As the eminence of the next influenza pandemic increases, so does the need for combined monitoring and modeling systems able to provide early quantitative predictions. An internet-based survey of influenza-like illness (ILI) – the Great Influenza Survey – was launched in the Netherlands and Belgium in the 2003/2004 influenza season (1). This innovative surveillance system is based on the voluntary online participation of the population who, on a weekly basis, respond to an internet questionnaire (3). The Dutch Great Influenza Survey has been carried out yearly in the Netherlands and Belgium. In Portugal it was implemented in the 2005/2006 season as “Gripenet”, and performed very well for its first year (5).
For the Dutch Great Influenza Survey, the representativeness of the population and ILI rates reported were validated by comparing the data collected through the internet system with data from the National Information Network of GPs (LINH) and the Dutch Sentinel Practice Network, respectively (2).
Within this system, we are also able to obtain real-time incidence data that, in conjunction with mathematical and computer models of transmission (9, 10), can be used as an early warning system to inform public health policy. The effectiveness of the system however, critically depends on the collection of representative and timely data.
Traditionally, influenza activity in Portugal, and in Europe as a whole, is monitored using a combination of clinical and virological information collected through a network of sentinel medical practitioners (GPs) and a network of reference laboratories. The data are gathered and compiled by the European Influenza Surveillance Scheme (EISS) (11, 12, 4). Comparison of the results from the internet-based system (1, 3) with the official Portuguese and EISS results are beginning to reveal important complementarities that this project intends to explore further (5).
The monitoring of ILI and influenza by EISS is well established and has proved to be reliable over the years. Moreover, the scheme functions as an important source of scientific information with regard to virological data and vaccine composition that are not possible in an internet-based surveillance system. The key role of the GPs in the EISS surveillance system ensures a high level of continuity and scrutiny, characteristics that remain to be proven with the internet-based system.
The internet-based system, relying on weekly questionnaires completed online, needs in average 4 days until reported incidence rates can be reliably deduced from real-time data, giving this internet-based approach a head start of 4 to 5 days compared to the information provided via sentinel GPs (2). The incidence curves determined by the internet-based system were qualitatively similar to those of EISS but considerably higher. According to the data collected in 2006/2007 the incidence of ILI in the Netherlands, Belgium and Portugal was approximately equal: 6.6% in the Netherlands, 6.1% in Belgium and 5.6% in Portugal, while according to EISS data the incidence in the Netherlands was 0.8%, 3.9% in Belgium, and 0.6% in Portugal (5). This is an important discrepancy whose investigation relies on the continuity of the internet-based data collection and further comparisons with the EISS data.
Moreover, the system includes a broad range of science communication and education activities and functions as an interactive enterprise. This is ensured by a task force that includes experts on virology, public health, mathematics, science communication, informatics, education and social sciences.
The use of the internet for surveys and health research has been previously considered, especially in association with themes likely to generate public anxiety, like bio-terrorist threats or anticipated outbreaks of pandemic infectious diseases (6). A major concern of web-based surveys is the non-representative nature of the participants and it is instinctively believed that results are likely to be biased. This belief, however, should be strongly contested and it has been argued that internet samples may actually be more representative that traditional samples (7, 8).
