Self-reported social contacts and the risk of influenza infection during a pandemic

Abstract

BACKGROUND: Variability in social behaviour such as the propensity to meet individuals and mix in groups must contribute to heterogeneities in the transmission of influenza and other respiratory pathogens, with data from questionnaire-based studies of these encounters now used routinely to inform age-specific parameters of transmission models. Although results from these models are being used with increasing frequency to inform the prioritisation of key resources such as vaccines, the association between self-reported social contacts and infection has been quantified only for populations, not for individuals. METHODS: We embedded a questionnaire within a longitudinal cohort survey of influenza to test the strength of association between self-reported contacts and serologically confirmed infection. We extracted variables related to individual contacts and the locations at which contacts were made. RESULTS: Location and contacts were only weakly correlated. Using logistic regression, we found overwhelming support for models that included age over those that did not across a large set of social contact variables. The best models all included location variables. Estimated increases in risk of infection were ≈1% per location and ≈15% per contact. Although risk estimates were only marginally significant, both were consistent across many alternate models. CONCLUSIONS: Self-reported social contacts did not explain age-specific patterns of infection. Therefore, pending additional empirical results, theoretical studies of interventions against respiratory pathogens should not assume a strong relationship between self-reported contacts and force-of-infection without appropriate sensitivity analyses. Within age groups, a potentially strong association between number of locations and risk of infection suggests that many infections occur via the aerosol route. Also, if confirmed, the link between number of locations and infection could help improve plans to reduce the impact of a moderate or severe pandemic by precisely characterising a novel modifiable risk factor.