The history and current activities of EWGLI are described further on its website (www.ewgli.org).

This paper provides results and commentary on cases of travel-associated legionnaires' disease reported to EWGLINET with onset in 2003.

Methods
All countries that participate in EWGLINET use standard case definitions. A single case is defined as a person who stayed, in the two to ten days prior to onset of illness, at a public accommodation site that has not been associated with any other previous cases of legionnaires' disease, or a person who stayed at an accommodation site linked to other cases of legionnaires' disease but after an interval of at least two years [1].

A cluster of travel-associated legionnaires' disease is defined as two or more cases who stayed at or visited the same accommodation site in the two to ten days before onset of illness and whose onset is within the same two year period [1].

National surveillance schemes detect and follow up each case within the country of residence and then report the case, travel and microbiology details to the EWGLINET coordinating centre at the Health Protection Agency's Communicable Disease Surveillance Centre (CDSC) in London. The details are entered onto a database, and the database is searched to check whether that case should form or become part of a cluster, or whether it is a single case.

In July 2002, European guidelines were introduced to standardise national responses to cluster alerts by EWGLINET [1-5]. The response to single cases is via the collaborator in the country of infection, who issues a checklist for minimising risk of legionella infection to any accommodation sites involved. Cluster sites require that more detailed investigations be carried out, including risk assessments, sampling and control measures. Countries report the progress of such investigations to the coordinating centre in London using a Form A (two-week investigation report) and Form B (six-week investigation report) for each cluster. If these forms are not received within the relevant time period, EWGLINET publishes details of the cluster on its public website (www.ewgli.org) to state that the coordinating centre cannot be confident that the accommodation site has adequate control measures in place. This notice is removed once the relevant form(s) have been received, to confirm that measures to minimise the risk of legionella infection at the site have been taken.

Results
Cases and outcomes
Thirty six countries participated in EWGLINET in 2003 [FIGURE 1] and reported a total of 632 cases of travel-associated legionnaires' disease to the coordinating centre with onset in 2003 (including one case reported by the United States, which is outside of EWGLINET). This compares with 676 cases reported in 2002.

Cases reported to EWGLINET follow a distinctive age and sex pattern. Each year, approximately three times as many male cases are reported as female cases, and most cases are aged 50 years or over. In 2003, male cases outnumbered female cases by 2.6 to 1, and the peak age group reported was 50-59 years for both sexes. The age range for males was 15 to 91 years, and for females, 15-89 years (with one case of unknown age).

EWGLINET sees a very seasonal pattern of reporting. There is often a peak in the number of cases with onset over the summer months, and a drop-off in cases over winter. This is, for the most part, because the scheme records only travel-associated cases, and the majority of people choose to take their holidays during the summer. In 2003, cases peaked in July, with a second, smaller peak, in September.

The case fatality rate in 2003 was 6% (38 deaths reported), a very slight decrease from previous years. The number of patient recoveries reported increased from 30% in 2002 (203 cases) to 38% in 2003 (238 recoveries). 192 cases were reported as 'still ill' (a similar number to previous years), whilst the number of cases with unknown outcomes decreased from 34% in 2002 to 26% (164 cases) in 2003. These are the case outcomes at time of report to EWGLINET, or final outcomes if follow-up information is forwarded to EWGLINET at a later date.

Microbiology
The proportion of cases diagnosed by urinary antigen detection as the main diagnostic method continued to increase (80.5% in 2002 [2], 81.2% of cases in 2003). The number of culture proven cases remained relatively constant (48), while the number of cases diagnosed by serology declined slightly from 2002 (in 2003, 23 cases were diagnosed by four-fold rise, and 43 cases by single-high titre, compared with 2002's 49 diagnoses by four-fold rise and 31 by single-high titre).

The main category of detected organism reported to the coordinating centre was Legionella pneumophila serogroup 1 (485 cases, 76.7%). The remaining cases were reported as 'L. pneumophila serogroup unknown' (90 cases, 14.2%), 'L. pneumophila other serogroup' (4 cases, 0.6%), 'Legionella species unknown' (50 cases, 7.9%) and 'Legionella other species' (3 cases, 0.5%. Two of these cases were L. bozemanii, the species of the third was not reported).

Travel
The main countries reporting cases of travel-associated legionnaires' disease in their citizens were England and Wales (159 cases), France (120) and the Netherlands (104) [FIGURE 2].

The main countries of infection were Italy (122 cases) and France (118), largely because both of these countries report many cases of the disease in their citizens who have been travelling within their own countries (Italy had 64 internal cases, France had 89). The countries visited by the third and fourth highest numbers of cases were Spain (91 cases) and Turkey (64), neither of which reports much internal travel (Spain reported 9 cases in 2003, Turkey did not report any). If external travel (i.e. foreign travel) only is considered, Spain (82 cases in 2003) and Turkey (64) become the countries with the highest number of infections, followed by Italy (58) and France (29). All other countries of infection had fewer external travel cases.

The proportion of cases linked to clusters for the main four countries of infection ranged from 26% (France) to 41% (Turkey). For Turkey, this is a big improvement on the 71% of cases linked to clusters in 2002. Italy had 29% linked to clusters in 2003, while Spain had 31% [FIGURE 3].

Twenty seven cases visited more than one European country, and eight visited more than one country including one or more non-European countries. An additional 68 cases (10.8%) visited countries outside the EWGLINET scheme.

Whilst 494 cases stayed in only one accommodation site during their 2-10 day incubation period, the remaining 138 stayed in more than one, with one Danish case staying in eight. The average number of sites per case was 1.42.

Clusters
Eighty nine new clusters were detected in 2003. Clusters were defined as two or more cases associated with the same accommodation site, where the second and subsequent cases had onset in 2003, and the first case had onset up to two years previously. These clusters varied in size, and although the majority consisted of only two cases (66 clusters), one cluster involved 17 cases. This cluster was located in England, and centred on a hotel and leisure centre. In addition to the 17 English travel-associated cases, of whom two died, there were three further cases of legionnaires' disease and two cases of Pontiac fever identified in the community (none of whom died), giving a total of 22 cases of disease associated with this outbreak. The source was traced to a spa pool located in the complex.

The second largest outbreak detected by EWGLINET in 2003 was located at a hotel in Spain, and was associated with eight Swedish cases of travel-associated legionnaires' disease. No deaths were reported to EWGLINET. The first six cases formed a new cluster in the early half of 2003, sampling for legionella at the hotel was positive, control measures were taken, and a Form B report was submitted. However, subsequent to this, two further cases stayed at the accommodation site and became ill, leading to a request from EWGLINET for new investigations and a new Form A and B to be submitted. The reports showed that samples were again found to be positive, and that further control measures had been carried out. The Spanish authorities reacted promptly to the EWGLINET alerts, and gave detailed updates on the situation at the hotel to the coordinating centre throughout the investigations. At the time of writing, no further cases have been reported with association to this hotel.

In contrast to the two clusters detailed above, 35 of the clusters detected in 2003 involved a single case from two or more countries, and so would not ordinarily have been detected by any individual country. Thus 39% of the clusters with onset in 2003 would not have been identified without the EWGLINET surveillance scheme.

The 2003 clusters occurred in a total of seventeen countries. France had the most (18 clusters), followed by Italy (14), Turkey (12) and Spain (11). Twelve clusters fell in countries outside the EWGLINET scheme (Bahamas, Cyprus, Dominican Republic, Egypt, Mexico, Sri Lanka and Thailand), and three were situated on cruise ships. Five clusters involved two or more accommodation sites, of which one spanned two countries. Most of the clusters had onset in summer, with peaks in July and October, but at least two clusters occurred in every month in 2003.

Investigations
The eighty nine new clusters in 2003 involved a total of 98 sites, one of which was already under investigation, and 12 of which were situated in non-EWGLI countries, leaving 85 that required EWGLINET investigations. In addition, 21 sites that had been associated with clusters in previous years were associated with additional cases ('cluster updates'), and so required re-investigation. In total, 106 investigations were required by EWGLINET for 2003 clusters and cluster updates. EWGLINET also requested the investigation of a cluster site in northern Cyprus (a non-EWGLI country). Turkey arranged this and returned a Form B, giving a total of 107 Form B reports received for the 2003 clusters and cluster updates.

Fifty nine 'Form B' reports (55%) stated that samples from the accommodation site had tested positive for L. pneumophila (at concentrations equal to or greater than 1000 cfu/litre), 46 (43%) reported that samples had not detected any L. pneumophila, and two Form B reports (2%) were unable to give sampling results for reasons accepted by the coordinating centre. The names of fifteen of the Turkish sites and one French site from new clusters or cluster updates in 2003 were published on the EWGLINET website for failure to return reports on time, or for failure to implement appropriate control measures on time. Four Turkish cluster sites and one French cluster identified in 2002 but where investigations were due in 2003 were also published, giving a total of 20 Turkish sites (some published more than once, giving 25 postings), and two French sites published on the EWGLI website in 2003.

In 2003, investigation reports were returned for 151 single sites, even though the EWGLI guidelines do not require such investigations to be carried out. Of these, 132 sites were sampled, and 72 (54.5%) were positive for L. pneumophila.

Discussion
The number of cases reported to the EWGLINET surveillance scheme in 2003 was not as high as in 2002, but still represented a significant burden of disease in European travellers. France and Italy were the main countries of infection for 2003, due in no small part to the large number of internal cases reported by these countries each year. If the internal travel were to be removed, Spain and then Turkey would have been the main countries of infection. The fact that countries such as Italy and France do report their internal travel cases allows an international surveillance scheme like EWGLINET to detect more clusters within those countries. In all, 12 countries reported cases of internal travel to EWGLINET in 2003, one more than in 2002.

In 2003, 17 out of the 18 clusters located in France would not have been detected without internal reporting (i.e. no more than one case in the cluster involved foreign travel), and six out of 14 clusters in Italy would not have been detected. The number of clusters detected because of internal reporting for Turkey (none out of 12 clusters) and Spain (two out of nine) are much smaller because of the low number of internally reported cases from those countries. If all countries began to report their internally acquired cases of travel-associated legionnaires' disease to EWGLINET, we would expect to see a large increase in the number of clusters detected by the scheme.

Not all cases of travel-associated legionnaires' disease are reported to EWGLINET each year. The coordinating centre collects an annual dataset [6] from each country detailing every case of legionnaires' disease detected by that country, including the number of travel-associated cases acquired abroad and internally. In 2003, the difference between the annual dataset and the EWGLINET dataset suggested that 290 travel-associated cases had not been reported to EWGLINET. This is in part due to legal restrictions on reporting in some countries. However, whilst 76% of cases acquired abroad in 2003 were reported, only 57% of internally acquired cases were. Countries may believe that EWGLINET is less interested in such cases, but they are a very valuable addition to the EWGLINET dataset, as discussed above.

The EWGLI guidelines for investigation of clusters were put in place in July 2002, so 2003 was the first full year of their use. Some countries have experienced difficulties implementing them efficiently, and EWGLINET is attempting to help these countries adapt to the new procedures. Turkey in particular encountered difficulties managing its investigations, with 20 of its sites being published on the website in 2003. Improvements in this country have now occurred as a result of their strengthened links between the collaborators and local public health officials.

EWGLINET collaborators and local health authorities in many countries have put a great deal of effort into thoroughly investigating the 107 sites that returned a Form A and B in 2003, and it is very encouraging that the vast majority of investigations are being carried out satisfactorily and on time. In addition, in one investigation, legionella isolates were obtained by England from a cruise ship associated with a cluster of legionnaires' disease, and were typed and matched with a clinical isolate from a German patient using sequence-based methods, confirming the site as the source of the outbreak [7]. The small number of clinical isolates obtained from patients limits the use of this technique, and countries should encourage samples for culture to be taken from patients with legionnaires' disease where at all possible.

Over the last few years, participants in the EWGLINET scheme have detected an increasing number of cases, and the crude fatality rate has decreased accordingly, as less serious cases are diagnosed and reported. Additionally, the urinary antigen test has made the process of diagnosis much faster, leading to earlier treatment of individual cases, earlier detection of clusters, and therefore earlier implementation of control measures. Despite the decreasing percentage of fatalities attributed to legionnaires' disease, the EWGLINET scheme continues to fulfil a very important role, emphasised by the 39% of clusters that would not have been detected in 2003 without its international reach.

Acknowledgements
This work is co-funded by the European Commission Health and Consumer Protection Directorate-General and the member states of EWGLI.
We would like to thank all the collaborators* for reporting their cases and all the people involved in public health control and prevention programs for travel-associated legionnaires' disease.
* The list of collaborators is available on the EWGLI website at www.ewgli.org