Introduction
The current Coronavirus (COVID-19) pandemic has become an Internet
phenomenon, leading newsfeeds and trending on news forums globally.
Understandably, there is widespread public interest, which is being met
by blanket media coverage of an unprecedented nature. The Internet is
now the favoured first port of call for those seeking healthcare
information (Diaz et
al., 2002; Andreassen et al., 2007). Therefore, such digital
information is likely to be playing a key role in public communication
during the current crisis.
Data generated though such Internet searching has long been known to be
useful for disease monitoring and surveillance
(Brownstein et al.,
2009; Eysenbach, 2011; Anema et al., 2014; Mavragani et al., 2018).
Resources, such as Google Trends, which provide data on the volumes of
Internet searching upon specific topics, have been identified as being
potentially useful sources of real time data
(Carneiro and
Mylonakis, 2009; Nuti et al., 2014). Such data sources may possibly
reflect disease occurrence quicker and more accurately than traditional,
but slower, disease monitoring through official channels. Studies
examining the relationship between Internet searching and disease
occurrence have become commonplace
(Carneiro and
Mylonakis, 2009; Mavragani et al., 2018). However, as recently
highlighted, although many studies describe relationships and seek
correlations, few studies use such data to its full potential utilising
it in disease forecasting and modelling
(Mavragani et al.,
2018). Additionally, whether relationships between Internet search data
and disease occurrence occur across national boundaries is rarely
examined; typically such studies examine such relationships within only
a single national country.
Thus, here the aim was not only to examine whether such correlations
between Google Trends data and COVID-19 cases occurred, but also to
utilise such data in modelling; could such data enhance traditionally
based models using reported case numbers? Additionally, were such
relationships apparent, and model enhancement occur, across a wider
geographical range than a single nation? Coronavirus is a pan-European
problem, with epidemics developing almost simultaneously across many
countries. This situation provides a unique opportunity to examine
whether such data can enhance modelling across multiple countries,
continent wide.