Influenzanet is a system to monitor the activity of influenza-like-illness (ILI) with the aid of volunteers via the internet

http://www.influenzanet.eu/

Epiwork Logo
Developing the framework for an epidemic forecast infrastructure.
http://www.epiwork.eu/

The Seventh Framework Programme (FP7) bundles all research-related EU initiatives.

7th Framework Logo
Participating countries and volunteers:

Participating countries and volunteers:

The Netherlands 12026
Belgium 3955
Portugal 1754
Italy 4331
Great Britain 4594
Sweden 6702
Austria 527
Switzerland 148
France 1
Spain 997
Ireland 296
Denmark 1417
InfluenzaNet is a system to monitor the activity of influenza-like-illness (ILI) with the aid of volunteers via the internet. It has been operational in The Netherlands and Belgium (since 2003), Portugal (since 2005) and Italy (since 2008), and the current objective is to implement InfluenzaNet in more European countries.

In contrast with the traditional system of sentinel networks of mainly primary care physicians coordinated by the European Influenza Surveillance Scheme (EISS), InfluenzaNet obtains its data directly from the population. This creates a fast and flexible monitoring system whose uniformity allows for direct comparison of ILI rates between countries.

Any resident of a country where InfluenzaNet is implemented can participate by completing an online application form, which contains various medical, geographic and behavioural questions. Participants are reminded weekly to report any symptoms they have experienced since their last visit. The incidence of ILI is determined on the basis of a uniform case definition.

Hide this information

Help

ILI case definitions

The following case definition for influenza-like illness is recommended by the ECDC. From season 2011-2012 on, all countries participating in Influenzanet will use the same questionnaire and will apply this case definition.

ILIecdc - ECDC recommendation
  • Sudden onset of symptoms
  • At least 1 of: Fever or feverishness (chills); Malaise; Headache; Muscle pain
  • At least 1 of: Cough; Sore throat; Shortness of breath

Many national surveillance systems do not apply the ECDC recommendation, but rather apply an ILI case definition which requires fever, especially to distinguish between a real influenza infection and a common cold. We therefore introduce a variation of the ILIecdc definition which does require fever:

ILIfever - Requiring fever
  • Sudden onset of symptoms
  • Fever or feverishness (chills)
  • At least 1 of: Malaise; Headache; muscle pain
  • At least 1 of: Cough; Sore throat; Shortness of breath

The ILI case definition in use by the CDC is defined by:

ILIcdc - CDC recommendation
  • Fever (at least 37.8 degrees Celsius)
  • At least 1 of: cough or sore throat
  • Without a known cause other than influenza

During the first seasons of Influenzanet, participants could only report a fever if they measured their temperature, and not all symptoms were reported. Therefore, the ILIecdc, ILIfever and ILIcdcdefinitions cannot be applied. We introduce the following case definition, based on the ILIfever definition:

ILIhist - All (historic) seasons
  • Measured temperature ≥ 38 degrees Celsius
  • Sudden onset of symptoms or fever
  • At least 1 of: Headache; Muscle pain
  • At least 1 of: Cough; Sore throat

The ILIhist can be equally applied to all countries and seasons. However, care should still be taken when comparing different countries. Due to cultural differences, participants could have different customs whether to actually measure their temperature in case of a (suspected) fever.

Within the Influenzanet system, multiple ILI case definitions can be used simultaneously. ILIecdc has a higher sensitivity, since more participants with influenza will fit the definition. ILIhist has a higher specificity, since fewer participants who do not have influenza, will fit the definition. See for example: The Netherlands 2011/12.

Influenzanet can monitor more syndrome or diseases based on the reported symptoms.

Allergy Allergy / Hay fever
  • Not ILIecdc
  • Not fever or chills or Temperature ≥38C
  • At least 1 of: Runny or blocked nose; sneezing; watery, bloodshot eyes
  • Participant suspects symptoms due to allergy
Gastro - Gastroenteritis / Stomach flu
  • Not ILIecdc
  • At least 2 of: diarrhoea; nausea; vomiting; stomach ache
Cold - Common cold
  • Not ILIecdc
  • Not Allergy
  • At least 2 of: Runny or blocked nose; Sneezing; Cough; Sore throat

Back to top

Active participants

People who are experiencing ILI symptoms are more inclined to register in Influenzanet. This is especially the case when many new persons join during the initial phase of a new influenza epidemic, due to media attention. Therefore, the first survey of every participant, in which a participant reports symptoms he/she had in the week(s) before joining, is ignored (ignore_first = True). The first day a participant is considered active is therefore defined as the day following his/her first completed survey.

A participant is considered active until the day of his/her last completed symptoms questionnaire (always_active = False). Although it is possible that participants who stop reporting do so because they do not have any symptoms, we do not have any information about this.

Participants are encouraged by a weekly email reminder to complete their symptoms' questionnaire every week, whether they have any (ILI) symptoms or not. This allows the calculation of a reliable incidence, based on the ratio of participants with and without symptoms. Participants who do not report regularly are excluded from the analyses, defined as participants who have completed less than 3 symptoms' questionnaires (min_surveys = 3), as well as participants who report on average less than every 3 weeks (max_freq = 21).

Back to top

ILI activity

The (ILI) activity for a certain week (Monday-Sunday) is defined as the number of onsets divided by the number of participant-weeks.

Activity = Onsets / Person-weeks

A participant who has been active during the full week is counted as 1 person-week. A participant who either registered during the week, or who completed his/her last symptoms' questionnaire during the week, is counted as a fraction.

For determining the day of onset, we use the fever onset data if provided, else the symptoms onset date if provided, and finally the reporting data if neither the symptoms nor the fever onset date is provided (onset = "fever").

Onsets which occurred before a person registered in Influenzanet and which are reported in the first survey, are ignored (ignore_first = True). For a participant who fits the case definition during two consecutive surveys, only the first onset is considered (ignore_doube = True). Note that if both the first survey and consecutive surveys are excluded, the onset of a participant who fits the case definition in both his/her first and second survey, is also excluded.

Back to top

Real-time reporting

In real-time reporting, participants continuously report symptoms that could have started in the past. The determined (ILI) activity for the most recent days therefore becomes more accurate when new information becomes available.

Back to top

Age correction

It is possible to do age correction on the determined (ILI) activity. The corrected (ILI) activity in the population is determined by:

(ILI) activity = Sum( fraction(age_group) * activity(age_group) )

where the fraction each age group represents in the population is take from Eurostat. The default age distribution is: 1-14, 15-64, and 65-100 (age_correction = 1-15-65-100).

Age correction is util to compensate for underrepresented groups. However, in Influenzanet the underrepresented groups could be so small, that within such a group there is no clear (ILI) signal. By doing age correction, these underrepresented noisy signals get amplified, leading to a more noisy signal in the age corrected ILI activity. Therefore, by default no age correction is applied.

Back to top

Attack rate

The (ILI) attack rate is defined as the percentage of all active participants who had an onset during a season.

Attack rate = All onsets / All active participants

Back to top

Plots

For plotting the (ILI) activity, the weekly activity (Monday-Sunday) is plotted on Sundays. The determined weekly activity can either be joined by straight lines (daily=False), or the weekly activity can be determined for each day, where for example on Fridays the weekly activity for the period Saturday-Friday is plotted (daily=True)

For graphs which compare various subgroups, these subgroups can be so small, which leads to very noisy incidences. Therefore, the incidences in those graphs are a moving average over three weeks.

When comparing the ILI activity determined by Influenzanet with the ILI activity determined by the medical doctors reporting to the ECDC, the absolute values could differ significantly. These differences could be attributed to the different case definitions used by medical doctors in different countries, as well as the different rate at which people with ILI symptoms visit their doctor. Also, there could be a selection bias in the people who decide to participate in Influenzanet, as well as a selection bias in people who decide to visit a GP. The trends in ILI activity as measured by both systems are however similar, which can be illustrated by plotting ILI activity from both system in the same diagram on different scales. To determine the scaling between the axes of both system, a linear regression between the ILI activity of both systems is done, and the "gradient" and "intercept" are defined, such that:

ILI (Influenzanet) = "gradient" * ILI (EISN) + "intercept"

The coefficients are based on the data for all previous seasons combined (linreg_seasons = "2003,...,2010"). We fix the intercept at 0 (plot.linreg_intercept0 = True), such that Influenzanet=0 always corresponds to EISN=0. Leaving the intercept free, in general shows that Influenzanet has a relatively higher baseline than EISN. However, plots where the origins of the Influenzanet and EISN axes are not aligned, such that an ILI activity of zero in Influenzanet corresponds to a negative ILI incidence in EISN, leads to confusing graphs.

Back to top

Extra notes

From 24 June 2009 - 24 October 2009, Influenzanet did not registered GP visit reports for participants in The Netherlands and Belgium. From 2 December 2009 - 8 January 2010, Influenzanet did not register GP visits and "Stayed home" for participants in Portugal. These periods are excluded from the calculation of the GP and stayed home rate.

Back to top

Extra cold definitions

Some extra cold definitions for testing.

Coldorig - Common cold definition in 2011/12
  • No sudden onset of symptoms
  • No fever
  • At least 2 of: Runny or blocked nose; Sneezing; Cough; Sore throat
Coldari - Acute respiratory infection
  • Not ILIecdc
  • Not Allergy
  • Sudden onset of symptoms
  • At least 1 of: Runny or blocked nose; Cough; Sore throat; Shortness of breath
Coldno sneeze - Exclude sneezing
  • Not ILIecdc
  • Not Allergy
  • At least 2 of: Runny or blocked nose; Cough; Sore throat
Coldshort - Include shortness of breath
  • Not ILIecdc
  • Not Allergy
  • At least 2 of: Runny or blocked nose; Sneezing; Cough; Sore throat; Shortness of breath
Coldno sneeze,short - Exclude sneezing, include shortness of breath
  • Not ILIecdc
  • Not Allergy
  • At least 2 of: Runny or blocked nose; Cough; Sore throat; Shortness of breath

Back to top

Epiwork