Contents © 2013 Flávio Codeço Coelho - Powered by Nikola
Sander van Noort
Instituto Gulbenkian de Ciência
Introduction. An important source of epidemiological data on infectious diseases is the number of infected persons over time. The incidence is usually collected based on clinical diagnosis of sentinel medical doctors, but the wide availability of Internet has given rise to new innovative methods. Influenzanet uses cohorts of self-reporting volunteers via the Internet and has been active in The Netherlands and Belgium (2003), Portugal (2005), Italy (2007), United Kingdom (2009), Sweden (2011), France (2012) and Spain (2012).
Method. Media attention is generated to encourage people (patients and non-patients) to enroll at one of the national websites by completing an intake questionnaire containing a set of demographic, medical, socio-economic and lifestyle questions. Subsequently, participants receive an e-mail on a weekly basis with a link to a short symptom questionnaire on eventual symptoms since their last visit. When symptoms are reported, some additional questions are presented about GP consultation and changes in activities. The incidence for influenza-like illness (ILI) is determined based on the percentage of reporting participants who fit a syndromic case definition.
Results. Based on the reports of tens of thousands of volunteers, the ILI incidence as reported by Influenzanet follows the same trends as the ILI incidence as reported by the sentinel network of medical doctors. In countries with many participants such as the Netherlands, Influenzanet has been able to detect the onset of an ILI epidemic on average two weeks earlier than the medical doctors.
Influenzanet is not only an independent source of ILI incidence, it can also expose biases in the incidence reported by the medical doctors. The rate at which participants with ILI visit their GP varies by country and by personal characteristics such as age, which causes relative over- and underestimation in the ILI incidences as reported by the medical doctors.
Based on the intake questionnaire, Influenzanet can also determine risk factors for ILI. The estimated risk factors are largely in correspondence with those found in previous studies, such as an increase in children, participants with chronic diseases and participants with children in their household. Influenzanet can also estimate risk factors which are normally not measured, such as an increased risk in participants with pets and the absence of a significant risk change in participants who travel by public transport. Based on the vaccination status of every participant, Influenzanet can also provide data on the vaccine efficiency for ILI. Since not all ILI is due to influenza, the determined vaccine efficiencies for ILI is lower than usually estimated for influenza.
Since Influenzanet is independent of a GP visiting rate and can apply the same ILI case definition in every country, the ILI incidences can be compared directly between countries. This allows the fitting of transmission models for influenza in multiple countries simultaneously and has led to new hypotheses that the burden of an ILI epidemic not only depends on the susceptibility of the population, but also increases when the epidemic unfolds during colder and dryer months.
Conclusion. Influenzanet provides a valuable new source for the ILI activity in the population. The integration of multiple data sources, self-reporting, medical doctors, weather data, online searches and social media, in combination with transmission models for influenza, can results in a much more detailed assessment on the spread of influenza through the population.