnosed but living with HIV. Recognizing the inherent bias in direct estimates of prevalence, “indirect” methods that utilize information from a variety of data sources are increasingly useful for difficult-to-quantify diseases and conditions such as substance use disorders and homelessness. Among those ‘indirect’ methods include the basic multiplier method [30], the Bayesian Multi-Parameter Estimation of Prevalence (MPEP) method [31, 32], and the multiple sample estimation (MSE) method (previously known as ‘capture-recapture’) [33, 34]. Multiplier methods estimate the size of the population by generating a “benchmark”, the number of an event, and then applying a “multiplier” which is the reciprocal of the rate of this event among a specific population. MPEP is a highly flexible modelling approach that aims to synthesize as much relevant data as feasible to estimate prevalence, allowing model assumptions to be interrogated through assessment of consistency of evidence from different data sources. MSE involves fitting a statistical model to the observed overlap between multiple administrative databases. The model is extrapolated to estimate the number of people not observed in these databases, and hence overall prevalence. Such an approach has been used at the local and state levels to identify undetected HIV diagnoses [27].

Each of these approaches has limitation and strengths but can be used to estimate the overall prevalence (those with diagnosed HIV and those undiagnosed living with HIV) as well as specific trends in specific subsets of the population (e.g., transgender populations, people experiencing homelessness) who may not be diagnosed due to social or structural factors (e.g., lack of health insurance, language barriers, stigma). Having a more comprehensive picture of who is being “missed,” can inform resource allocation and facilitate the evaluation of penetration and reach of the deployed interventions.

An important consideration in implementing a new approach to HIV prevalence estimation is the costs and benefits. The more sophisticated approaches (i.e., MPEP and MSE, for example, require data that are linked at the individual-level). Such a process is resource intensive in terms of personnel, data sharing, and linkage software. Moreover, this approach also requires community engagement. While HIV surveillance has been broadly accepted by the community of individuals living with HIV, distrust remains, particularly in sharing additional personal data with government entities. A concerted effort is, therefore, required across local, state, and national public health entities to gain the trust of this community and involve them in the data collection process. In doing so, we might improve data granularity and accuracy to focus resources on the populations at the highest risk of infection.

Democratizing Testing to the Individual and Community

There are significant disparities in testing, as one in seven persons who have HIV are not aware of having a diagnosis [2]. These disparities exist at every step of the care continuum for racially minoritized populations, people who inject drugs, and people experiencing homelessness or incarceration. The system is not optimized for all people.

Expanding testing to non-traditional health care settings is known to increase case findings. Several studies have demonstrated the impact of testing availability at syringe-service programs, methadone programs, pharmacies, mobile health units, emergency departments, and jails and prisons [35,36,37,38,39]. As seen in the above example with D.C., democratizing testing to individuals and in the community (i.e., non-clinical settings) is a necessary next step.

Community-based testing methods may identify more new HIV diagnosis than traditional risk assessment testing. A study in South Africa used a novel expanded social network recruitment to HIV testing (E-SNRHT) to assess new HIV diagnosis [40]. Among the E-SNRHT group, newly diagnosed individuals were provided with educational material about HIV transmission and asked to recruit anyone they knew whom they thought would benefit from HIV testing [40]. The study compared the rates of new HIV diagnosis between the E-SNRHT intervention versus traditional risk network recruitment [40]. The authors found that E-SNRHT participants detected a significantly higher number of new HIV diagnosis (70.3% vs. 40.4%) and located significantly more previously-undiagnosed cases of HIV than risk network recruitment (rate ratio = 9.40; p < 0.0001) [40]. Testing at the community level is an effective intervention to bring more individuals with HIV into care and, thus, deepen our epidemiological understanding of HIV.

Furthermore, several studies have demonstrated that test availability at trusted or familiar community settings facilitates HIV screening uptake, particularly among marginalized populations. Church-based testing has been shown to reach people who have not been tested before and for whom health insurance status is a barrier to testing [41]. Barbershops have successfully partnered with local health agencies to offer free HIV testing [42]. Alcohol-serving establishments (e.g., bars) have demonstrated varying success at improving HIV testing and prevention [43, 44]. Same day rapid HIV tests as well as education, counselling, retesting recommendations, and referrals in these stigma-free community settings improves case detection and linkage to care.

Additionally, home-based self-tests have excellent sensitivity and specificity and provide individuals with the opportunity to test in private. A study of an HIV self-test distribution program between March 2023 to 2024 delivered 440,000 mail self-home tests to those who were at risk for HIV and found that 24.1% had never received testing and 1.9% had a positive HIV result [45]. As part of most home-based testing programs, specimen kits are mailed to an individual’s home and contain supplies to collect saliva or blood from a fingerstick. The kit is then mailed back to the lab. Within days, test results are returned to a clinician or laboratory. The clinician or laboratory will contact the individual if the home test was positive. Alternatively, rapid home tests may be collected at home and yield results within 40 min. In both cases, patients with positive results should have the opportunity to self-report their positive test results, receive access to confirmatory testing, and be referred for treatment. Many home test kits also include educational materials and resources if a test result returns positive. Due to advances in home-based testing made during the COVID-19 pandemic, opportunities to couple self-testing with telehealth has grown expeditiously and many HIV home and rapid tests provide resources for follow up telehealth visits. To fully realize the benefits of community-based and home-based testing, community partners and individuals must be able to conveniently report their results directly to the public health department. A modernized HIV surveillance system leverages community test sites, home tests, and rapid HIV tests to empower marginalized individuals.

Leveraging Technological Advances and Real-Time Informatics

Despite HIV risk factor and transmission knowledge, we continue to experience outbreaks such as those in Scott County, Indiana [46] Lowell and Lawrence, Massachusetts [47] and elsewhere which begs the question, “What are we missing?” We need to move beyond the concept that risk factors apply across populations and, instead, assess community-specific risk. Harnessing big data and using machine learning methods, which are designed to generate predictions and infer unknown risk factors can help us understand the interplay between patient-, provider-, and system-level factors as well as epidemiologic realities that influence and predict HIV-related risk in our heterogenous population [48, 49].

A recent study found that machine learning may lead to improved HIV testing [50]. The study found that among over 85,000 individuals, the machine learning algorithm was able to accurately identify the small percentage (< 3%) of individuals who were newly diagnosed with HIV [50]. Factors such as age at first STI diagnosis, previous STI history, and social vulnerability index were found to be predictors of incident HIV infection among males and females [50]. Such methods have also been applied for PrEP candidates to prevent HIV [51] as well as other vulnerable populations such as people leaving prison who are at risk for overdose [52]. Machine learning techniques can be leveraged to accurately detect HIV diagnosis among high-risk individuals and identify specific risk factors for HIV.

Justifiably, there is widespread skepticism of the use of artificial intelligence and machine learning in health care. The benefits of formally incorporating these methods into the HIV surveillance system need to be weighed against the ethical considerations. It is possible to realize the potential of artificial intelligence to improve the HIV surveillance system while addressing and acknowledging ethical concerns [53]. First, algorithms should not be implemented without robust systems and processes that protect communities. Second, it is crucial that algorithms do not reinforce biases, stigma, or disparities. Finally, we need to partner with community and advocacy organizations in the development and use of artificial intelligence. Borrowing from community based participatory research, the “nothing about us without us” should be applied here too.

Developing an Integrated System with no Wrong Door for Entry

As discussed, the current HIV surveillance system requires laboratories or clinicians to report to local public health agencies who then report to the CDC. Accurately estimating prevalence, increasing community-based testing and case identification, and leveraging technology requires an integrated system with multiple portals of entry. If we expect individuals and communities to embrace a new wave of analytics-based surveillance, they need to be involved in building the system and share a sense of ownership so that they know what it is, how it works, and who are the key players. As such, investments in public health to partner with the community, build capacity around the data, streamline and improve communication, and defragment the system are critical. Communities need to participate in the design, develop the methods, and institute the surveillance system to improve data outcomes. Creating an integrated, community-based system is not easy, but is necessary to achieve our EHE goals.