Parastu Kasaie

Toward optimal resource-allocation for control of epidemics: an agent-based-simulation approach

Employing mathematical modeling and analytical optimization techniques, traditional approaches to the resource-allocation (RA) problem for control of epidemics often suffer from unrealistic assumptions, such as linear scaling of costs and benefits, independence of populations, and positing that the epidemic is static over time. Analytical solutions to more realistic models, on the other hand, are often difficult or impossible to derive even for simple cases, which restricts application of such models. We develop an agent-based simulation model of epidemics, and apply response-surface methodology to seek an optimum for the RA output in an iterative procedure. Validation is demonstrated through comparison of the results with the mathematical solution in an RA example for which the analytical solution is known. We apply the proposed approach to a more complicated RA problem in which a number of previous restricting assumptions are relaxed.

Simulation Optimization for Allocation of Epidemic-Control Resources

We consider the problem of resource allocation (RA) in the control of epidemics where a fixed budget is allocated among competing healthcare interventions to achieve the best health benefits, and propose a simulation-optimization framework to address a general form of the problem. While traditional approaches to the epidemic RA problem suffer from restrictive assumptions to facilitate exact analytical solutions, a simulation-based technique relaxes such assumptions and provides a more realistic representation of the epidemic. Coupling the simulation model with optimization techniques enables us to analyze the behavior of RA outcomes with regard to different investment strategies and seek optimal allocations. We discuss implementation steps and illustrate our approach for an RA problem in the control of influenza pandemic with several interacting healthcare interventions.

An agent-based simulation of a tuberculosis epidemic: understanding the timing of transmission

Tuberculosis (TB) transmission is a key factor for disease-control policy, but the timing and distribution of transmission and the role of social contacts remain obscure. We develop an agent-based simulation of a TB epidemic in a single population, and consider a hierarchically structured contact network in three levels, typical of airborne diseases. The parameters are adopted from the literature, and the model is calibrated to a setting of high TB incidence. We model the dynamics of transmission at the individual level, and study the timing of secondary infections from a single source throughout the duration of the disease. We compare the patterns of transmission among different networks and discuss implications. Sensitivity analysis of outputs indicates the robustness of the results to variations in the parameter values.

The Impact of Pre-Exposure Prophylaxis Among Men Who Have Sex With Men: An Individual-Based Model

Objectives: Preexposure prophylaxis (PrEP) is recommended for preventing HIV infection among individuals at high risk, including men who have sex with men (MSM). Although its individual-level efficacy is proven, questions remain regarding population-level impact of PrEP implementation. Design: We developed an agent-based simulation of HIV transmission among MSM, accounting for demographics, sexual contact network, HIV disease stage, and use of antiretroviral therapy. We use this framework to compare PrEP delivery strategies in terms of impact on HIV incidence and prevalence. Results: The projected reduction in HIV incidence achievable with PrEP reflects both population-level coverage and individual-level adherence (as a proportion of days protected against HIV transmission). For example, provision of PrEP to 40% of HIV-negative MSM reporting more than one sexual partner in the last 12 months, taken with sufficient adherence to provide protection on 40% of days, can reduce HIV incidence by 9.5% (95% uncertainty range: 8%–11%) within 5 years. However, if this could be increased to 80% coverage on 80% of days (eg, through mass campaigns with a long-acting injectable formulation), a 43% (42%–44%) reduction in HIV incidence could be achieved. Delivering PrEP to MSM at high risk for HIV acquisition can augment population-level impact up to 1.8-fold. Conclusions: If highly ambitious targets for coverage and adherence can be achieved, PrEP can substantially reduce HIV incidence in the short-term. Although the reduction in HIV incidence largely reflects the proportion of person-years protected, the efficiency of PrEP delivery can be enhanced by targeting high-risk populations.OBJECTIVES Pre-exposure prophylaxis (PrEP) is recommended for preventing HIV infection among individuals at high risk, including men who have sex with men (MSM). Although its individual-level efficacy is proven, questions remain regarding population-level impact of PrEP implementation. DESIGN We developed an agent-based simulation of HIV transmission among MSM, accounting for demographics, sexual contact network, HIV disease stage and use of antiretroviral therapy. We use this framework to compare PrEP delivery strategies in terms of impact on HIV incidence and prevalence. RESULTS The projected reduction in HIV incidence achievable with PrEP reflects both population-level coverage and individual-level adherence (as a proportion of days protected against HIV transmission). For example, provision of PrEP to 40% of HIV-negative MSM reporting more than one sexual partner in the last 12 months, taken with sufficient adherence to provide protection on 40% of days, can reduce HIV incidence by 9.5% (95% uncertainty range: 8-11%) within five years. However, if this could be increased to 80% coverage on 80% of days (e.g., through mass campaigns with a long-acting injectable formulation), a 43% (42-44%) reduction in HIV incidence could be achieved. Delivering PrEP to MSM at high risk for HIV acquisition can augment population-level impact up to 1.8-fold. CONCLUSIONS If highly ambitious targets for coverage and adherence can be achieved, PrEP can substantially reduce HIV incidence in the short-term. While the reduction in HIV incidence largely reflects the proportion of person-years protected, the efficiency of PrEP delivery can be enhanced by targeting high-risk populations.

Advancing global health and strengthening the HIV response in the era of the Sustainable Development Goals: the International AIDS Society—Lancet Commission

Author(s): Bekker, Linda-Gail; Alleyne, George; Baral, Stefan; Cepeda, Javier; Daskalakis, Demetre; Dowdy, David; Dybul, Mark; Eholie, Serge; Esom, Kene; Garnett, Geoff; Grimsrud, Anna; Hakim, James; Havlir, Diane; Isbell, Michael T; Johnson, Leigh; Kamarulzaman, Adeeba; Kasaie, Parastu; Kazatchkine, Michel; Kilonzo, Nduku; Klag, Michael; Klein, Marina; Lewin, Sharon R; Luo, Chewe; Makofane, Keletso; Martin, Natasha K; Mayer, Kenneth; Millett, Gregorio; Ntusi, Ntobeko; Pace, Loyce; Pike, Carey; Piot, Peter; Pozniak, Anton; Quinn, Thomas C; Rockstroh, Jurgen; Ratevosian, Jirair; Ryan, Owen; Sippel, Serra; Spire, Bruno; Soucat, Agnes; Starrs, Ann; Strathdee, Steffanie A; Thomson, Nicholas; Vella, Stefano; Schechter, Mauro; Vickerman, Peter; Weir, Brian; Beyrer, Chris

Guidelines for design and analysis in agent-based simulation studies

Agent-based simulation (ABS) continues to grow in popularity and in its fast-expanding application in various fields. Despite the increased interest, however, a common protocol or standard curriculum for development and analysis of ABS models hardly exists. As originally discrete-event simulation (DES) modelers, self-taught and still new to the world of ABS modeling, we have occasionally observed a gap between traditional simulation theory and current practices of ABS in the literature. This points to great unevenness among existing ABS applications in terms of concepts and design, quantitative and computational techniques used in analysis of models, as well as domain-specific issues in different fields. In this paper, we review a number of important topics and issues in the design and analysis of ABS models that deserve attention. Our discussion is supported by some illustrative examples from ABS models of disease epidemics, but its applicable to a fairly general class of ABS models.

Estimating the proportion of tuberculosis recent transmission via simulation

Tuberculosis (TB) is an infectious disease that can progress rapidly after infection or enter a period of latency that can last many years before reactivation. Accurate estimation of the proportion of TB disease representing recent versus remote (long ago) transmission is critical to disease-control policymaking (e.g., high rates of recent transmission demand more aggressive diagnostics). Existing approaches to this problem through cluster analysis of TB strains in population-based studies of TB molecular epidemiology are crude and prone to bias. We propose an agent-based simulation of TB transmission in conjunction with molecular epidemiologic techniques that enables study of clustering dynamics in relation to disease incidence, diversity of circulating strains, sampling coverage, and study duration. We perform a sequence of simulation experiments with regard to different levels of each factor, and study the accuracy of estimates from the cluster-analysis method relative to the true proportion of incidence due to recent transmission.

A Novel Tool Improves Existing Estimates of Recent Tuberculosis Transmission in Settings of Sparse Data Collection

In any setting, a proportion of incident active tuberculosis (TB) reflects recent transmission (“recent transmission proportion”), whereas the remainder represents reactivation. Appropriately estimating the recent transmission proportion has important implications for local TB control, but existing approaches have known biases, especially where data are incomplete. We constructed a stochastic individual-based model of a TB epidemic and designed a set of simulations (derivation set) to develop two regression-based tools for estimating the recent transmission proportion from five inputs: underlying TB incidence, sampling coverage, study duration, clustered proportion of observed cases, and proportion of observed clusters in the sample. We tested these tools on a set of unrelated simulations (validation set), and compared their performance against that of the traditional ‘n-1’ approach. In the validation set, the regression tools reduced the absolute estimation bias (difference between estimated and true recent transmission proportion) in the ‘n-1’ technique by a median [interquartile range] of 60% [9%, 82%] and 69% [30%, 87%]. The bias in the ‘n-1’ model was highly sensitive to underlying levels of study coverage and duration, and substantially underestimated the recent transmission proportion in settings of incomplete data coverage. By contrast, the regression models’ performance was more consistent across different epidemiological settings and study characteristics. We provide one of these regression models as a user-friendly, web-based tool. Novel tools can improve our ability to estimate the recent TB transmission proportion from data that are observable (or estimable) by public health practitioners with limited available molecular data.

Resource allocation for controlling epidemics: Calibrating, analyzing, and optimizing an agent-based simulation

This is a companion article to Kasaie and Kelton (2013), and provides an extended discussion on the calibration, analysis, and optimization of an agent-based simulation (ABS) model of an epidemic. The detailed information is presented for the illustrative case of a resource-allocation (RA) problem in the control of an influenza pandemic as described in Section 4 of Kasaie and Kelton (2013). The suggested protocol, however, can be adapted to address other instances of the RA problem in the context of other infectious disease epidemics. Section 1 defines all the parameters and run conditions for the ABS model. In Section 2 we discuss calibration of the RA objective function (DALYs function) and related sensitivity analyses. An investigation of variance reduction for the ABS model is presented in Section 3. A short discussion on application of mixture design and the corresponding triangular output displays for this design is discussed in Section 4. Finally, Section 5 presents the numerical results of the RA problem using a response surface methodology optimization approach.

What Will It Take to Reduce HIV Incidence in the United States: A Mathematical Modeling Analysis

Abstract Background The National HIV/AIDS Strategy has set ambitious goals to improve the epidemic in the United States. However, there is a paucity of usable program-level benchmarks tied to population-level epidemiologic goals. Our objective was to define tangible benchmarks for annual rates along the care continuum that are likely to translate to meaningful reductions in incidence. Methods We used a validated mathematical model of HIV transmission and care engagement to characterize care continuum parameters that would translate into 50% reductions in incidence by 2025, compared with a base case scenario of the current US care continuum. We generated a large pool of simulations in which rates of screening, linkage, and retention in care were varied across wide ranges to evaluate permutations that halved incidence by 2025. Results Among all simulations, 7% achieved a halving of incidence. It was impossible for our simulations to achieve this target if the annual rate of disengagement from care exceeded 20% per year, even at high rates of care reengagement. When retention in care was 95% per year and people living with HIV (PLWH) out of care reengaged within 1.5 years (on average), the probability of halving incidence by 2025 was approximately 90%. Conclusions HIV programs should aim to retain at least 95% of PLWH in care annually and reengage people living with HIV into care within an average of 1.5 years to achieve the goal of halving HIV incidence by 2025.