In 2009, there was a pandemic influenza outbreak of the pH1N1 swine flu. The United States declared a public health emergency on April 26, and the WHO declared it a pandemic on June 12 (pg 1).
The optimization model used in this article is based on which path of actions over time maximizes the number of people not infected in the first 12 months of the epidemic (pg 2). The algorithm used is an “Upper Confidence Bounds Applied to Trees” (UCT) algorithm.
To model transmission of the disease between travelers through cities, a random variable is drawn based on Binomial(N,
1-), where N is the number of infected travelers and S is the number of susceptible people in the city (pg 3).
The model tested 11 SNS actions (distribution of 0, 5, 10, 15, 25, or 50 million antiviral courses apportioned based on overall population or to current disease prevalence) over 12 months. A measure of waste, W, was included (as a 2-month half life) to account for expirations, misplacements, treatment of false positives (pg 3). Uptake is the rate at which people move from asymptomatic to symptomatic, thus requiring treatment by an antiviral course (pg 3).
A 31 million course release 2 months in, followed by 1 million/month thereafter (both proportionally adjusted), was used as an infinite supply benchmark (pg 4.)
$latex R_0$ around 1.6±.2, simple release policies (e.g. 5M or 10 M per month) worked as well as the infinite supply. Around 2.0, only 10M/month worked as well as the infinite supply, and at higher
values (2.4+) only optimized strategies could work as well as the benchmark (pg 6).
Dimitrov, Nedialko B., et al. “Optimizing Tactics for Use of the U.S. Antiviral Strategic National Stockpile for Pandemic Influenza.” PLoS ONE, edited by Benjamin J. Cowling, vol. 6, no. 1, Jan. 2011, p. e16094. DOI.org (Crossref), doi:10.1371/journal.pone.0016094.