A community outbreak of Legionnaires disease linked to hospital cooling towers: an epidemiological method to calculate dose of exposure

Transcription

1 International Epidemiological Association 1999 Printed in Great Britain International Journal of Epidemiology 1999;28: A community outbreak of Legionnaires disease linked to hospital cooling towers: an epidemiological method to calculate dose of exposure Clive M Brown, a Pekka J Nuorti, b Robert F Breiman, b A Leroy Hathcock, c Barry S Fields, b Harvey B Lipman, b Gerald C Llewellyn, c Jo Hofmann b and Martin Cetron b Background From July to September 1994, 29 cases of community-acquired Legionnaires disease (LD) were reported in Delaware. The authors conducted an investigation to a) identify the source of the outbreak and risk factors for developing Legionella pneumophila serogroup 1 (Lp-1) pneumonia and b) evaluate the risk associated with the components of cumulative exposure to the source (i.e. distance from the source, frequency of exposure, and duration of exposure). Methods Results Conclusion A case-control study matched 21 patients to three controls per case by known risk factors for acquiring LD. Controls were selected from patients who attended the same clinic as the respective case-patients. Water samples taken at the hospital, from eight nearby cooling towers, and from four of the patient s homes were cultured for Legionella. Isolates were subtyped using monoclonal antibody (Mab) analysis and arbitrarily primed polymerase chain reaction (AP-PCR). Eleven (52%) of 21 case-patients worked at or visited the hospital compared with 17 (27%) of 63 controls (OR 5.0, 95% CI : ). For those who lived, worked, or visited within 4 square miles of the hospital, the risk of illness decreased by 20% for each 0.10 mile from the hospital; it increased by 80% for each visit to the hospital; and it increased by 8% for each hour spent within miles of the hospital. Lp-1 was isolated from three patients and both hospital cooling towers. Based on laboratory results no other samples contained Lp-1. The clinical and main-tower isolates all demonstrated Mab pattern 1,2,5,6. AP-PCR matched the main-tower samples with those from two case-patients. The results of our investigation suggested that the hospital cooling towers were the source of a community outbreak of LD. Increasing proximity to and frequency of exposure to the towers increased the risk of LD. New guidelines for cooling tower maintenance are needed. Knowing the location of cooling towers could facilitate maintenance inspections and outbreak investigations. Keywords Legionnaires disease, disease outbreaks, dose-response, water microbiology, environmental exposure, aerosol exposure Accepted 17 September 1998 a Centers for Disease Control and Prevention Epidemiology Program Office, Division of Field Epidemiology. b Centers for Disease Control and Prevention, National Center for Infectious Diseases, Division of Bacterial and Mycotic Diseases. c Delaware Health and Social Services, Division of Public Health, Epidemiology Section. Reprint requests: Martin Cetron, Surveillance and Epidemiology Branch, Division of Quarantine, National Center for Infectious Diseases, Centers for Disease Control and Prevention,1600 Clifton Rd, MS-E03, Atlanta, GA 30333, USA. In July 1994, infection-control nurses at Hospital A in Wilmington, Delaware, observed an increase in the number of employees presenting to the employee health clinic with pneumonia. On 25 July 1994, the hospital notified the Delaware Division of Public Health (DE-DPH) that nine people with community-acquired pneumonia, including two hospital employees, had been admitted to the hospital. Among the nine, two were subsequently diagnosed with Legionnaires disease (LD) based on demonstration of urinary-antigen (UAg) for 353

2 354 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Legionella pneumophila serogroup 1 (Lp-1). Ultimately, from July through September 1994, a total of 29 cases of LD were reported to public health officials. Investigations of outbreaks of LD have demonstrated that the infection can be transmitted by aerosol-producing devices (e.g. cooling towers, 1 4 evaporative condensers, 5,6 whirlpool spas, 7 humidifiers, decorative fountains and mist machines 8,9 ) and by potable water aerosolized by shower-heads and tapwater faucets. 10 The primary objectives of this investigation were to determine the magnitude of the outbreak, to identify sources of transmission, to implement control measures to prevent further transmission of infection, and to evaluate the effectiveness of these control measures. The large number of confirmed cases in our outbreak and the clustering of cases in space and time allowed us to evaluate the risk associated with the components of cumulative exposure (i.e. distance from the source, frequency of exposure to the source, and duration of exposure to the source). Materials and Methods Case definition A case of community-acquired pneumonia was defined as, the acute onset of lower respiratory symptoms accompanied by evidence of a pulmonary infiltrate on chest radiograph in a person who lived in, worked in, or visited New Castle County after 1 June A case of community-acquired LD was defined as community-acquired pneumonia accompanied by laboratory evidence of acute infection with Legionella in any of the following ways: a) the isolation of Lp-1 from respiratory secretions; or b) the detection of Lp-1 antigens in urine by radioimmunoassay; or c) a fourfold or greater rise in titre to 1:128 of serum antibodies against Lp-1 by indirect fluorescent antibody (IFA) assay. A nosocomial case of LD was defined as a patient with a laboratory-confirmed diagnosis of LD who a) was an inpatient at Hospital A after 1 June 1994; b) demonstrated an acute onset of new, lower respiratory symptoms supported by chest radiograph evidence of pulmonary infiltrates; c) had either symptom onset 48 hour but 10 days after admission to hospital or symptom onset within 10 days of discharge from the hospital. Case finding and case investigation Cases of LD were identified retrospectively and prospectively by investigators from the DE-DPH, the Centers for Disease Control and Prevention (CDC), and the infection control staff of area hospitals for the period 1 June to 30 September Admission/discharge records, patient registers at the emergency room and the employee health clinic, and microbiology laboratory records at four local hospitals were reviewed to identify patients who had community-acquired pneumonia or a diagnosis of LD. Beginning on 27 July, individuals who had signs and symptoms consistent with community-acquired pneumonia were identified prospectively at local hospitals by daily review of a) admission records, b) the logs from emergency rooms and employee health clinics, and c) radiology reports. Patients were evaluated for infection with Legionella by urineantigen detection. When available, for patients identified prospectively and for those patients identified retrospectively who were still in the hospital, respiratory secretions were cultured for Legionella, and paired acute- and convalescentphase serum antibody levels against Lp-1 were measured by IFA. Attack rates for pre-intervention cases for whom place of residence was known, were calculated by census tract of residence based on population projections from the 1990 census. Information from open-ended interviews with casepatients was used to generate hypotheses about exposures to sources of Legionella and to design a standard questionnaire for use in a case-control study. Matched case-control study Based on the attack rates in the various census tracts, the hospital cooling tower was implicated as a likely source of the outbreak. A case-control study to confirm the association and to investigate further the importance of distance and duration of exposure was conducted. Entry into the case-control study was restricted to people who had confirmed LD, who lived in or had visited New Castle County and who had a date of onset from 2 June 1994 to 12 August Controls were randomly selected from among eligible outpatients who had visited each case-patient s primary or consulting physician within the last year. Controls who had a clinical diagnosis of pneumonia in the past 6 months were excluded, however controls were not tested for evidence of antibodies to LD. Three controls were matched to each patient based on the following criteria: 1) Age within 10 years of the age of the patient. 2) Health conditions ranked by increasing risk of acquiring LD Category 1: Healthy non-smoker (i.e. never smoked or quit smoking 1 year before study period); Category 2: Healthy smoker; Category 3: Chronic non-immunosuppressive illness (e.g. chronic obstructive pulmonary disease, congestive heart failure, diabetes mellitus, or chronic renal insufficiency); Category 4: Active immunosuppressive illness or therapy (e.g. HIV/AIDS, cancer, renal dialysis, or immunosuppressive medications). Assessment of exposure and analysis A standard questionnaire was used to interview case-patients and controls about specific exposures; it focused on the area near Hospital A where the attack rate was highest. To define the areas of exposure, interviewers were given maps that had the streets of the areas of investigation clearly marked. The maps were divided into a grid by marking concentric square blocks, defined by the streets in the area, of increasing distances around the hospital. The respective distance of the perimeter of each block from the hospital was miles, 0.25 miles, 0.5 miles, 0.75 miles and 1 mile; these distances were measured to the closest point on the outer perimeter of each block. Each case-patient and their three matched controls were asked about possible sites of exposure in the 2 weeks before the date of onset of illness of the case-patient. To evaluate the components of cumulative exposure, they were asked if they had lived in, worked in, or visited the Wilmington area. To determine which block was likely to contain the source of infection, the authors looked at the risk of illness from being within each

3 LEGIONNAIRES DISEASE AND HOSPITAL COOLING TOWERS 355 block as compared to not being within that block. To determine the frequency and duration of exposure, we asked individuals about the number of visits they made and the length of time they spent in each block. Separate logistic regression models were used to determine the change in risk with changes in frequency and duration of exposure within each block. To determine if there was a change in risk with change in distance from the cooling towers at Hospital A, a logistic regression model was designed in which a respondent s point of closest contact to that block was considered the only relevant exposure for that respondent. This model was used to evaluate whether a doseresponse effect existed between the amount of time spent in close proximity to the cooling towers at Hospital A and the likelihood of having LD. The dose variable, Aerosol Exposure Units (AEU), was defined as the time (t) spent at distance (d) from the source, divided by the distance (d) from the source (i.e. AEU = t/d, expressed in hours per mile from the source) and was calculated as follows: The average time spent per day in each block (average distances from Hospital A: miles, miles to 0.25 miles, 0.25 miles to 0.5 miles, 0.5 miles to 0.75 miles) was estimated by adding the time spent visiting, working (assuming an 8-hour workday), or living in each block. Individuals living in a block were assumed to have spent the remainder of their day otherwise unaccounted for (to complete 24 hours) in that block. Individuals not living within one mile of the outbreak area were assumed to have spent the remainder of their day at a distance of at least one mile from the cooling tower. The dose was computed as the sum of the average time spent in each block divided by the average distance to the cooling towers at Hospital A for that block. The doses were standardized to assign a minimum dose of 0 to anyone who spent all 24 hours per day more than 0.75 miles from the cooling tower. For example; A person spending 12 hours at a distance of 0.25 mile from the tower would be determined to have been exposed to a dose of 12/0.25 = 48 AEU. Data were entered in Epi Info 11 version 6. Univariate analyses were done in Epi Info and matched logistic-regression analyses were done using SAS 12 version Odds ratios were used to estimate the risk of disease. Environmental investigation and intervention The attack rates for illness (Table 1) and the distribution of cases by census tract of residence (Figure 1) guided the locations for environmental sampling. Area cooling towers were identified by taking panoramic photographic views of the area from the roof of Hospital A, the highest point in the city. These photographs were used to identify buildings likely to have cooling towers. Investigators drove along streets within a 2-mile radius of Hospital A to identify buildings with aerosol-producing devices. Ten evaporative cooling towers within one mile of the hospital were identified from 27 July to 16 August 1994, and water from these was cultured for Legionella. Water samples were cultured from potable water within the inpatient areas in Hospital A and from the hospital s hot water supply. On the basis of results that suggested the main hospital tower was the most likely source (Tables 1 and 2), the maintenance procedure and logs for the two cooling towers and the air-handling Table 1 Attack rate and distribution of pre-intervention cases by census tract of residence and attack rate among hospital staff Tract LD cases Population Rate/ Rate ratio Hospital A (Staff) A B C D E F G H I J K L M N O (referent) Total Census tracts are labelled by their proximity to the census tract with the cooling tower where Lp-1 was isolated. Tracts A to E (Figure 1) were within an approximate 0.5 mile radius of the hospital. Tracts F and G were just 1 mile from the hospital, and tracts H to O were 2 miles from the hospital. system at the hospital were reviewed. Water samples from Hospital A and the other environmental samples were sent to CDC. The techniques for culturing water samples 13 and identifying isolates by direct flourescent assay 14 followed procedures outlined by CDC. Laboratory investigation Urine specimens from people whose illness met the case definition for community-acquired pneumonia were tested for presence of antigens of Lp-1 by radioimmunoassay (RIA). 15 Culturing of respiratory secretions for Legionella species was done at the microbiology laboratory of each hospital. Paired acute- and convalescent-phase sera were collected and sent to the DE-DPH laboratory for serologic determination of antibody titre against Lp-1. Environmental and clinical isolates were sub-typed using monoclonal-antibody (Mab) 16 analysis and arbitrarily primed polymerase-chain reaction (AP-PCR). 17 Results Case finding and descriptive epidemiology Sixty-five patients with community-acquired pneumonia were identified; 29 met the case definition for LD based on detection of antigens of Lp-1 in their urine. For 22 of the patients with LD, onset of illness occurred during the study period (25 June to 12 August 1994) (Figure 2); these 22 patients were included in our case-control study. In three of the 22 patients, respiratory secretions were positive for Legionella. Peak onsets of illness were from 27 to 31 July The median age of the 22 patients was 53.5 years (range: years). By health status, 14% were in category 1, 59% in category 2, and 27% in category 3. No deaths occurred among case-patients during this period. No cases of nosocomial LD were identified.

4 356 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY Figure 1 Spot map of Legionnaires disease cases by census tract of patient residence (not to scale). The 15 individual cases of Legionnaires disease are represented by dots. Seven other case-patients (not shown) lived in census tracts farther from the contaminated cooling towers. The shaded areas of increasing size around Hospital A represent the blocks used for the epidemiological investigation Table 2 Results of cooling tower testing at Hospital A, July August 1994 Result Test Colony forming Result (Sample no.) a Date units (CFU) date Main tower samples (1) 07/22/ /27/94 (2) 07/25/ /29/94 (3) 07/25/ /29/94 (4) 07/28/ /02/94 (5) 07/28/ /02/94 Small tower samples (1) 07/22/ /27/94 (2) 07/25/ /29/94 (3) 07/25/ /29/94 a Main tower: All samples were taken from a part of the same closed system. Samples (2 & 3) & (4 & 5) were duplicate. Small tower: Samples 2 & 3 were duplicate. Attack rates for the 3-month period of July to September 1994 (Table 1) were highest among the hospital staff (124.7 cases/ ) and among residents within the census tract immediately adjacent to the hospital (i.e. tract B; cases/ ). Case-control study and analysis Twenty-one of 22 case-patients identified during the study period were available for interview, and three matched controls were identified for each patient. Eleven (52%) of the 22 casepatients worked at or visited the hospital, compared with 17 (27%) of 63 controls (odds ratio [OR] 5.0; 95% confidence interval [CI] : ). The risk was lower for visiting areas 0.25 mile from Hospital A. In matched logistic regression analyses, the risk of illness decreased by 20% for each 0.1 mile increase in distance from Hospital A up to one mile from the hospital (Figure 3). Risk of illness varied with the frequency and duration of visits to the area. Within the 2-week period of interest, for every visit made to the block where Hospital A was located, the risk of illness increased by 80%. The risk of illness was lower for visits to blocks further from the hospital; there was no increased risk for visits to areas 1 mile from Hospital A. For each hour spent in blocks within miles of Hospital A, the risk of illness increased by 8%. The increase in risk of illness due to duration of time spent in a block was lower for time spent in blocks more distant from Hospital A; the risk was negligible 1 mile from the block with Hospital A. The results of the dose-response model demonstrated that the median dose of exposure was higher for the 21 case-patients (58 AEU) compared with that for the 63 controls (6 AEU) (Figure 4). Environmental and laboratory investigation Samples from both evaporative cooling towers at the hospital contained Lp-1 (Table 2). Legionella was not isolated from samples from the eight other cooling towers or from potable water samples obtained at four patients homes. The Mab subtype of isolates from the main cooling tower was subtype 1,2,5,6, which was the same as three clinical isolates. Two of these clinical isolates also were indistinguishable from isolates from the main cooling tower, as indicated by AP-PCR subtyping. The other clinical isolate had a different AP-PCR pattern and was from a person who stated that they had not been within a mile of the hospital. The isolate from the small hospital cooling tower was inadvertently discarded. The main tower was hyperchlorinated on 27 and 28 July; it was hyperchlorinated again and mechanically cleaned on 29 July. The small tower was hyperchlorinated and mechanically cleaned on 2 August.

5 LEGIONNAIRES DISEASE AND HOSPITAL COOLING TOWERS 357 Figure 2 Cases of Legionnaires disease by date of onset of symptoms and date of environmental interventions. Black bars represent pre-intervention cases. Grey bars represent post-intervention cases. The asterisks represent the two cases that had a different AP-PCR sub-typing pattern from that which was found in samples from the hospital s main cooling tower Figure 3 Change in risk of illness as a function of proximity to Hospital A. This graph shows the result of the model estimating the change in the OR with increasing distance from the hospital. There was a 20% reduction in risk for each additional tenth of a mile from the hospital; at approximately one mile, there was no increased risk of illness. The model was statistically significant, P 0.04 Cultures of post-intervention water samples taken at biweekly intervals starting on 29 July were still negative for Lp-1 on 30 November Hospital engineering staff indicated that the maintenance procedures that were used for the cooling towers and airhandling system were in accord with protocols described in the manufacturer s guidelines; the staff also indicated that the main cooling tower at Hospital A worked at maximal capacity throughout the summer. The air-intake vents of the air-handling system for the administrative and family practice areas at the hospital which are located about 100 yards from the main tower were protected by single filters. The air-intake vents of the airhandling system for the inpatient and Intensive Care Unit (ICU) areas which are located about 100 yards from the small tower, were protected by double high-efficiency particulate air (HEPA) filters. Seven cases of LD occurred during the post-intervention period. In contrast to cases that occurred in the pre-intervention period, the median age of patients was 64 years. Of the seven patients, none were healthy, 43% had illness in the category 3 classification, and 57% had illness in category 4. Two (29%) Figure 4 Cumulative exposure by case-control status. This is a model estimating the dose of exposure for patients and controls. The solid line represents the percentage of people that had Legionnaires disease exposed at each dose; the broken line represents the percentage of controls exposed. The exposure dose (AEU) was based on the length of time spent at each distance (AEU = t/d) of the seven died as a result of LD. By hospital of diagnosis, 55% of the pre-intervention cases were diagnosed at Hospital A compared with other area hospitals, whereas none of the postintervention cases were diagnosed at Hospital A. In addition to the three isolates from case-patients identified in the preintervention period, a fourth isolate was obtained from a postintervention patient. This patient stated that they had not been within a mile of Hospital A. The isolate from this patient had a different AP-PCR pattern from the hospital tower isolates. Discussion The epidemiological and microbiological data suggest that contaminated aerosols from one or both cooling towers at Hospital A were responsible for the outbreak. The dose-response data suggests that transmission occurred primarily within 0.25 miles of the cooling towers; however, the possibility cannot be ruled out that aerosols travelled further to transmit disease to some patients. In 1986, an investigation in Wisconsin 2 suggested a

6 358 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY range of distance of transmission up to 2 miles from the source; however, other studies 4,18 have demonstrated more limited ranges. In the present study case-patients were more likely than controls to have had frequent exposure or long duration of exposure to the source, suggesting that in addition to proximity, cumulative exposure may have been an important risk factor for illness. Other studies, 2 5,18,19 have shown that risk of infection was associated with proximity to the source. Our study attempted to quantify the level of exposure to the source by combining the effects of exposure due to proximity to the source and the duration of time spent at that point. Our dose variable, aerosol exposure units, was much higher for case-patients than for controls suggesting that it may be a good proxy for the infecting dose of organism. This approach should be tried in other studies to quantify the level of exposure to LD and other environmentally-acquired diseases. Manufacturers of cooling towers should consider modifications that reduce the release of aerosolized water, thereby reducing the infecting dose of Legionella in the atmosphere. After the interventions to the cooling towers at Hospital A, area physicians and health officials continued active, prospective surveillance for cases of LD for 10 weeks after the completion of the investigation. Active surveillance was done by screening all admissions for pneumonia using the UAg test for detecting Lp-1. 13,20 Because there was no active background surveillance for LD, we cannot say whether the seven cases of LD that were identified at 2 weeks post-intervention reflected the background of sporadic, community-acquired LD, or whether they were related to a second outbreak. However, information from the investigation of these seven cases did not show a cluster based on area of residence, work, or other routine activities. Epidemiologically, these seven cases did not appear to be related to the outbreak cluster and molecular subtyping by AP-PCR indicated that at least one of these seven patients was infected with a strain of Lp-1 that was unrelated to the outbreak strain. Because the outbreak was widely reported in the media, study participants may have been more likely to remember and report visiting the block where Hospital A was located versus visiting other blocks. Because registration of cooling towers is not required, identification of towers and other aerosolproducing devices depended on the author s inspection, which could have missed potential sources of disease transmission. Contaminated cooling towers have transmitted LD by exposing occupants of buildings to contaminated aerosols via open windows 4,18 and through air-intake vents and air-handling systems. 21,22 The absence of nosocomial cases of LD was notable. Nosocomial transmission may have been prevented by the hospital s use of double HEPA filters on air-intake vents of the inpatient and ICU areas. The filters were not cultured, but no Lp-1 isolates were found in water from the condensate tray between the two HEPA filters. Windows in patient areas were kept closed. The meteorological office at the local aerodrome reported that during the period of the outbreak the prevailing winds were mostly from the south-west. From our vantage point on top of the hospital we saw that the mist from the cooling tower was carried in different directions at different times of the day. In addition, Hospital A, a multi-storied structure lay to the north of both cooling towers. This is the probable reason why most cases lived to the south and east of the hospital in tract B. Mab subtyping of the clinical and environmental isolates supported the epidemiological data linking the cases and the hospital cooling towers. AP-PCR added further discrimination to Mab subtyping. The cooling tower s maintenance logs suggest that routine procedures were followed. Intensive use of cooling towers, such as may occur in conditions of extreme temperature and/or humidity, may have provided more opportunity for contamination of the towers by increasing the intake of make-up water. Under such conditions there may be a need for increased biocide over that required routinely. Most of the estimated annual cases of LD in the US occur sporadically. 23,24 Although contaminated cooling towers are an important cause of outbreaks of LD, 1 4,18,21,25 the proportion of disease occurring as a result of exposure to cooling towers is unknown. Prospective epidemiological studies are needed to better define the attributable risk for contracting LD from cooling towers and other disseminators of Legionella and for developing cost-effective prevention guidelines. Practices based on current recommendations do not ensure that a cooling tower will not serve as a source of Legionella, we recommend that maintenance procedures for cooling towers be reviewed and new guidelines be implemented, especially under conditions that lead to prolonged or intensive use. New towers, and towers recently started up after shut-down, are at increased risk for the growth of Legionella. 3,25 We recommend that mechanical cleaning be done routinely at the start-up of towers. The registration of cooling towers and evaporative condensers would facilitate their identification as potential sources during outbreak investigations. Acknowledgements We thank Dr Paul Edelstein, Clinical Microbiology Laboratory, Hospital of the University of Pennsylvania, who performed the urinary antigen tests on the samples; Robert Benson and Janet Pruckler, Respiratory Diseases Laboratory Section, Childhood and Respiratory Diseases Branch, National Center for Infectious Diseases, Centers for Disease Control and Prevention, who performed culture of the water samples, monclonal analysis, and AP-PCR; Dr Dave Verma and the staff of the Public Health Laboratory, Delaware Health and Social Services, who analysed the acute- and convalescent-phase sera; Dr Edward Montz, President, Indoor Air Solutions, Inc., Pottstown, PA; and Dr Chin Yang, P&K Microbiology Services, Cherry Hill, NJ. References 1 Centers for Disease Control and Prevention. Legionnaires disease associated with cooling-towers. MMWR 1994;43:491 3,9. 2 Addis DG, Davis JP, Martin L et al. Community-acquired Legionnaires disease associated with a cooling-tower: evidence for longer distance transport of Legionella pneumophila. Am J Epidemiol 1989;130: Addis DG, Davis JP, Wand PJ et al. Two cases of community-acquired Legionnaires disease: evidence for association with a cooling-tower. J Infect Dis 1989;159: Klaucke DN, Vogt DL, LaRue D et al. Legionnaires disease: the epidemiology of two outbreaks in Burlington, Vermont Am J Epidemiol 1984;119: Cordes GC, Fraser DW, Skaliy P et al. Legionnaires disease outbreak at an Atlanta, Georgia, country club: evidence for spread from an evaporative condenser. Am J Epidemiol 1980;111:

7 LEGIONNAIRES DISEASE AND HOSPITAL COOLING TOWERS Breiman RF, Cozen W, Fields BS et al. Role of air sampling in investigation of an outbreak of Legionnaires disease associated with exposure to aerosols from an evaporative condenser. J Infect Dis 1990; 161: Jernigan D, Hofmann J, Cetron MS et al. and the Cruise Ship Working Group. Outbreak of Legionnaires Disease among cruise ship passengers exposed to a contaminated whirlpool spa. Lancet 1996;347: Hlady WG, Mullen RC, Mintz CS et al. Outbreak of Legionnaires disease linked to a decorative fountain by molecular epidemiology. Am J Epidemiol 1993;138: Mahoney FJ, Hoge CW, Farley TA et al. Communitywide outbreak of Legionnaires disease associated with a grocery store mist machine. J Infect Dis 1992;165: Hanrahan JP, Dale LM, Scharf VB et al. A community outbreak of legionellosis: transmission by potable hot water. Am J Epidemiol 1987; 125: Dean AG, Dean JA, Coulombier D et al. Epi Info, Version 6: A Word Processing, Database, and Statistics Program for Epidemiology and Microcomputers. Atlanta, Georgia: Centers for Disease Control and Prevention, USA, SAS for Personal Computers, Release 6.08, Cary, NC: SAS Institute, Barbaree JM, Fields BS, Martin WT et al. Procedures for the Recovery of Legionella from the Environment. Centers for Disease Control and Prevention, Atlanta, Nov Wilkinson HW. Hospital Laboratory Diagnosis of Legionella Infections. Centers for Disease Control and Prevention, Atlanta, Jan Kohler RB, Zimmerman SE, Wilson SD. Rapid radioimmunoassay diagnosis of Legionnaires disease, detection and partial characterization of urinary antigen. Ann Intern Med 1981;94: Luck PC, Helbig JH, Gunter U et al. Epidemiologic investigation by macrorestriction analysis and by using monoclonal antibodies of nosocomial pneumonia caused by Legionella pneumophila serogroup 1. J Clin Microbiol 1994;32: Belkum AV, Struelens M, Quint W. Typing of Legionella pnuemophila strains by polymerase chain reaction mediated DNA fingerprinting. J Clin Microbiol 1993;31: Garbe PL, Davis BJ, Weisfeld JS et al. Nosocomial Legionnaires disease, epidemiologic demonstration of cooling-towers as a source. JAMA 1985;254: Bhopal RS, Fallon RJ, Buist EC, Black RJ, Urquhart JD. Proximity of the home to a cooling tower and risk of non-outbreak legionnaires disease. Br Med J 1991;302: Plouffe JF, File TM Jr, Breiman R et al. Reevaluation of the definition of Legionnaires disease: use of the urinary antigen assay. Community Based Pneumonia Incidence Study Group. Clin Infect Dis 1995;20: Dondero TJ, Rendtorff RC, Mallison GF et al. An outbreak of Legionnaires disease associated with a contaminated air-conditioning cooling-tower. N Engl J Med 1980;302: Maesaki S, Kohno S, Koga H et al. An outbreak of Legionnaires disease in a nursing home. Intern Med 1992;31: Marston BJ, Plouffe JF, File TM et al. Incidence of communityacquired pneumonia requiring hospitalization: population-based active surveillance in Ohio. Arch Intern Med 1997;157: Marston BJ, Lipman HB, Breiman RF. Surveillance for Legionnaires disease. Risk factors for morbidity and mortality. Arch Intern Med 1994;154: Bentham RH, Broadbent CR. A model for autumn outbreaks of Legionnaires disease associated with cooling-towers, linked to system operation and size. Epidemiol Infect 1993;111: