Pandemic Flu: Hit hard, but should we hit early?

Image courtesy of CU Boulder Applied Math
Mathematical Biology Group

When a new pandemic flu strain emerges, as happened in Mexico in 2009, public health interventions such as social distancing are often implemented in the hope of reducing transmission. It is often assumed that the more measures that can be implemented early in the epidemic, the better. However, social distancing policies, while sounding simple, are very costly and typically of finite duration.  It is also unclear, what the policy 'objectives' of the intervention should be. Is the aim to reduce total morbidity/mortality or is it to reduce peak prevalence? How are these aims to be balanced against the societal impact of the interventions? Heady questions indeed.  Fortunately, a new paper in PLoS Computational Biology by Deirdre Hollingsworth and colleagues (including Roy Anderson) poses many of these questions and attempts to explore the intertwined policy objectives of various pandemic flu mitigation strategies.

Using a mathematical model (deterministic SIR model) and assuming a mean infectious period of 2.6 days, an Ro=1.8 and a population size of 58 million (the population of the island of Great Britain), the authors estimated the impact of social distancing interventions on influenza dynamics. They looked at a three possible durations of the intervention: (1) indefinite (2) 12 weeks - the current US policy maximum and (3) until a pandemic vaccine becomes available at 6 months. The also incorporated the use of a limited stockpile of antiviral drugs to limit disease severity and reduce transmission and also looked at the utilization of a partially protective pre-pandemic vaccine during the first 6 months of the pandemic.

What did they find? Well, it is pretty messy, like the truth typically is. One finding that sticks out is that in long-term interventions, there is very little disease incidence before week five, so there is little overall benefit from starting social distancing policies too early. The authors estimate that a few weeks delay in implementing a long-term intervention may result in a higher peak prevalence, but a considerably shorter duration of the epidemic.

What about short-term (12-week) interventions, such as those recommended by the current US pandemic plan? In this scenario, strategies that might contain an epidemic size below a certain level would not be the same ones that could limit peak prevalence. In these scenarios there would be a second peak after the intervention is lifted and they may even result in almost no change in overall epidemic size. For these "short" duration interventions there are no easy answers. For example, if the social distancing intervention is 33% effective in reducing transmission and is started at week 5, it might minimize peak prevalence. However, the same intervention with 22% effectiveness and similar timing would be expected to minimize the epidemic size.

The authors describe and discuss various other scenarios and provide a multitude of estimates for what the effects might be.  However, I think their key contribution is to force us to confront very important policy questions head on.  We need to have discussions about what our policy objectives should be in a pandemic.  Should we aim to limit the total number of cases or should we care about the peak size of the pandemic when our hospital and other public health services are stretched to the limit?  Maybe we should just want the epidemic to end as quickly as possible, so that society can get back to the new normal? We should probably have these discussions before the next pandemic, so we can design the optimal policy to achieve our goal(s). 

Hollingsworth et al. PLoS Computational Biology February 2011