Whilst all available data (2004–13) were downloaded, analyses for this study were focused on the most recent five years (2009–13) as preliminary data analyses indicated that Google Trends data were not available prior to 2009 for numerous search terms (Figure 1; panels 2, 4, 9, 12, 16 and 17). Additionally, shifts in language are known to affect surveillance systems built upon textual data [4]. The shortened period (2009–13) was selected to minimise the effects of language shifts. However, this period still provides the requisite 50 pairs of observations for performing cross correlations [26].