A computer simulation used by public health officials to plan pandemic response overlooks the social disorder likely to follow such an event, suggest researchers this week in the Bulletin of the World Health Organization. Experts say evidence on how social groups might react is lacking, but even without them models can still be a useful tool in the early stages of an outbreak.

 

“The unpredictability of social order during disasters was not adequately addressed by simulation methods,” write Toomas Timpka, of Linköpings University in Sweden, and colleagues. “[E]ven minor disruptions of the social order may invalidate key infrastructural assumptions underpinning current pandemic simulation models.”

 

Computer models can help public health officials understand what effect interventions might have on the spread of an influenza virus during a pandemic. The simulations are designed to be flexible, enabling policy makers to replace assumptions with real data once they become available during an outbreak.

 

To get a better understanding of the usefulness of these models, Timpka and colleagues ran a case-study simulation and invited nine experts, from different areas within the field of public-policy, to assess it. The model looked at what effect the distribution of antiviral drugs and public health interventions could have on the spread of influenza in an urban district of Sweden with a population of 140,000 people.

 

Predicting exactly how people will behave in the event of a pandemic is very difficult, say Timpka et al. During the outbreak of severe acute respiratory syndrome (SARS) in Toronto, disruptions to social order meant that quarantine compliance rates fell to 57%, they point out.

 

Alessandro Vespignani, from the Indiana University School of Informatics in the USA, says the public’s perception of risk during a major event is important to consider when modelling disease patterns in a population. “[But] we don’t have any data or quantitative research that shows how society will change [in the event of a pandemic],” he stresses.

 

How extensively social disruption might limit public health responses will depend on the country hit by a pandemic, Timpka tells EHTF News. “In certain parts of the world the population are very aware of the possibility of outbreaks, so the risk for social disorder and disruption is very low.”

 

By contrast, in a country like Sweden there is a much larger risk that current response plans will not function effectively in the event of a pandemic. This is because key public-sector workers such as nurses, teachers, and bus drivers might not turn up to work, he says. “It won’t be chaos, but the point is that calculations based on an institutional society will not be the case in real life, therefore [predictions generated by models] will not be true.”

 

Vespignani agrees that people in different parts of the world will react differently to a pandemic, but feels models can still be useful — so long as accurate data are available to describe the success of interventions and other aspects as the incident unfolds. Once information is known about any disruptions to the antiviral drug distribution system, for example, it can be entered into the model for a quick update of the forecast.

 

The difficulties have more to do with real-time acquisition of data, explains Vespignani. “There are so many different factors at play when forecasting, we cannot just blame the models themselves.” The infrastructure to provide the needed data may not exist in poorer countries, he highlights.

 

But according to Timpka and colleagues, updating the models with fresh data may not happen fast enough. It took five hours to update the case-study model when the researchers keyed in a new value to replace the theoretical number of people who could catch the virus from one infected person, they point out.

 

This relationship between model flexibility and calculation time may put off policy-makers from using them at a national level in the early stages of a pandemic, when time is critical, say the authors. Vespignani believes it’s possible to find a trade-off between computational efficiency and flexibility in modelling. Scientists can update a model with more accurate information as soon as the data arrive, and in some cases almost in real-time, he suggests. “Models provide another instrument for policy makers, another piece of information that might prove to be very useful.”

 

In the case of influenza, Timpka says that more empirical evidence is needed to make the models more relevant. Research into how influenza transmission really occurs during seasonal outbreaks will yield good baseline data that can inform pandemic models, he suggests.

 

It would help to look at how the disease spreads between groups of people of various ages, ethnicities, and areas of residence, according to Timpka. Studying social networks that people form on a daily basis would also add valuable information, he says. “Reliable data on this topic is one of the key areas that will make [pandemic influenza] models better.”