On average, 1675 cases of legionellosis are reported to the Centers for Disease Control and Prevention (CDC) annually by state health departments, the majority of which are Legionnaires' disease (LD) (CDC, unpublished data). It is estimated that between 8,000 and 18,000 persons are hospitalized with LD in the United States each year, suggesting that under-diagnosis or under-reporting are common [1]. One species of Legionella, Legionella pneumophila, is responsible for 90% of reported cases of LD in the United States and the majority of these are caused by Legionella pneumophila serogroup 1 (Lp1) [2, 3].

Eighty percent of cases of LD are sporadic [4] and the source of Legionella for these cases is rarely known. In contrast, outbreaks, while rare, provide an opportunity to identify common sources of Legionella. Community LD outbreaks have frequently been associated with cooling towers [5–13], and whirlpool spas [14–17], and less frequently with evaporative condensers [18], air-scrubbers [19], supermarket misters [20], potable water [21, 22], and decorative fountains [23].

In a three week period in June and July 2005, seven cases of LD among residents of Rapid City were reported to the South Dakota Department of Health (SDDH). In the previous 10 years, only two confirmed cases of LD had been reported to the SDDH from Rapid City. (Personal communication, Dr. Kightlinger) No common exposures among case-patients were initially apparent. In July 2005, SDDH and CDC began an epidemiological and environmental investigation to identify the source of the outbreak and to prevent additional cases.

We believe that the source of this outbreak was the decorative fountain in Restaurant A for a number of reasons. The epidemiological evidence showed that even with poor case-patient recall the odds ratio from the case-control study showed a strong association between eating at restaurant A and having LD. The environmental investigation found Benidorm strain in the fountain in Restaurant A but not in any other potential environmental source. In addition, the strain isolated from the fountain and the three patient isolates were identical by SBT analysis. The colony count observed in the fountain, while very high, should be viewed as supportive but not decisive evidence of the fountain being the source.

This outbreak was unusual in that it lasted several months, a source that on observation appeared to generate only a small aerosol was implicated, and the case-patients had very short exposure periods. Previously a decorative fountain located in a hotel lobby was associated with an outbreak of LD [23] and a decorative fountain in a restaurant was linked with an outbreak of Pontiac Fever [27]. In addition, a decorative fountain located outdoors was the suspected source of an outbreak of LD in Portugal [28]. These fountains were significantly larger than the fountain implicated in this outbreak and were likely to have generated a larger aerosol. The fountain in Restaurant A was one of the many potential sources investigated early in the outbreak. Only sources that had a reasonable epidemiological link and that were considered to be a plausible source were sampled. Our initial assessment of the fountain in Restaurant A in July led us to conclude that it was an unlikely source because it appeared to generate very little aerosol, and because at that point only two case-patients reported having eaten at Restaurant A. Therefore, it was not sampled at that time but in October when we found a stronger epidemiological link. The fountain was reported to be out of operation for approximately 30 consecutive days during the time of this outbreak. Although the exact dates were unavailable to us, they appeared to correspond with the period on the epidemic curve when there were no cases. Rather than remediate the fountain in Restaurant A the fountain was permanently removed.

Our initial hypothesis was that a cooling tower was the source of this outbreak for three reasons. Firstly, we could find no common exposures and could not initially place more than two case-patients in any one building, leading us to believe that the source was outdoors. Previous studies have shown infections occurring as far as 6 km from an external source indicating that the wind is capable of transporting Legionella plumes [5]. However, exposure to contaminated cooling tower drift can occur indoors but in this situation there was no geographical clustering of patient homes which might have been expected if infection was as a result of a drift. Secondly, we found several cooling towers with Legionella in the city close to areas where people shopped and worked, and to map grids through which case-patients were significantly more likely than controls to have traveled. Apart from exposure to cooling towers (using map grids as a proxy), analysis of potential exposures where more than half the case-patients had visited did not initially show any significant differences between case-patients and controls. This hypothesis was strengthened by the timing of the outbreak at the start of summer with many towers bearing heavier cooling loads and running in excess of 100% capacity due to a hotter than usual summer in South Dakota. Thirdly, as outlined in the background, cooling towers are common sources of community outbreaks of LD.

Clinicians at the local hospital in Rapid City had adopted a policy in November 2004 of more intensive diagnostic testing for patients with CAP so that antibiotic therapy could be targeted. Legionella urine antigen testing therefore increased prior to detection of this outbreak. This highlights the importance of appropriate diagnostic testing in patients with CAP. While many patients with LD can be treated successfully and empirically with antibiotics currently recommended for CAP, this outbreak shows that the use of etiology-specific diagnostic tests can lead to a public health intervention that prevents future cases of LD. Furthermore, increasing physician awareness and use of the urinary antigen testing, through our recommendation during the outbreak to test all new cases of CAP for LD, may have improved patient management and contributed to the relatively low case-fatality rate. The availability of clinical isolates of Legionella from 4 case-patients was critical in allowing us to identify the source of the outbreak. In the United States there has been a steady decline in the proportion of LD cases diagnosed by culture since the introduction of urine antigen testing [2]. However, as this outbreak demonstrates, diagnostic testing of persons with CAP should include collection of urine for antigen testing and respiratory specimens for culture of Legionella whenever possible. This recommendation is included in the recently updated Infectious Disease Society of America/American Thoracic Society (IDSA/ATS) guidelines on the management of CAP in adults [29].

Legionella was cultured from over half (55%) of the cooling towers tested, several of which were positive for Lp1. Although the outbreak was assumed to be caused by Benidorm strain, which was not found in any of the cooling towers, the presence of Legionella in so many sources, some with high colony counts, and with the potential to aerosolize is worrying. Other studies have shown that detectable levels of Legionella are present in many cooling towers and other building water systems [30, 31]. Some researchers advocate measurement of colony counts as a predictor of disease risk [30]. However, because measurement of colony counts is not standardized, no other source was initially evident, and because there is no known safe level of Legionella, we applied the precautionary principle and recommended that all positive towers be remediated regardless of their colony count. SDDH employed a Legionella consultant (TK) to manage remediation of the cooling towers in accordance with published guidelines [32], and to hold cooling tower maintenance workshops for industry and businesses in Rapid City and two other South Dakota cities.

Our investigation had certain limitations. While it is understandable that individuals recovering from serious illnesses would have some difficulties with recall of activities 1 to 2 weeks earlier, poor recall delayed our ability to identify Restaurant A as a potential source of the outbreak. We could have asked our case-patients and controls whether they visited each individual shop, restaurant, etc. in Rapid City. However, this would have made our already lengthy questionnaire unwieldy. Case-patients were re-interviewed about their exposure to Restaurant A but controls were not re-interviewed. Therefore, we have not presented new matched odds ratios as they may be biased. Given the magnitude of the odds ratio (mOR 32.7) based on the initial case-patients' reported attendance at Restaurant A and given that the Benidorm strain was found only in the fountain in Restaurant A, we believe that not re-interviewing controls has not weakened our conclusions that the fountain in Restaurant A was the source of this outbreak. We excluded the last five cases from the case-control study because they were reported after the source of the outbreak was published and we were concerned that their responses may have been biased. We believe that these cases appeared rather suddenly because media attention may have led to increased medical care-seeking, increased diagnostic testing, increased reporting, or all three.

We were unable to explain the presence of Denver strain in the clinical isolate of one case-patient. Although this patient ate at Restaurant A, it is possible that this case-patient was infected by a different source and was a sporadic case unrelated to the outbreak. This assumes that the case-patients that were not culture confirmed were infected with Benidorm strain. It is also possible that Denver strain was present in the fountain but was masked by the predominance of Benidorm strain, or that Denver was present in the fountain in early August when this case-patient became ill, but was no longer present in late October when the fountain was sampled. These hypotheses also make it possible that Benidorm strain may have been present in cooling towers but was not present at the time of sampling. If that were the case however, one might have expected more people who had not eaten at Restaurant A to be case-patients. We interviewed approximately half of the controls by phone whereas all case-patients were interviewed in person. This may have led to better recall among cases. Interviews with controls by phone facilitated more rapid recruitment of controls and thus a more timely analysis.

Most of the fountains we sampled had little or no routine maintenance although this is recommended by fountain manufacturers and by the American Society of Heating, Refrigeration, and Air-conditioning Engineers (ASHRAE). ASHRAE guidelines are non-specific for fountain maintenance and are not disseminated widely in the restaurant industry [32]. Proper care and maintenance of ornamental water fixtures, such as decorative fountains, is essential to prevent outbreaks of LD and can be achieved by increasing awareness among fountain operators of the importance of adequate maintenance.

Small decorative fountains pose a previously unrecognized source of Legionnaires' disease outbreaks as we believe this is the first time that such a small fountain with apparently limited aerosol-generating capability has been implicated as the source of an LD outbreak. Investigations of future community outbreaks of LD should consider exposures to indoor decorative fountains, including small ones such as those that might be present in restaurants, hotels, or other businesses, as potential sources of Legionella.

The findings and conclusions in this report are those of the authors and do not necessarily represent the view of the Centers for Disease Control and Prevention.