Development of an Agent-Based Model to Investigate the Impact of HIV Self-Testing Programs on Men Who Have Sex With Men in Atlanta and Seattle

Modeling the impact of changing sexual behaviors with opposite-sex partners and STI testing among women and men ages 15-44 on STI diagnosis rates in the United States 2012-2019

Objective

To estimate the potential contributions of reported changes in frequency of penile-vaginal sex (PVS), condom use and STI screening to changes in gonorrhea and chlamydial diagnoses from 2012 to 2019.

Methods

An agent-based model of the heterosexual population in the U.S. simulated the STI epidemics. Baseline was calibrated to 2012 diagnosis rates, testing, condom use, and frequency of PVS. Counterfactuals used behaviors from the 2017-2019 NSFG, and we evaluated changes in diagnosis and incidence rates in 2019.

Results

Higher testing rates increased gonorrhea and chlamydia diagnosis by 14% and 13%, respectively, but did not reduce incidence. Declining frequency of PVS reduced the diagnosis rate for gonorrhea and chlamydia 6% and 3% respectively while reducing incidence by 10% and 9% respectively. Declining condom use had negligible impact on diagnosis and incidence.

Conclusion

Understanding how changing behavior drives STI incidence is essential to addressing the growing epidemics. Changes in testing and frequency of PVS likely contributed to some, but not all, of the changes in diagnoses. More research is needed to understand the context within which changing sexual behavior and testing are occurring.

Uptake of HIV testing and its correlates among sexually experienced college students in Southwestern, China: a Web-Based online cross-sectional study

BACKGROUND

The prevalence of human immunodeficiency virus (HIV) is becoming more common among college students in China. However, latest data on the prevalence and correlates of HIV testing among sexually experienced college students is rarely.

METHODS

An online survey was conducted among college students aged 18 years or older using multistage stratified cluster sampling from 16 colleges. Data on socio-demographic, HIV testing, HIV-related awareness, attitudes, sexual education and behaviors were collected. Propensity score matching (PSM) and logistic regression model were used to identify factors associated with HIV testing.

RESULT

A total of 108,987 students participated the survey, of which 13,201 sexually experienced college students were included in this study. 1,939 (14.69%) college students with sexual experience reported uptake of HIV testing in the preceding year. The uptake of HIV testing increased for college students with a rising HIV knowledge score and sexual health knowledge. Being awareness of HIV-related knowledge (aOR = 1.15, 95%CI: 1.01-1.30), accepting one-night stands (aOR = 1.16, 95%CI:1.03-1.32), obtaining satisfactory sexual interpretation from parent(s) (aOR = 1.24, 95%CI: 1.07-1.43), ever had unintended pregnancy (aOR = 1.78, 95%CI: 1.32-2.38), ever had received HIV-related preventive service(s) (aOR = 1.37, 95%CI: 1.10-1.70), ever had participated HIV-related preventive services (aOR = 3.76, 95%CI: 2.99-4.75) and ever had anal sex (aOR = 2.66, 95%CI: 2.11-3.34) were positively associated with uptake of HIV testing. However, accepting premarital sex (aOR = 0.76, 95%CI: 0.66-0.88), accepting cohabitation (aOR = 0.75, 95%CI: 0.61-0.92), occasionally discussing sex with parent(s) (aOR = 0.68, 95%CI: 0.50-0.91), and being with moderate satisfaction of school sex courses (aOR = 0.74, 95%CI: 0.58-0.95) were negatively associated with uptake of HIV testing.

CONCLUSION

The prevalence of HIV testing was relatively low. Participation in HIV-related services and high-risk sexual behaviors were important enablers for testing. Improving sex education for students, increasing HIV preventive services on campus, and improving family sex education are necessary to increase HIV testing among college sexually experienced students.

Calibration and Validation of the Colorectal Cancer and Adenoma Incidence and Mortality (CRC-AIM) Microsimulation Model Using Deep Neural Networks

OBJECTIVES

Machine learning (ML)-based emulators improve the calibration of decision-analytical models, but their performance in complex microsimulation models is yet to be determined.

METHODS

We demonstrated the use of an ML-based emulator with the Colorectal Cancer (CRC)-Adenoma Incidence and Mortality (CRC-AIM) model, which includes 23 unknown natural history input parameters to replicate the CRC epidemiology in the United States. We first generated 15,000 input combinations and ran the CRC-AIM model to evaluate CRC incidence, adenoma size distribution, and the percentage of small adenoma detected by colonoscopy. We then used this data set to train several ML algorithms, including deep neural network (DNN), random forest, and several gradient boosting variants (i.e., XGBoost, LightGBM, CatBoost) and compared their performance. We evaluated 10 million potential input combinations using the selected emulator and examined input combinations that best estimated observed calibration targets. Furthermore, we cross-validated outcomes generated by the CRC-AIM model with those made by CISNET models. The calibrated CRC-AIM model was externally validated using the United Kingdom Flexible Sigmoidoscopy Screening Trial (UKFSST).

RESULTS

The DNN with proper preprocessing outperformed other tested ML algorithms and successfully predicted all 8 outcomes for different input combinations. It took 473 s for the trained DNN to predict outcomes for 10 million inputs, which would have required 190 CPU-years without our DNN. The overall calibration process took 104 CPU-days, which included building the data set, training, selecting, and hyperparameter tuning of the ML algorithms. While 7 input combinations had acceptable fit to the targets, a combination that best fits all outcomes was selected as the best vector. Almost all of the predictions made by the best vector laid within those from the CISNET models, demonstrating CRC-AIM's cross-model validity. Similarly, CRC-AIM accurately predicted the hazard ratios of CRC incidence and mortality as reported by UKFSST, demonstrating its external validity. Examination of the impact of calibration targets suggested that the selection of the calibration target had a substantial impact on model outcomes in terms of life-year gains with screening.

CONCLUSIONS

Emulators such as a DNN that is meticulously selected and trained can substantially reduce the computational burden of calibrating complex microsimulation models.

HIGHLIGHTS

Calibrating a microsimulation model, a process to find unobservable parameters so that the model fits observed data, is computationally complex.We used a deep neural network model, a popular machine learning algorithm, to calibrate the Colorectal Cancer Adenoma Incidence and Mortality (CRC-AIM) model.We demonstrated that our approach provides an efficient and accurate method to significantly speed up calibration in microsimulation models.The calibration process successfully provided cross-model validation of CRC-AIM against 3 established CISNET models and also externally validated against a randomized controlled trial.

Agent-based model projections for reducing HIV infection among MSM: Prevention and care pathways to end the HIV epidemic in Chicago, Illinois

Our objective is to improve local decision-making for strategies to end the HIV epidemic using the newly developed Levers of HIV agent-based model (ABM). Agent-based models use computer simulations that incorporate heterogeneity in individual behaviors and interactions, allow emergence of systemic behaviors, and extrapolate into the future. The Levers of HIV model (LHM) uses Chicago neighborhood demographics, data on sex-risk behaviors and sexual networks, and data on the prevention and care cascades, to model local dynamics. It models the impact of changes in local preexposure prophylaxis (PrEP) and antiretroviral treatment (ART) (ie, levers) for meeting Illinois' goal of "Getting to Zero" (GTZ) -reducing by 90% new HIV infections among men who have sex with men (MSM) by 2030. We simulate a 15-year period (2016-2030) for 2304 distinct scenarios based on 6 levers related to HIV treatment and prevention: (1) linkage to PrEP for those testing negative, (2) linkage to ART for those living with HIV, (3) adherence to PrEP, (4) viral suppression by means of ART, (5) PrEP retention, and (6) ART retention. Using tree-based methods, we identify the best scenarios at achieving a 90% HIV infection reduction by 2030. The optimal scenario consisted of the highest levels of ART retention and PrEP adherence, next to highest levels of PrEP retention, and moderate levels of PrEP linkage, achieved 90% reduction by 2030 in 58% of simulations. We used Bayesian posterior predictive distributions based on our simulated results to determine the likelihood of attaining 90% HIV infection reduction using the most recent Chicago Department of Public Health surveillance data and found that projections of the current rate of decline (2016-2019) would not achieve the 90% (p = 0.0006) reduction target for 2030. Our results suggest that increases are needed at all steps of the PrEP cascade, combined with increases in retention in HIV care, to approach 90% reduction in new HIV diagnoses by 2030. These findings show how simulation modeling with local data can guide policy makers to identify and invest in efficient care models to achieve long-term local goals of ending the HIV epidemic.

Agent-Based Modeling of Autosomal Recessive Deafness 1A (DFNB1A) Prevalence with Regard to Intensity of Selection Pressure in Isolated Human Population

An increase in the prevalence of autosomal recessive deafness 1A (DFNB1A) in populations of European descent was shown to be promoted by assortative marriages among deaf people. Assortative marriages became possible with the widespread introduction of sign language, resulting in increased genetic fitness of deaf individuals and, thereby, relaxing selection against deafness. However, the effect of this phenomenon was not previously studied in populations with different genetic structures. We developed an agent-based computer model for the analysis of the spread of DFNB1A. Using this model, we tested the impact of different intensities of selection pressure against deafness in an isolated human population over 400 years. Modeling of the "purifying" selection pressure on deafness ("No deaf mating" scenario) resulted in a decrease in the proportion of deaf individuals and the pathogenic allele frequency. Modeling of the "relaxed" selection ("Assortative mating" scenario) resulted in an increase in the proportion of deaf individuals in the first four generations, which then quickly plateaued with a subsequent decline and a decrease in the pathogenic allele frequency. The results of neutral selection pressure modeling ("Random mating" scenario) showed no significant changes in the proportion of deaf individuals or the pathogenic allele frequency after 400 years.

Effective strategies to promote HIV self-testing for men who have sex with men: Evidence from a mathematical model

BACKGROUND

HIV testing is the gateway to HIV treatment and prevention. HIV self-testing (HIVST) has potential to increase testing; however, the potential population-level impact of HIVST on the HIV epidemic and the best strategies for promoting HIVST are unknown. Our aim is to inform public health approaches for promoting HIVST as part of a comprehensive strategy to reduce HIV incidence.

METHODS

Stochastic network-based HIV transmission models were used to estimate how different HIVST strategies would affect HIV incidence in Seattle and Atlanta over 10 years. We included four types of HIV testers and implemented nine replacement and eleven supplementation strategies for HIVST.

RESULTS

Replacement of clinic-based tests with HIVST increased HIV incidence in Seattle and Atlanta. The benefits of supplementary strategies depended on the tester type using HIVST. Targeting non-testers averted the highest number of cases per test. In Seattle 2.2 (95%SI=-77, 100.4) and 4.7 (95%SI=-35.7, 60.1) infections were averted per 1000 HIVST when non-testers used HIVST once or twice per year respectively. In Atlanta the comparable rates were 8.0 (95%SI=-60.3 to 77.7) and 6.7 (95%SI=-37.7, 41.0). Paradoxically, increasing testing among risk-based testers using HIVST increased incidence.

CONCLUSIONS

The population-level impact of HIVST depends on who is reached with HIVST, how kits are used, and by characteristics of the underlying epidemic and HIV care infrastructure. Targeted HIVST can be an effective component of a comprehensive HIV testing strategy. More work is needed to understand how to identify and target non-testers for self-testing implementation.

Temporal Variation in One-Time Partnership Rates Among Young Men Who Have Sex With Men and Transgender Women

BACKGROUND

Volatility in sexual contact rates has been recognized as an important factor influencing HIV transmission dynamics. One-time partnerships may be particularly important given the potential to quickly accumulate large number of contacts. Yet, empirical data documenting individual variation in contact rates remain rare. This study provides much needed data on temporal variation in one-time partners to better understand behavioral dynamics and improve the accuracy of transmission models.

METHODS

Data for this study were obtained from a longitudinal cohort study of young men who have sex with men and transgender women in Chicago. Participants provided sexual network data every 6 months for 2 years. A series of random effects models examined variation in one-time partnership rates and disaggregated within and between associations of exposure variables. Exposure variables included prior number of one-time partners, number of casual partners, and having a main partner.

RESULTS

Results indicated substantial between-person and within-person variation in one-time partners. Casual partnerships were positively associated and main partnerships negatively associated with one-time partnership rates. There remained a small positive association between prior one-time partnerships and the current number of one-time partnerships.

CONCLUSIONS

Despite the preponderance of a low number of one-time partners, substantial variation in one-time partnership rates exists among young men who have sex with men and transgender women. Accordingly, focusing on high contact rate individuals alone may be insufficient to identify periods of highest risk. Future studies should use these estimates to more accurately model how volatility impacts HIV transmission and better understand how this variation influences intervention effectiveness.

Agent-based evolving network modeling: a new simulation method for modeling low prevalence infectious diseases

Agent-based network modeling (ABNM) simulates each person at the individual-level as agents of the simulation, and uses network generation algorithms to generate the network of contacts between individuals. ABNM are suitable for simulating individual-level dynamics of infectious diseases, especially for diseases such as HIV that spread through close contacts within intricate contact networks. However, as ABNM simulates a scaled-version of the full population, consisting of all infected and susceptible persons, they are computationally infeasible for studying certain questions in low prevalence diseases such as HIV. We present a new simulation technique, agent-based evolving network modeling (ABENM), which includes a new network generation algorithm, Evolving Contact Network Algorithm (ECNA), for generating scale-free networks. ABENM simulates only infected persons and their immediate contacts at the individual-level as agents of the simulation, and uses the ECNA for generating the contact structures between these individuals. All other susceptible persons are modeled using a compartmental modeling structure. Thus, ABENM has a hybrid agent-based and compartmental modeling structure. The ECNA uses concepts from graph theory for generating scale-free networks. Multiple social networks, including sexual partnership networks and needle sharing networks among injecting drug-users, are known to follow a scale-free network structure. Numerical results comparing ABENM with ABNM estimations for disease trajectories of hypothetical diseases transmitted on scale-free contact networks are promising for application to low prevalence diseases.

A large-scale location-based social network to understanding the impact of human geo-social interaction patterns on vaccination strategies in an urbanized area

Cities play an important role in fostering and amplifying the transmission of airborne diseases (e.g., influenza) because of dense human contacts. Before an outbreak of airborne diseases within a city, how to determine an appropriate containment area for effective vaccination strategies is unknown. This research treats airborne disease spreads as geo-social interaction patterns, because viruses transmit among different groups of people over geographical locations through human interactions and population movement. Previous research argued that an appropriate scale identified through human geo-social interaction patterns can provide great potential for effective vaccination. However, little work has been done to examine the effectiveness of such vaccination at large scales (e.g., city) that are characterized by spatially heterogeneous population distribution and movement. This article therefore aims to understand the impact of geo-social interaction patterns on effective vaccination in the urbanized area of Portland, Oregon. To achieve this goal, we simulate influenza transmission on a large-scale location-based social network to 1) identify human geo-social interaction patterns for designing effective vaccination strategies, and 2) and evaluate the efficacy of different vaccination strategies according to the identified geo-social patterns. The simulation results illustrate the effectiveness of vaccination strategies based on geosocial interaction patterns in containing the epidemic outbreak at the source. This research can provide evidence to inform public health approaches to determine effective scales in the design of disease control strategies.