Introduction
A European surveillance scheme for travel associated legionnaires' disease (now called EWGLINET) was established by the European Working Group on Legionella Infections (EWGLI) in 1987. Its objectives are to identify cases of legionella infection in returning travellers, to detect outbreaks and clusters of legionnaires' disease and to collaborate in the control and prevention of further cases. Its history and current activities are described in detail on its website (1). This paper provides results and commentary on reports of cases that occurred in 2000 and 2001. The impact of a major change in the cluster definition in 2001 and an increase in reports from specific countries will be discussed.

Methods
A case of travel associated legionnaires' disease is defined as a person with:
- Clinical or radiological evidence of pneumonia and a laboratory diagnosis demonstrating evidence of legionella infection (2);
- A history of travel in the 10 days before the onset of illness. Travel is defined as staying away from home for one or more nights in accommodation used by travellers, such as hotels, holiday apartments etc.
Details on cases ascertained by national or regional surveillance systems of participant countries are entered into the scheme's web database. When a new case is added, the database is searched by the coordinating centre at CDSC for any previous cases reported to have stayed at the same accommodation site.

From 1 January 2001, a cluster is defined by EWGLINET as 2 or more cases associated with an accommodation site, which occur within two years of each other. All other cases are treated as single cases. Under the old definition, cases occurring at sites with previous cases more than six months earlier were categorised as "linked".

Single cases are immediately notified to national collaborators and/or the Ministry of Health. Clusters are reported to all collaborators and the World Health Organization (WHO). In some countries public health authorities choose to report clusters to their national organisations of tour operators. Different levels of intervention are expected: issuing a checklist for minimising risk of legionella infection to sites associated with single cases, and conducting risk assessments, sampling for legionella and implementing control measures at sites associated with clusters (3). Closure of accommodation sites is at the discretion of public health authorities, but tour operators sometimes withdraw from sites associated with large or extended outbreaks of legionnaires' disease.

Results
In 2000, 31 countries and 38 collaborating centres took part, increasing to 33 countries and 43 centres in 2001 (figure 1). The number of cases by year of onset has increased from 3 cases in 1987 to 360 in 2000 and 481 in 2001, the highest number reported to the scheme in any one year (table 1).

Characteristics of cases
In 2000 age and sex profile of cases were similar to previous years, with more than twice as many men as women reported. The mean age of cases was 57 years in 2000 (median age 57 years, range 24-88 years). In 2001 the age distribution showed similar characteristics with a mean and median age of 59 years (range 17-96 years). The proportion of men to women increased to 3:1 with most of the increase occurring in men in the 50-59 and 60-70 year age groups. In women, the main decrease occurred in the 40-59 year age groups. Most countries reported cases in both years in proportion to the overall male/female ratio except for France where the male to female ratio of cases was 7.3/1 in 2000 and 5.3/1 in 2001.

The number of cases diagnosed by culture has remained small, with 36 in 2000 and 34 in 2001. The urinary antigen detection method was used to confirm 262 (73%) cases in 2000 and 378 (79%) cases in 2001 (figure 2), resulting in both an absolute and a proportionate decline in the numbers of cases diagnosed by all other methods. Diagnosis by a four-fold rise in antibody levels declined to 26 (5%) cases in 2001 from 36 (10%) in 2000 compared with 61 (25%) in 1997. Cases diagnosed by single high titre serology have remained at similarly low numbers in recent years and accounted for 6-7% of all reports in 2000 and 2001. Use of other techniques such as PCR accounted for 0-2% of cases in the same period.

In 2000, 117 cases (28%) were linked to L. pneumophila sgp1 infection and 182 (50.5%) to L. pneumophila other or unknown serogroup. In 2001 these figures were 186 (21%) for L. pneumophila sgp1 and 225 (47%) for L. pneumophila other or unknown serogroup. The remaining cases in both years were reported as legionella with no further details. Only 2 cases had other species detected and reported, both in 2001.
An outcome of illness was reported for 74% of cases in 2000 and 73% in 2001. Deaths were reported for 23 cases in 2000, and 41 cases in 2001, case fatality rates of 6.4% and 8.5% respectively. Cases reported as still ill at the time of report or known to have recovered were similar in both years (32% and 29.5% still ill, and 35.5% and 35% recovery in 2000 and 2001 respectively). Cases were further classified as having a known outcome (recovery or death) or unknown outcome (still ill or unknown). In 2000, 42% had a known outcome, and 44% in 2001. This compares with an average of 65% with a known outcome between 1992 and 1999 (figure 3).

Reporting
Infections are usually diagnosed after cases return to their country of residence and 18 countries reported cases in either 2000 or 2001. The Netherlands reported the greatest number of cases: 103 (29%) in 2000 and 118 (24.5%) in 2001 (figure 4) followed by England and Wales, 91 (25%) cases in 2000 and 89 (18.5%) cases in 2001. France almost doubled and Italy more than doubled the number of cases reported in 2001 compared with the previous year.

Median interval between onset of illness and report of the case to EWGLINET were 33 days in 2000. This was reduced to 23 days in 2001 (figure 3). A median of 38 days was calculated between 1992 and 1999. The Pearson's correlation coefficient between the proportion of known outcomes and the median delay to report was -1.2 (p=0.010; 95% CI [-2.077; 0.337]), indicating a decrease of 1.2% in known outcome for every day's reduction in the time from onset to report of a case.

It was notable that 4 countries in 2001 were responsible for reporting 71% of the cases: the Netherlands, England and Wales, France and Italy. The travel patterns for cases from these countries were distinct, with 87% of Italy's cases travelling in Italy and 66% of France's in France. In contrast, only 4% of Dutch cases and 8% of English cases travelled in their own country. The Dutch traveled to a far wider range of countries, 28% of them to Turkey and 11% to more than one European country, while 29% of the English traveled to Spain, 13% to Turkey and 16% to countries outside of Europe.

Countries visited
Cases with onset in 2000 stayed at an average of 1.7 accommodation sites in 56 countries. For the first time, France was the most visited country with 55 cases (15%) traveling only in France. Fifty-four (15%) cases visited only Spain, 40 (11%) only Turkey and 36 (10%) cases only Italy. Forty two (12%) cases travelled in more than one country in Europe, mostly through France, Italy and Spain and a further 89 cases to other European countries. Nine cases were associated with Mexico, the remaining 35 (10%) cases with countries outside the EWGLINET scheme.

Cases with onset in 2001 visited an average of 1.5 sites in 46 countries. Italy was the most visited country for the first time with 91 (19%) cases travelling only in Italy. France was associated with 86 (18%) cases travelling only in France, Turkey with 78 (16%) cases and Spain with 74 (15%) cases only in Spain. Twenty-five cases travelled to more than one country in Europe, mostly through France, Italy, Germany and Spain; the remaining 42 cases with other European countries. Eleven cases were associated with travel to Mexico, and a further 74 (15%) cases with countries outside the EWGLINET scheme.

Rates of infection in travellers from the United Kingdom were calculated using data from the Office for National Statistics Travel and Tourism Survey (4) (table 2). The highest rates of travel associated legionnaires' disease were in travellers visiting Mexico and Turkey in both years, and the lowest rates were in France and the United States.

Travel itineraries
In 2000, 254 (71%) cases stayed at only one accommodation site in the two to ten days before onset. The remaining 106 (29%) cases visited between two and eight sites and between one and five countries per incubation period. In 2001, 373 cases (77.5%) cases had stayed at only one site, the remainder again staying at between two and eight sites in one to five countries.

Cases used a variety of types of accommodation. Hotels and hotel apartments were still the most popular with 77% and 78% of cases staying in at least one such site during their incubation period in 2000 and 2001 respectively. As in 1999 (3), many itineraries included stays at smaller sites, such as bed and breakfast, farmhouse and camping accommodation. Travellers visiting these smaller sites were often touring and moving on every few days. Business travel was also reported. These cases typically spent short visits in large city hotels or at motels while on the road. Eleven cases in 2000 and four in 2001 were associated with stays on ferries or ships.

Clusters
In 2000, 28 clusters were detected and associated with 70 cases, 19% of the total reports (table 1). Of these clusters, 10 (36%) would not have been detected without EWGLINET since each included only one national from different countries. The clusters detected were small, ranging in size from 2 to 5 cases. As in previous years, most of the clusters occurred in the most visited countries. For example, Spain had the most clusters (6), 2 of which were large with 5 cases each. Turkey had 5 clusters, France 4, and Italy 3. Countries outside of the EWGLINET scheme were involved in 4 clusters, 2 in Mexico and 1 each in Morocco and Moldova. Of the 70 cases associated with clusters, 10 deaths (14%) were reported.

Under the new cluster definition, 72 clusters were detected in 2001 and associated with 149 cases, 31% of the total reports (table 1). Without EWGLINET, 23 (32%) clusters would not have been detected. The largest cluster identified, in Spain, involved 8 cases and 3 further clusters, in Spain and Turkey, involved 5 or 6 cases. Fifty-two of the clusters had 2 cases associated with them. Turkey had the most clusters identified (20). Half of all cases reported as travelling in Turkey were associated with a cluster.

In 2001, 3 clusters were identified outside of the EWGLINET scheme, 2 in Mexico and 1 in the US. Amongst the clusters in the collaborating countries, 4 involved multiple sites, 3 of them involving nationals from the same country who travelled on the same itinerary: Turkey in 2 multiple clusters, Germany and Italy in the third. In a fourth cluster, two nationals from different countries followed a similar travel pattern at different times in Italy. Eleven (7%) of the 149 cases associated with clusters were reported to have died.
In 2001, 43 of the 72 clusters identified (60%) would have met the old definition. This means that EWGLINET 'gained' 29 extra clusters. France was the country most affected by the new definition with eight 'additionnal' clusters identified. Spain had 6, Italy 4 and Greece 3.

Environmental investigations
Greater use of the electronic database was made in 2001 for reporting the results of environmental investigations at accommodation sites associated with cases. Over 200 such reports were received in 2001 compared with less than 40 in 2000. However, many of these reports were for investigations carried out in previous years. Hence the following results only include environmental investigations known to have been carried out recently: 55 investigations were carried out in 2000 and 140 in 2001 (table 3). Of these 20 were for the 28 clusters detected in 2000 and 31 for the 72 clusters detected in 2001. Legionellae were detected in 9 of the 20 sites associated with clusters in 2000 and 15 of the 31 sites in 2001.

In 3 of the sites associated with single cases in 2000, a clinical isolate from a case was indistinguishable by subgrouping and molecular subtyping analysis from the environmental isolate. This occurred five times in 2001. For cluster sites in 2000, no such comparison could be made. However, in 2001 clinical and environmental isolates were indistinguishable from each other in 2 cluster sites, 1 in Norway and 1 in Italy. Control measures were reported in just under half of all the investigations carried out in both years, although information on the closure and re-opening of sites was not very evident or reliable.

Discussion
The EWGLINET surveillance scheme continued to expand in 2000 and 2001 with substantial increases in the numbers of cases reported together with more information on environmental investigations. The rise in cases is thought to be due to the improved detection and reporting from national surveillance schemes, typified by the activities of France, Italy and the Netherlands, rather than any new increase in overall incidence. The Netherlands has now replaced England and Wales as the most frequent reporter of cases. The comparison of the travel habits of cases from France and Italy compared with those from the Netherlands, England and Wales is interesting. Cases from England, Wales and the Netherlands were almost all acquired abroad, whereas cases from France and Italy mainly acquired their infection within their home country. Thus, improved control measures in these latter two countries will substantially benefit their own residents as well as nationals from other countries who visit France and Italy in significant numbers. The different seasonal distribution seen in 2001 may be the result of the high number of cases reported by France and Italy, perhaps related to their different national habits of holiday-making.

As in previous years (3) the marked increase in the use of the urinary antigen detection test may also be contributing to the rise in the number of cases reported. The test offers a rapid diagnostic facility compared with other methods, allowing detection and reporting at an earlier stage of illness. The correlation between the rise in the use of urinary antigen detection test and the rise in the number of cases with an unknown or 'still ill' outcome at the time of report is noteworthy. The rising proportion of cases with these outcomes means that a number of deaths may not be reported and the case fatality ratio seen in EWGLINET may consequently be a large underestimation. The level of data completion in this area would be improved with follow up surveys of the collaborators and/or the reporting local health authorities.

A further effect in the rise of the urinary antigen detection method is that mainly only L. pneumophila infections are identified. It is useful for clinicians and public health authorities to bear in mind that a clinical specimen for culture may provide a positive diagnosis in a suspected case where the result of the urinary antigen method is negative. A higher number of cases diagnosed by culture would not only provide better opportunities to match clinical and environmental isolates for identifying sources of infection but would also detect other species or serogroups of legionella infections.

The rates of infection calculated for visitors from the United Kingdom (the only group for whom comprehensive information on the number of travellers making trips abroad is available) indicate that countries such as Mexico and Turkey are associated with a disproportionate number of cases compared with the number of people from the UK visiting them. Mexico is growing as a tourist destination whereas Turkey has experienced an increase in popularity over the last ten years or so. Their high rates of infection per million tourists should be addressed through a greater emphasis on hotel legionella prevention programmes, in order to minimise risk to people taking holidays there.

The response to clusters in the country of infection has been enhanced in the last two years by the change in cluster definition at the start of 2001. The large increase in clusters identified, 29 of which would not have been identified as such under the old definition, has ensured that in more instances an accommodation site will have a risk assessment and environ- mental sampling conducted. Many of the linked cases that the old definition would have identified have been absorbed into these 'new' clusters. In the notification to the country of infection those cases identified as single rather than linked under the new definition were flagged as having stayed at a site associated with a previous case more than two years earlier, allowing that country to respond according to its own public health policy. The clusters identified in 2000 and 2001 also demonstrate clearly the value of the international collaboration in detecting otherwise unnoticed public health risks. One third of all clusters would not have been identified without EWGLINET, as the cluster cases each came from a different country.

The large proportion of cases whose itineraries included visits to more than one site in a country or more continue to add to the logistic and resource demands placed on local public health and environmental authorities within the collaborating scheme. Response to single case reports varies across the collaborating scheme. Authorities in Italy and Spain for example investigate all sites associated with single cases (risk assessment and/or environmental sampling) while other countries typically issue the standard checklist of good practice in legionella control (5) the minimum action required by EWGLINET.

A major undertaking throughout 2000 and 2001 was the development of the European Guidelines for Control and Prevention of Travel Associated Legionnaires' Disease that became operational in July 2002 (6). During the creation of these guidelines a consensus arose within the collaborating countries on the management of clusters from 2002, and the importance of reporting environmental investigations from the country of infection became more evident. This is reflected in the increased activity both at the co-ordinating centre in London in entering pre-2000 and 2001 reports into the electronic database and in the rise in reporting of investigations in 2001 by the collaborating countries. A full analysis of these reports will be published separately.

A significant proportion of cluster sites, and also of single sites had legionellae isolated from samples. This information is important for assessing the impact of control measures at a site, as well as for providing evidence for any legal action arising from an infection. However, interpretation of the significance of environmental data results is limited when there are no matching clinical isolates from associated cases. The environmental investigations reported show that investigation and reporting processes take varying amounts of time, depending on the structure of public health services in each country. The European guidelines are expected to improve the investigation reports, especially for cluster sites, and should facilitate useful and interesting data in future years.

Acknowledgements

This work is funded by the European Commission DG SANCO General Health and Consumer Protection.

We would like to thank all the collaborators (7) for reporting their cases and all the people involved in public health control and prevention programmes for travel associated legionnaires' disease.