The impact of influenza epidemics on public health is important, and in general, underestimated (1). Surveillance of this illness is essential and is usually composed of two different approaches. First, clinical surveillance is based on weekly reports of consultations for influenza-like illness (ILI) by sentinel practitioners. These data provide information on the intensity of the epidemics. Second, virological surveillance is based on a selection of practitioners who take nasopharyngal swabs from patients presenting with ILI for virus isolation. The presence of influenza virus is detected by cell culture combined with immunofluorescence reaction or polymerase chain reaction (PCR). Virus detection reveals the antigenic characteristics of the circulating strains. This information contributes to the choice of influenza vaccine composition, and allows early detection of new variants. To stress the importance of such surveillance, WHO has urged member states to contribute to Europe-wide heightened preparedness for epidemics and pandemics by strengthening national surveillance and laboratory capacity (2). In response to this, the European Influenza Surveillance Scheme (EISS, http://www.eiss.org/) was set up to increase the sensitivity of existing early warning systems and facilitate communication between European Networks, by sharing and exchanging epidemiological, virological, and clinical information on influenza (3-6).
Another aspect of the surveillance is to provide general practitioners with information on epidemics and thus assist their diagnostic and therapeutic decisions. Clinical diagnosis of influenza is correct in 65-85% of the cases when the epidemic is confirmed in the community (7).
Surveillance requires time for data analysis and communication to physicians. It can take several days from the detection of a significant increase of influenza activity to the day the information is communicated.. In order to reduce this delay, a new approach has been tested in the past four years. This new system is based on the use of a near patient test (NPT), together with the declaration of medical consultation for ILI.

Several commercial products are available for the detection of influenza virus antigen in less than 30 minutes and do not require any sophisticated materials. The highest sensitivity of these tests was evaluated around 80-85% (8-12) of the cell culture sensitivity and so their use for individual diagnosis should be considered with caution. In contrast, the highest specificity of these tests was quite high (around 95%) (8-12). This specificity, together with the speed with which the result is obtained, make NPT an attractive tool for the surveillance of influenza epidemic in community practice. Such techniques have been used in several countries as in France, Germany, Italy, Poland, the United States (Hawaii) and Switzerland (13-15).
Results of this type of surveillance in Switzerland obtained over four years are reported here. An enhancement of the surveillance was obtained through training which is explained in this article too. Because of this, on behalf of EISS, Switzerland was put in charge of organising an expert task group made up of members from other European countries. The objective of the group was an evaluation of this new surveillance network as an early warning system. A statement on the use of NPTs for influenza surveillance was produced and is communicated in the present article.

Material and methods
Clinical surveillance
On a weekly basis between 1999 and 2003, 150 to 250 physicians notified the Federal Office of Public Health of morbidity due to influenza-like illness (ILI). Based on these reports, the rate of consultations for ILI was calculated.

NPTs surveillance
The influenza A/B Rapid Test NPT (Roche Diagnostics) was used, which is not a commercial test. It enables rapid detection of the influenza virus. The test consists of an immunochromatographic assay that will detect the presence of both Influenza A and influenza B viruses without distinguishing between them. This test has 77.4% sensitivity and 93% specificity of cell culture (16).
Between 150 and 200 sentinel participants per season were provided with an influenza A/B Rapid Test kit. Most participants were general practitioners, and some were paediatricians. Patients with ILI were screened for influenza antigens by obtaining a throat swab. Although selection criteria were the same for both methods, no patient simultaneously underwent NPT and cell culture for the detection of influenza virus. Practitioners communicated numbers and test results daily by fax to our laboratory, the National Influenza Centre. This information was collected and communicated twice a week to sentinel participants via a webpage (http://www.influenza.ch).

Virological surveillance
A restricted number of sentinel participants (range 55-65) took nasopharyngal swabs. Samples were sent in transport medium to our laboratory where influenza viruses were detected by cell culture combined with immunofluorescence. Subtyping was done by a haemagglutination inhibition test using standard antisera.
Results were published on our webpage (http://www.influenza.ch) once a week from the clinical and virological surveillance and twice per week from the NPTs surveillance. Criteria used by the sentinel practitioners to detect patients with influenza-like illness were the same for the three different surveillance systems.

Results
Surveillance with near patient test
Surveillance using the near patient test was conducted during four seasons between 1999 and 2003. Characteristics of the epidemics observed during that period are summarised in Table 1. Seasons between 2001 and 2003 were comparable. The percentage of medical consultations for ILI and the number of influenza viruses detected were found to be related. The nature of viral strains that circulated were the same and were covered by influenza vaccine. However, the 1999-2000 and 2000-2001 seasons were rather different as concerns the intensity and types of strains detected. The heterogeneity of the different epidemics (four seasons) allows an efficient evaluation of the system.

Results of the surveillance between 1999 and 2003 are summarised in Table 2. As shown, the number of participants has decreased over the four years. The number of positive results also varied considerably during the different seasons, as did the percentage of positive results which will be discussed later. Clinical, laboratory, and NPT data are shown in the figure. The two seasons from 1999 to 2001 showed similar patterns for the two virus detection systems and the clinical incidence (14). The significant increase of the virus circulation, the maximal rate of detection, and the decrease in the quantity of viruses detected by the two systems were systematically observed in the same week, or with a difference of one week.

The time when test results were made available was analysed. For cell culture, results were available after 11 days (± 3 days). This includes seven days needed for cell culture and four additional days needed for immunofluorescence and for the weekly communication (14). Results of the ratio of consultations for ILI from the previous week were in general available on Tuesdays (4-5 days delay). Using the NPT, 95% of the results arrived within two days. With two weekly updates of our website, results using the NPT were obtained an average of nine days faster than those obtained by cell culture.

Training
To enhance the quality of results obtained with the NPT by the general practitioners, training was organised in two consecutive years at the beginning of the season. This training consisted of sending three anonymised control samples to the participants: one negative, one weak and one strongly positive. A viral protein was used as the positive control. Results are presented in Table 3. In 2001-2002, only 45% of physicians found the correct combination of results with the three controls. Thirty seven per cent detected only the strong positive control and 18% did not detect any of the positive samples. Every participant who did not find the correct combination received new samples to repeat the test. In 2002-03, the rate of participants who found the correct combination increased to 92%. Six per cent did not detect the low positive and 2% did not detect any positive control. The NPT was therefore used much more correctly the second year. This training is obviously essential and is highly recommended for every new participant in order to assure the quality of the surveillance system.

Statement elaborated and approved by EISS
One major objective of EISS is to increase the sensitivity and efficiency of early warning system for the surveillance of influenza epidemic. To achieve this, the EISS coordination centre selected seven members* based on their virological, clinical, and epidemiological expertise to form a task group under the leadership of Yves Thomas. They were in charge of the evaluation of use of the NPTs in influenza surveillance and the subsequent integration into the EISS network. Additionally, one member of the EISS Steering Committee was represented in the Task Group (John Watson). A statement was produced by the task group after a meeting that occurred 29 November 2002 in Geneva. The final text has been submitted and endorsed by all the EISS members during a plenary lecture in Upsala on 25 April 2003 (Table 4).

Discussion
Sensitivity of the near patient test is lower than that of cell culture and PCR. So, the use of NPT for individual diagnosis is of limited benefit because in case of a negative result, interpretation is difficult. However, the high specificity of some of these tests (8,12,17) make them a valuable tool for the surveillance of influenza. The major point of interest of this surveillance is the speed with which information about circulation of influenza viruses can be obtained, information which is usually confirmed an average of nine days later using the classical surveillance systems (14). This time gain was observed in Switzerland during all four seasons, in France (13) and in Hawaii (15). This rapid detection and reporting of influenza viruses allows the public health services to take the necessary precautions for an epidemic and helps physicians to confirm their diagnoses of influenza (7). As a consequence, this information allows appropriate use of antiviral drugs against influenza only during the epidemic phase. The use of PCR would also reduce time in the surveillance of influenza. Two days for transport and several hours for the PCR analysis are necessary to obtain results about the presence of viral agent in a sample.
On behalf of EISS, a task group of experts was in charge of the evaluation of this new surveillance system. Their conclusions were that surveillance using the NPTs is a valuable tool for an early warning of a change in influenza activity. The statement produced should be helpful to countries that would like to use such a system for surveillance of influenza epidemics. A crucial point is that an existing surveillance system using classical detection methods should not be negatively affected by the surveillance with the NPTs. Clinical and virological surveillance based on cell culture are essential for the reasons already mentioned. Surveillance with the NPTs provides additional information. However, interpretation of the results of such surveillance should be made in combination with the results obtained with the classical systems.

One advantage is that surveillance with NPT requires neither expensive equipment nor the complicated infrastructure of a laboratory. Because these tests are easy to use, they can be performed in any physician´s office. In addition, participants in this new surveillance network reported that they were very pleased to have the NPTs available in their offices, offering them the possibility of a rapid confirmation of the presence of influenza virus in their community practice. The use of the NPT could also enhance the accuracy of their diagnosis. Such a surveillance scheme is easy to organise. However, organising a surveillance scheme with the NPTs on parallel with existing surveillance system requires additional expense of money and effort. This aspect should not be ignored and cost-benefit must be evaluated. In any case, the introduction of this new surveillance system should remain optional for the surveillance of influenza epidemic. Training the sentinel practitioners to use the NPTs improved their use of these tests. This training is highly recommended for the installation of such surveillance.

Another potential added value of this system is that it increases virological information. Indeed, a double swab could be planned for patients with a positive result detected with NPT. The second sample could then be sent to the laboratory for cell culture. Further studies on the virus would give information on the evolution of the strains circulating. The periods before and after the season could give predictive indications about the strain that might circulate during the following season.
During the different seasons, the number of positive samples detected with cell culture and with NPTs per week could be compared. However, the percentage of positive samples detected with the NPTs are considerably lower than the one observed with cell culture. This indicates that some positive samples will be missed because of the general lower sensitivity of such NPTs. The heterogeneity of the network and of the participation of physicians could contribute to such differences. These results show that additional studies are needed to complete the evaluation of the usefulness of surveillance with the NPTs.
 
This paper was written on behalf of the EISS Task Group* on ´The use of NPTs for influenza surveillance´.

* Members of the Task group: Aad Bartelds, Isabel Burckhardt 1, Anne Mosnier, Yves Thomas, John Watson, Sylvie van der Werf, Werner Wunderli (National Centre of Influenza, Switzerland).
1 Participated in the Task Group meeting but not in the writing of the EISS Statement

Acknowledgements
This work was a collaboration between the Swiss Sentinel Surveillance Network, the National Centre of Influenza and Roche-Pharma AG. We would like to thank all the sentinel practitioners.