IDM’s HIV team provides modeling and quantitative analysis to support governments, NGOs, and researchers working to reduce HIV burden and transmission in generalized epidemic settings.
A majority of modeling questions that we receive from partner governments and NGOs concern the optimization and affordability of antiretrovirals for HIV treatment and prevention. For example, we have teamed with Health Economics and Epidemiology Research Office (HE2RO) in Johannesburg to produce treatment scale-up cost and impact estimates for South Africa’s Ministry of Health, and are working to improve these estimates in areas including geospatial distribution of care delivery, orphans and vulnerable children (OVCs), and the role of the private sector. We also participate in collaborations coordinated by the HIV Modelling Consortium, such as their model comparison study performed for the World Health Organization (WHO) ahead of the 2013 revision of the WHO HIV treatment guidelines. Other collaborations focus on emerging biomedical approaches, such as pre-exposure prophylaxis (PrEP) and sustained remission. For example, we are collaborating with CONRAD to model the impact of microbicide-based HIV prevention at the individual, couple, and population level. Unlike PrEP, evidence about the health and prevention benefits of antiretroviral therapy (ART) is relatively plentiful, and most of the questions from our collaborators concern the timecourse of implementation and the model of care delivery. To this end, we are developing a flexible model input framework that defines the properties of the HIV “cascade of care” as an external model input file. The steps toward receiving treatment, delays and retention rates at each step, and possibility of overlapping or branching pathways to care will be configurable in an input file, allowing them to be altered without changing or re-compiling the model code. These features will be available in our 2015 software release.
Modeling trends in HIV epidemics requires reconciling and triangulating between diverse and imperfect datasets. These data include risk behavior, prevention activities, prevalence, diagnosis, treatment, and AIDS-related mortality. Sources from which we draw include health program statistics, medical laboratory data, clinical trials, census or cluster-randomized survey results, demographic and actuarial estimates, demographic surveillance, and production and purchasing of biomedical products. We have found that age- and gender-disaggregated data can be especially informative about epidemic dynamics when interpreted using a mechanistic, age-structured model. Fitting diverse data sources allows users to pressure-test model assumptions and to identify inconsistencies or data gaps to be prioritized. Antenatal clinic (ANC) based HIV surveillance and prevention of mother-to-child transmission (PMTCT) programs constitute a particularly important data source for HIV prevalence, especially where nationally representative HIV serosurveys are infrequent. However, extrapolation of epidemic trends from antenatal data is challenging because of the complex interplay of age, partnerships, sexual activity, and contraceptive usage in determining HIV risk in pregnant and non-pregnant women. We are currently extending our mechanistic model to include the effects of sexual activity on fertility, as well as the effects of hormonal contraceptives on HIV susceptibility. Our goal is to fit age-dependent fertility estimates while accounting for sexual activity in both fertility and HIV exposure, allowing for more accurate extrapolation of antenatal data to full-population epidemic trends. These features would also enable studies of the role of hormonal contraceptives in HIV epidemic patterns, as well as more detailed outputs involving paternity, such as single and dual orphanhood.
Interrupting HIV transmission requires an understanding of the drivers of the epidemic. Why did HIV spread so widely in the general population in sub-Saharan Africa, but not in other geographies? To what extent are these drivers active in modern times? What would it take to collapse the chains of transmission that fuel the epidemic? What are the largest contributors to new HIV infections, and would targeting them reverse the spread of HIV in the general population? We have explored these questions using a network transmission model that fits behavioral patterns of partnership formation as well as biological patterns of infectiousness and disease progression. We have developed a novel algorithm to ensure that our age-structured network model can fit any measured age pattern of partnerships, including multiple categories of partnership types. Using this model, we have investigated the age patterns of transmission, such as the proportion of transmission chains that ‘age out’ of the population, compared to those that infect younger individuals and perpetuate the epidemic. We have also explored the impact and cost-effectiveness of age-targeted interventions, as well as interventions targeted to high-risk populations such as migrant workers. Lastly, we have explored biological aspects of HIV risk, challenging commonly used models of infectiousness during acute HIV infection and correcting them to account for between-individual heterogeneity in susceptibility or infectiousness. These analyses provide insight into high-priority data gaps, structural reasons why models may disagree with each other or fail to predict epidemic trends, and potentially faster and more scalable approaches to interrupting HIV transmission.