Paper I establishes the benefits of linking epidemiological modeling with international health resource allocation decisions, reviewing the recent modeling literature on pandemic influenza control. The review indicates that outbreaks in resource-poor settings are controllable with moderate resource intensity and complexity of effort for viral strains of moderate infectiousness. However, very high resource allocations for preparedness in industrialized nations - at low geographic risk for the pandemic - are predicated on containment failure in countries at higher risk of outbreaks. Without assuming the infectiousness of a future flu virus, a redistribution of resources to the developing countries at primary risk reduces overall systemic risk of containment failure. The payoffs in terms of reduced global mortality and morbidity are higher with increased infectiousness.

The two other papers are associated with implementing the experimental desktop models for the context of India. Paper II first constructs a scenario-based non-epidemiological model of pandemic influenza introduction to, and subsequent spread within India under various assumptions. The model uses published data on attack rates in Asia during previous pandemics as well as seasonal influenza. The model exploits geographical risk variations across provinces of India as well as the provinces' demographics, transport networks, and rural-urban settings.

Paper III re-estimates the estimates of people living with HIV/AIDS (PLWHA) in India by combining the available prevalence data from the latest sero-surveillance data as well as the National Family Health Survey (NFHS-3) of 2005-06. The estimated total prevalence after accounting for biases in 2008 is 0.4%, within the bounds of the recent official announcement of 0.36% for year 2007 (0.27%-0.47%). The related 2.41 million PLWHA confirms that the prevalence had been overestimated by 230% till 2007.

The paper continues to comprehensively analyze antiretroviral (ARV) policy in India, beginning with the estimation of total costs of utilization under public and private market rates for first-line ART. A cohort simulation is conducted using a desktop model of disease progress in the population without access to ARVs. Deaths and morbidities are measured, and the availability of opportunistic infection prophylaxis is accounted for.

Cost-effectiveness (CE) is estimated based on years of life saved compared to no treatment. The desktop model results confirm cost-effectiveness of both one and two lines ART compared to no treatment, and CE values are within bounds of a recent stochastic, individual-based simulation (Freedberg et al. 2007), as well as prior studies. Advantages of the model—besides tractability and customizability—includes the inclusion of treatment failure and the ability to model physician decision to continue first-line treatment even on failure, a reality considered in India given the low availability of second-line treatment. Here, the model yields novel results: continuing first-line therapy till end of follow-up (the realistic treatment horizon) is cost-effective compared to ending first-line treatment at physician-determined treatment failure. Compared to Freedberg et al., PI-based second-line ART is not found cost-effective by WHO standards compared to NNRTI-based first-line therapy, at prevailing PI-combination drug costs. (Abstract shortened by UMI.)