A hybrid simulation model of HIV program interventions: from transmission behavior to macroeconomic impacts.

Metadata-grounded summary

Citation abstract

BACKGROUND: Considerable progress has been made in the estimation of drug safety treatment effects-particularly with observational medical claims and electronic medical record data. The use of the Target Trial framework, along with developments in statistical methodology such as doubly robust and G-estimation methods, has improved the ability to draw reliable causal inferences about drug and vaccine safety from observational data. However, such models rarely relate drug safety outcomes to other domains such as economic impacts. As shown by the COVID-19 pandemic, this is a significant limitation with potentially serious and long-term consequences.

OBJECTIVE: The goal of this paper is to demonstrate that the availability of simulation tools would enable policy-makers to assess drug safety and effectiveness outcomes associated with alternative policies, as well as examine these effects in the context of other domains, such as economic outcomes.

DESIGN: We develop an agent-based simulation model (ABM) of a peer navigator program to support engagement in HIV therapy in Tanzania. Results from the ABM are weighted to reflect the Tanzanian population and fed into the SPECTRUM model. This generates detailed demographic forecasts that are translated into macroeconomic impacts using labor force participation rates from the International Labor Organization, along with an econometric model of gross domestic product.

RESULTS: The ABM simulation estimated that the peer navigation program increased ART participation for men and women by about 12%-15% with no strong trend over time. The impact on VLS, however, was cumulative and significant for both men and women. By year 3, VLS was improved by 33.9 percentage points for women and 32.6 percentage points for men. However, the overall impact of these estimates on mortality was modest-ranging from less than 500 lives per year at the start of the forecast period to about 2500 lives per year in 2030. Consequently, the associated macroeconomic impacts were also small. The relatively modest impacts were due to the limited opportunity for HIV control in Tanzania, which had already met its 95/95/95 goals.

CONCLUSION: Although the simulated macroeconomic effects of the peer navigator program in Tanzania were modest, the paper demonstrates the feasibility of linking behavioral ABM simulations of program impacts to subsequent demographic effects and, finally, macroeconomic performance. Moreover, given the clinical response in ART and VLS in the exposed population in Tanzania, it is likely that the same peer navigator intervention conducted in another country with a larger at-risk HIV population would be much larger.

Simulating the effects of a peer navigation program on antiretroviral treatment and outcomes in Tanzania Pharmacoepidemiology and drug utilization studies using data from medical claims or electronic medical records are important tools for understanding the performance of drugs and other interventions in real world practice. Typically, however, such studies focus on treated patients and do not address the behavioral aspects of how patients decide upon treatment in the first place. We use an agent-based simulation model to estimate the impact of a peer navigation program in Tanzania on antiretroviral therapy (ART) participation and adherence, as well as viral load suppression (VLS) for HIV patients. We then estimate the impacts of the program on Tanzanian mortality rates by weighting the patient level estimates to reflect population estimates. Finally, we evaluate the impacts of estimated mortality changes on the productive capacity of the labor force and economic growth. The agent-based simulation finds that the peer navigation program had substantial positive impacts on ART participation for men and women, as well as VLS. However, the ultimate impacts of these effects on mortality and associated economic outcomes were relatively small. This appeared to be due to the relatively small size of the at-risk population who could potentially benefit from the intervention. To our knowledge, the current study represents the first successful linkage of a behavioral agent-based simulation model of a behavioral health policy intervention with its demographic implications and subsequent macroeconomic ramifications. Although the simulated effects of our study focused on a peer navigator program, rather than a drug per se, the outcomes measured utilization of ART and VLS. The ability to incorporate behavioral aspects of drug treatment patterns through agent-based simulation offers the potential to better understand drug treatment safety and effectiveness, as well as how such outcomes might be improved with augmentative approaches such as peer navigation programs.