Think about all the people you came into contact with today – your family members, friends, co-workers, jostling strangers on the bus, the barista who made your coffee. Chances are many of these were fleeting connections, but for infectious disease researchers, those contacts are essential for understanding how infections are transmitted in a population. Highly infectious diseases such as the cold or flu are often transmitted through a simple touch or through sneezes, coughs, and conversations, but for the most part researchers can only guess an individual’s daily contacts and thus can’t fully understand how these infections are passed along.
University of Washington Statistics doctoral student Gail Potter, working with VIDI members Dr. Ira Longini and Dr. Betz Halloran, wants to integrate data on social contact into models studying flu transmission. Using data from previous extensive surveys conducted in Belgium and surveys being conducted by the IRD, a research institute in Senegal, Potter hopes to model social contacts in different types of populations and use this knowledge to inform how the flu may spread in different communities.
“We really want to use statistics to understand the whole social network,” Potter said.
Her findings may then inform a larger project in Longini and Halloran’s group, a model that simulates influenza transmission in a population that the group has used to predict effectiveness of various interventions. Using the model, the group predicted that vaccination of school-aged children would go the farthest in stemming the novel H1N1 pandemic in a situation with limited vaccine supply.
Currently, the model makes best guesses on social contacts in the simulation, Potter said, but she hopes that integrating data of large-scale social contacts into the model may improve the model’s predictive powers. For example, the model currently hasn’t disentangled contact from disease transmission, Potter said, meaning that it calculates a single probability of contact plus transmission. In reality, each person will have one probability of contacting an infected person and then a different probability of catching the flu from that contact.
The surveys on social contact in Belgium have already been completed as part of the POLYMOD study, a survey of contact behavior carried out in 8 European countries. The survey asked 750 Belgian participants to keep a diary of all social contacts for one weekday and one weekend day. Some of the results were surprising, Potter said. For example, in households with two 0-5 year olds and two 19-35 year olds, her model estimates that the chance that all household members contact each other on a given day is only 50 percent.
In Senegal, Potter plans to analyze data from a similar social contact survey carried out in conjunction with an influenza vaccine trial underway there. The IRD conducts quarterly censuses of the population using community members as census takers. IRD and PATH are collaborating on a flu vaccine trial currently taking place in 20 Senegalese villages. The contact survey, designed by University of Washington Epidemiology graduate student Jonathan Sugimoto, asks people who come down with the flu where they went for the previous three days and who they contacted in that time, allowing the researchers to track the social contacts of infected persons.
One of the questions Potter’s work could address is why children are higher transmitters of the flu. It could be because they make more contacts in a day than adults do, or because their hygiene is worse, Potter said.
Potter plans to investigate whether the data from the two studies on social contact would change the outcome of the group’s flu simulation, and if so she will incorporate her models on social contact into the larger flu transmission model. Results on social contact could influence the estimated impact of small-scale flu interventions, Potter said, such as how to vaccinate a single school or small community if one or more of its members became sick.