A community outbreak of Legionnaires' disease linked to hospital

International Journal of Epidemiology 1999;28:353–359

In July 1994, infection-control nurses at Hospital A inWilmington, Delaware, observed an increase in the number ofemployees presenting to the employee health clinic with pneu-monia. On 25 July 1994, the hospital notified the DelawareDivision of Public Health (DE-DPH) that nine people withcommunity-acquired pneumonia, including two hospitalemployees, 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

© International Epidemiological Association 1999 Printed in Great Britain

A community outbreak of Legionnaires’disease linked to hospital cooling towers: an epidemiological method to calculate doseof exposureClive 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 Hofmannb and Martin Cetronb

Background From July to September 1994, 29 cases of community-acquired Legionnaires’disease (LD) were reported in Delaware. The authors conducted an investigationto a) identify the source of the outbreak and risk factors for developing Legionellapneumophila serogroup 1 (Lp-1) pneumonia and b) evaluate the risk associatedwith the components of cumulative exposure to the source (i.e. distance from thesource, frequency of exposure, and duration of exposure).

Methods A case-control study matched 21 patients to three controls per case by knownrisk factors for acquiring LD. Controls were selected from patients who attendedthe 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 werecultured for Legionella. Isolates were subtyped using monoclonal antibody (Mab)analysis and arbitrarily primed polymerase chain reaction (AP-PCR).

Results Eleven (52%) of 21 case-patients worked at or visited the hospital compared with17 (27%) of 63 controls (OR 5.0, 95% CI : 1.1–29). For those who lived, worked,or visited within 4 square miles of the hospital, the risk of illness decreased by20% for each 0.10 mile from the hospital; it increased by 80% for each visit tothe hospital; and it increased by 8% for each hour spent within 0.125 miles ofthe hospital. Lp-1 was isolated from three patients and both hospital coolingtowers. Based on laboratory results no other samples contained Lp-1. The clinicaland main-tower isolates all demonstrated Mab pattern 1,2,5,6. AP-PCR matchedthe main-tower samples with those from two case-patients.

Conclusion The results of our investigation suggested that the hospital cooling towers werethe source of a community outbreak of LD. Increasing proximity to and fre-quency of exposure to the towers increased the risk of LD. New guidelines forcooling tower maintenance are needed. Knowing the location of cooling towerscould 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 InfectiousDiseases, 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 forDisease Control and Prevention,1600 Clifton Rd, MS-E03, Atlanta, GA 30333,USA.

353

Legionella pneumophila serogroup 1 (Lp-1). Ultimately, from July through September 1994, a total of 29 cases of LD werereported 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 whirlpoolspas,7 humidifiers, decorative fountains and mist machines8,9)and by potable water aerosolized by shower-heads and tap-water faucets.10 The primary objectives of this investigationwere to determine the magnitude of the outbreak, to identifysources of transmission, to implement control measures to pre-vent further transmission of infection, and to evaluate theeffectiveness of these control measures. The large number ofconfirmed cases in our outbreak and the clustering of cases inspace and time allowed us to evaluate the risk associated withthe components of cumulative exposure (i.e. distance from thesource, frequency of exposure to the source, and duration ofexposure to the source).

Materials and MethodsCase definition

A case of community-acquired pneumonia was defined as, theacute onset of lower respiratory symptoms accompanied by evid-ence of a pulmonary infiltrate on chest radiograph in a personwho lived in, worked in, or visited New Castle County after 1 June 1994. A case of community-acquired LD was defined ascommunity-acquired pneumonia accompanied by laboratoryevidence of acute infection with Legionella in any of the follow-ing ways:

a) the isolation of Lp-1 from respiratory secretions; orb) the detection of Lp-1 antigens in urine by radioimmuno-assay; orc) a fourfold or greater rise in titre to >1:128 of serum anti-bodies against Lp-1 by indirect fluorescent antibody (IFA) assay.

A nosocomial case of LD was defined as a patient with alaboratory-confirmed diagnosis of LD who a) was an inpatientat Hospital A after 1 June 1994; b) demonstrated an acute onsetof new, lower respiratory symptoms supported by chest radio-graph evidence of pulmonary infiltrates; c) had either symptomonset .48 hour but <10 days after admission to hospital orsymptom onset within 10 days of discharge from the hospital.

Case finding and case investigation

Cases of LD were identified retrospectively and prospectively byinvestigators from the DE-DPH, the Centers for Disease Controland Prevention (CDC), and the infection control staff of areahospitals for the period 1 June to 30 September 1994. Admis-sion/discharge records, patient registers at the emergency roomand the employee health clinic, and microbiology laboratoryrecords at four local hospitals were reviewed to identify patientswho had community-acquired pneumonia or a diagnosis of LD.

Beginning on 27 July, individuals who had signs and symp-toms consistent with community-acquired pneumonia wereidentified prospectively at local hospitals by daily review of a)admission records, b) the logs from emergency rooms andemployee health clinics, and c) radiology reports.

Patients were evaluated for infection with Legionella by urine-antigen detection. When available, for patients identified

prospectively and for those patients identified retrospectivelywho were still in the hospital, respiratory secretions werecultured for Legionella, and paired acute- and convalescent-phase serum antibody levels against Lp-1 were measured byIFA. Attack rates for pre-intervention cases for whom place ofresidence was known, were calculated by census tract ofresidence based on population projections from the 1990census. Information from open-ended interviews with case-patients was used to generate hypotheses about exposures tosources of Legionella and to design a standard questionnaire foruse in a case-control study.

Matched case-control study

Based on the attack rates in the various census tracts, thehospital cooling tower was implicated as a likely source of theoutbreak. A case-control study to confirm the association and toinvestigate further the importance of distance and duration ofexposure was conducted. Entry into the case-control study wasrestricted to people who had confirmed LD, who lived in or had visited New Castle County and who had a date of onsetfrom 2 June 1994 to 12 August 1994.

Controls were randomly selected from among eligible out-patients who had visited each case-patient’s primary or consult-ing physician within the last year. Controls who had a clinicaldiagnosis of pneumonia in the past 6 months were excluded,however controls were not tested for evidence of antibodies toLD. Three controls were matched to each patient based on thefollowing criteria:

1) Age within 10 years of the age of the patient.2) Health conditions ranked by increasing risk of acquiring LDCategory 1: Healthy non-smoker (i.e. never smoked or quitsmoking >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 immunosuppressivemedications).

Assessment of exposure and analysis

A standard questionnaire was used to interview case-patientsand controls about specific exposures; it focused on the areanear Hospital A where the attack rate was highest. To define theareas of exposure, interviewers were given maps that had thestreets of the areas of investigation clearly marked. The mapswere divided into a grid by marking concentric square blocks,defined by the streets in the area, of increasing distances aroundthe hospital. The respective distance of the perimeter of eachblock from the hospital was 0.125 miles, 0.25 miles, 0.5 miles,0.75 miles and 1 mile; these distances were measured to theclosest point on the outer perimeter of each block.

Each case-patient and their three matched controls wereasked about possible sites of exposure in the 2 weeks before thedate of onset of illness of the case-patient. To evaluate the com-ponents of cumulative exposure, they were asked if they hadlived in, worked in, or visited the Wilmington area. To deter-mine which block was likely to contain the source of infection,the authors looked at the risk of illness from being within each

354 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY

block as compared to not being within that block. To determinethe frequency and duration of exposure, we asked individualsabout the number of visits they made and the length of timethey spent in each block. Separate logistic regression modelswere used to determine the change in risk with changes in fre-quency and duration of exposure within each block. To deter-mine if there was a change in risk with change in distance fromthe cooling towers at Hospital A, a logistic regression model was designed in which a respondent’s point of closest contact tothat block was considered the only relevant exposure for thatrespondent. This model was used to evaluate whether a dose-response effect existed between the amount of time spent inclose proximity to the cooling towers at Hospital A and thelikelihood of having LD.

The dose variable, Aerosol Exposure Units (AEU), was definedas the time (t) spent at distance (d) from the source, divided bythe distance (d) from the source (i.e. AEU = t/d, expressed inhours per mile from the source) and was calculated as follows:

The average time spent per day in each block (average distancesfrom Hospital A: ,0.125 miles, 0.125 miles to ,0.25 miles, 0.25 miles to ,0.5 miles, 0.5 miles to 0.75 miles) was estimatedby adding the time spent visiting, working (assuming an 8-hourworkday), or living in each block.

Individuals living in a block were assumed to have spent the remainder of their day otherwise unaccounted for (to com-plete 24 hours) in that block. Individuals not living within one mile of the outbreak area were assumed to have spent theremainder of their day at a distance of at least one mile from thecooling tower.

The dose was computed as the sum of the average time spentin each block divided by the average distance to the coolingtowers at Hospital A for that block. The doses were standardizedto assign a minimum dose of 0 to anyone who spent all 24hours 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 doseof 12/0.25 = 48 AEU.

Data were entered in Epi Info11 version 6. Univariateanalyses were done in Epi Info and matched logistic-regressionanalyses were done using SAS12 version 6.08. Odds ratios wereused to estimate the risk of disease.

Environmental investigation and intervention

The attack rates for illness (Table 1) and the distribution of casesby census tract of residence (Figure 1) guided the locations forenvironmental sampling. Area cooling towers were identifiedby taking panoramic photographic views of the area from theroof of Hospital A, the highest point in the city. These photo-graphs were used to identify buildings likely to have coolingtowers. Investigators drove along streets within a 2-mile radiusof Hospital A to identify buildings with aerosol-producingdevices. Ten evaporative cooling towers within one mile of thehospital were identified from 27 July to 16 August 1994, andwater from these was cultured for Legionella. Water sampleswere cultured from potable water within the inpatient areas inHospital A and from the hospital’s hot water supply. On thebasis of results that suggested the main hospital tower was the most likely source (Tables 1 and 2), the maintenance pro-cedure and logs for the two cooling towers and the air-handling

system at the hospital were reviewed. Water samples from Hos-pital A and the other environmental samples were sent to CDC.The techniques for culturing water samples13 and identifyingisolates by direct flourescent assay14 followed procedures outlinedby CDC.

Laboratory investigation

Urine specimens from people whose illness met the case defini-tion for community-acquired pneumonia were tested for presenceof antigens of Lp-1 by radioimmunoassay (RIA).15 Culturing ofrespiratory secretions for Legionella species was done at themicrobiology laboratory of each hospital. Paired acute- andconvalescent-phase sera were collected and sent to the DE-DPHlaboratory for serologic determination of antibody titre againstLp-1. Environmental and clinical isolates were sub-typed usingmonoclonal-antibody (Mab)16 analysis and arbitrarily primedpolymerase-chain reaction (AP-PCR).17

ResultsCase finding and descriptive epidemiology

Sixty-five patients with community-acquired pneumonia wereidentified; 29 met the case definition for LD based on detectionof antigens of Lp-1 in their urine. For 22 of the patients withLD, onset of illness occurred during the study period (25 Juneto 12 August 1994) (Figure 2); these 22 patients were includedin our case-control study. In three of the 22 patients, respiratorysecretions were positive for Legionella. Peak onsets of illness werefrom 27 to 31 July 1994. The median age of the 22 patients was 53.5 years (range: 29–80 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. Nocases of nosocomial LD were identified.

LEGIONNAIRES’ DISEASE AND HOSPITAL COOLING TOWERS 355

Table 1 Attack rate and distribution of pre-intervention cases bycensus tract of residence and attack rate among hospital staff

Tract LD cases Population Rate/100 000 Rate ratio

Hospital A (Staff) 2 1604 124.7 12.2

A 1 2301 43.5 4.3

B 5 3585 139.5 13.7

C 1 3840 26.0 2.6

D 1 4595 21.8 2.1

E 2 3410 58.7 5.8

F 2 3767 53.1 5.2

G 1 3122 32.0 3.2

H 1 3174 31.5 3.1

I 1 3810 26.2 2.6

J 1 3350 29.9 2.9

K 1 2794 35.8 3.5

L 1 2138 46.8 4.6

M 1 4781 20.9 2.1

N 1 5367 18.6 1.8

O 1 9845 10.2 (referent)

Total 21 59 879 35.1

Census tracts are labelled by their proximity to the census tract with thecooling tower where Lp-1 was isolated. Tracts A to E (Figure 1) were withinan approximate 0.5 mile radius of the hospital. Tracts F and G were just .1mile from the hospital, and tracts H to O were .2 miles from the hospital.

Attack rates for the 3-month period of July to September1994 (Table 1) were highest among the hospital staff (124.7cases/100 000) and among residents within the census tractimmediately adjacent to the hospital (i.e. tract B; 139.5 cases/100 000).

Case-control study and analysis

Twenty-one of 22 case-patients identified during the studyperiod were available for interview, and three matched controlswere identified for each patient. Eleven (52%) of the 22 case-patients worked at or visited the hospital, compared with 17(27%) of 63 controls (odds ratio [OR] 5.0; 95% confidenceinterval [CI] : 1.1–29). The risk was lower for visiting areas.0.25 mile from Hospital A.

In matched logistic regression analyses, the risk of illnessdecreased by 20% for each 0.1 mile increase in distance fromHospital A up to one mile from the hospital (Figure 3). Risk ofillness varied with the frequency and duration of visits to thearea. Within the 2-week period of interest, for every visit madeto the block where Hospital A was located, the risk of illnessincreased by 80%. The risk of illness was lower for visits toblocks further from the hospital; there was no increased risk forvisits to areas >1 mile from Hospital A. For each hour spent in blocks within 0.125 miles of Hospital A, the risk of illnessincreased by 8%. The increase in risk of illness due to durationof time spent in a block was lower for time spent in blocks moredistant from Hospital A; the risk was negligible >1 mile fromthe block with Hospital A.

The results of the dose-response model demonstrated that themedian dose of exposure was higher for the 21 case-patients (58AEU) compared with that for the 63 controls (6 AEU) (Figure 4).

Environmental and laboratory investigation

Samples from both evaporative cooling towers at the hospitalcontained Lp-1 (Table 2). Legionella was not isolated fromsamples from the eight other cooling towers or from potablewater samples obtained at four patients’ homes. The Mabsubtype of isolates from the main cooling tower was subtype1,2,5,6, which was the same as three clinical isolates. Two ofthese clinical isolates also were indistinguishable from isolatesfrom the main cooling tower, as indicated by AP-PCR subtyping.The other clinical isolate had a different AP-PCR pattern andwas from a person who stated that they had not been within amile of the hospital. The isolate from the small hospital coolingtower was inadvertently discarded. The main tower washyperchlorinated on 27 and 28 July; it was hyperchlorinatedagain and mechanically cleaned on 29 July. The small towerwas hyperchlorinated and mechanically cleaned on 2 August.


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 aroundHospital A represent the blocks used for the epidemiological investigation

Table 2 Results of cooling tower testing at Hospital A, July–August 1994

ResultTest Colony forming Result(Sample no.)a Date units (CFU) date

Main tower samples

(1) 07/22/94 7290 07/27/94

(2) 07/25/94 2320 07/29/94

(3) 07/25/94 3140 07/29/94

(4) 07/28/94 6840 08/02/94

(5) 07/28/94 9150 08/02/94

Small tower samples

(1) 07/22/94 105 07/27/94

(2) 07/25/94 2280 07/29/94

(3) 07/25/94 2340 07/29/94

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 wereduplicate.

Cultures of post-intervention water samples taken at biweeklyintervals starting on 29 July were still negative for Lp-1 on 30November 1994.

Hospital engineering staff indicated that the maintenanceprocedures that were used for the cooling towers and air-handling system were in accord with protocols described in themanufacturer’s guidelines; the staff also indicated that the maincooling tower at Hospital A worked at maximal capacity through-out the summer. The air-intake vents of the air-handling systemfor the administrative and family practice areas at the hospitalwhich are located about 100 yards from the main tower wereprotected by single filters. The air-intake vents of the air-handling 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-interventionperiod. In contrast to cases that occurred in the pre-interventionperiod, the median age of patients was 64 years. Of the sevenpatients, none were healthy, 43% had illness in the category 3classification, and 57% had illness in category 4. Two (29%)

of the seven died as a result of LD. By hospital of diagnosis, 55% of the pre-intervention cases were diagnosed at Hospital Acompared with other area hospitals, whereas none of the post-intervention cases were diagnosed at Hospital A. In addition to the three isolates from case-patients identified in the pre-intervention period, a fourth isolate was obtained from a post-intervention patient. This patient stated that they had not beenwithin a mile of Hospital A. The isolate from this patient had adifferent AP-PCR pattern from the hospital tower isolates.

DiscussionThe epidemiological and microbiological data suggest that con-taminated aerosols from one or both cooling towers at HospitalA were responsible for the outbreak. The dose-response datasuggests that transmission occurred primarily within 0.25 milesof the cooling towers; however, the possibility cannot be ruledout that aerosols travelled further to transmit disease to somepatients. In 1986, an investigation in Wisconsin2 suggested a


Figure 2 Cases of Legionnaires’ disease by date of onset of symptoms and date of environmentalinterventions. Black bars represent pre-intervention cases. Grey bars represent post-interventioncases. The asterisks represent the two cases that had a different AP-PCR sub-typing pattern fromthat which was found in samples from the hospital’s main cooling tower

Figure 3 Change in risk of illness as a function of proximity toHospital A. This graph shows the result of the model estimating thechange in the OR with increasing distance from the hospital. Therewas a 20% reduction in risk for each additional tenth of a mile fromthe hospital; at approximately one mile, there was no increased risk of illness. The model was statistically significant, P , 0.04

Figure 4 Cumulative exposure by case-control status. This is a modelestimating the dose of exposure for patients and controls. The solid linerepresents the percentage of people that had Legionnaires’ diseaseexposed at each dose; the broken line represents the percentage ofcontrols exposed. The exposure dose (AEU) was based on the length oftime spent at each distance (AEU = t/d)

range of distance of transmission up to 2 miles from the source;however, other studies4,18 have demonstrated more limitedranges. In the present study case-patients were more likely thancontrols to have had frequent exposure or long duration ofexposure to the source, suggesting that in addition to proximity,cumulative exposure may have been an important risk factorfor illness. Other studies,2–5,18,19 have shown that risk of infec-tion was associated with proximity to the source. Our studyattempted to quantify the level of exposure to the source by com-bining the effects of exposure due to proximity to the sourceand the duration of time spent at that point. Our dose variable,‘aerosol exposure units’, was much higher for case-patientsthan for controls suggesting that it may be a good proxy for theinfecting dose of organism. This approach should be tried inother studies to quantify the level of exposure to LD and otherenvironmentally-acquired diseases. Manufacturers of coolingtowers should consider modifications that reduce the release of aerosolized water, thereby reducing the infecting dose ofLegionella in the atmosphere.

After the interventions to the cooling towers at Hospital A,area physicians and health officials continued active, prospect-ive surveillance for cases of LD for 10 weeks after the com-pletion of the investigation. Active surveillance was done byscreening all admissions for pneumonia using the UAg test fordetecting Lp-1.13,20 Because there was no active backgroundsurveillance for LD, we cannot say whether the seven cases ofLD that were identified at .2 weeks post-intervention reflectedthe background of sporadic, community-acquired LD, or whetherthey were related to a second outbreak. However, informationfrom the investigation of these seven cases did not show a clusterbased on area of residence, work, or other routine activities.Epidemiologically, these seven cases did not appear to be relatedto the outbreak cluster and molecular subtyping by AP-PCRindicated that at least one of these seven patients was infectedwith 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 andreport visiting the block where Hospital A was located versusvisiting other blocks. Because registration of cooling towers is not required, identification of towers and other aerosol-producing devices depended on the author’s inspection, whichcould have missed potential sources of disease transmission.

Contaminated cooling towers have transmitted LD by ex-posing occupants of buildings to contaminated aerosols via openwindows4,18 and through air-intake vents and air-handling sys-tems.21,22 The absence of nosocomial cases of LD was notable.Nosocomial transmission may have been prevented by thehospital’s use of double HEPA filters on air-intake vents of the inpatient and ICU areas. The filters were not cultured, butno Lp-1 isolates were found in water from the condensate traybetween the two HEPA filters. Windows in patient areas werekept closed.

The meteorological office at the local aerodrome reported thatduring the period of the outbreak the prevailing winds weremostly from the south-west. From our vantage point on top ofthe hospital we saw that the mist from the cooling tower wascarried in different directions at different times of the day. Inaddition, Hospital A, a multi-storied structure lay to the northof both cooling towers. This is the probable reason why mostcases lived to the south and east of the hospital in tract B.

Mab subtyping of the clinical and environmental isolates sup-ported the epidemiological data linking the cases and the hos-pital cooling towers. AP-PCR added further discrimination toMab subtyping.

The cooling tower’s maintenance logs suggest that routineprocedures were followed. Intensive use of cooling towers, suchas may occur in conditions of extreme temperature and/orhumidity, may have provided more opportunity for contamina-tion of the towers by increasing the intake of make-up water.Under such conditions there may be a need for increasedbiocide over that required routinely.

Most of the .10 000 estimated annual cases of LD in the USoccur sporadically.23,24 Although contaminated cooling towersare an important cause of outbreaks of LD,1–4,18,21,25 the propor-tion of disease occurring as a result of exposure to cooling towersis unknown. Prospective epidemiological studies are needed tobetter define the attributable risk for contracting LD from coolingtowers and other disseminators of Legionella and for developingcost-effective prevention guidelines. Practices based on currentrecommendations do not ensure that a cooling tower will notserve as a source of Legionella, we recommend that maintenanceprocedures for cooling towers be reviewed and new guidelinesbe implemented, especially under conditions that lead to pro-longed or intensive use. New towers, and towers recentlystarted up after shut-down, are at increased risk for the growthof Legionella.3,25 We recommend that mechanical cleaning bedone routinely at the start-up of towers. The registration ofcooling towers and evaporative condensers would facilitate theiridentification as potential sources during outbreak investigations.

AcknowledgementsWe thank Dr Paul Edelstein, Clinical Microbiology Laboratory,Hospital of the University of Pennsylvania, who performed theurinary antigen tests on the samples; Robert Benson and JanetPruckler, Respiratory Diseases Laboratory Section, Childhoodand Respiratory Diseases Branch, National Center for InfectiousDiseases, Centers for Disease Control and Prevention, whoperformed culture of the water samples, monclonal analysis,and AP-PCR; Dr Dave Verma and the staff of the Public HealthLaboratory, Delaware Health and Social Services, who analysedthe acute- and convalescent-phase sera; Dr Edward Montz,President, Indoor Air Solutions, Inc., Pottstown, PA; and DrChin Yang, P&K Microbiology Services, Cherry Hill, NJ.

References1 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 distancetransport of Legionella pneumophila. Am J Epidemiol 1989;130:557–68.

3 Addis DG, Davis JP, Wand PJ et al. Two cases of community-acquiredLegionnaires’ disease: evidence for association with a cooling-tower. J Infect Dis 1989;159:572–75.

4 Klaucke DN, Vogt DL, LaRue D et al. Legionnaires’ disease: theepidemiology of two outbreaks in Burlington, Vermont. 1980. Am JEpidemiol 1984;119:382–91.

5 Cordes GC, Fraser DW, Skaliy P et al. Legionnaires’ disease outbreakat an Atlanta, Georgia, country club: evidence for spread from anevaporative condenser. Am J Epidemiol 1980;111:425–31.


6 Breiman RF, Cozen W, Fields BS et al. Role of air sampling ininvestigation of an outbreak of Legionnaires’ disease associated withexposure to aerosols from an evaporative condenser. J Infect Dis 1990;161:1257–61.

7 Jernigan D, Hofmann J, Cetron MS et al. and the Cruise Ship WorkingGroup. ‘Outbreak of Legionnaires’ Disease among cruise ship passen-gers exposed to a contaminated whirlpool spa’. Lancet 1996;347:494–98.

8 Hlady WG, Mullen RC, Mintz CS et al. Outbreak of Legionnairesdisease linked to a decorative fountain by molecular epidemiology.Am J Epidemiol 1993;138:555–62.

9 Mahoney FJ, Hoge CW, Farley TA et al. Communitywide outbreak ofLegionnaires’ disease associated with a grocery store mist machine. J Infect Dis 1992;165:736–39.

10 Hanrahan JP, Dale LM, Scharf VB et al. A community outbreak oflegionellosis: transmission by potable hot water. Am J Epidemiol 1987;125:639–49.

11 Dean AG, Dean JA, Coulombier D et al. Epi Info, Version 6: A Word Pro-cessing, Database, and Statistics Program for Epidemiology and Microcom-puters. Atlanta, Georgia: Centers for Disease Control and Prevention,USA, 1994.

12 SAS for Personal Computers, Release 6.08, Cary, NC: SAS Institute, 1993.

13 Barbaree JM, Fields BS, Martin WT et al. Procedures for the Recovery ofLegionella from the Environment. Centers for Disease Control andPrevention, Atlanta, Nov 1994.

14 Wilkinson HW. Hospital Laboratory Diagnosis of Legionella Infections.Centers for Disease Control and Prevention, Atlanta, Jan 1988.

15 Kohler RB, Zimmerman SE, Wilson SD. Rapid radioimmunoassaydiagnosis of Legionnaires’ disease, detection and partial character-ization of urinary antigen. Ann Intern Med 1981;94:601–05.

16 Luck PC, Helbig JH, Gunter U et al. Epidemiologic investigation bymacrorestriction analysis and by using monoclonal antibodies ofnosocomial pneumonia caused by Legionella pneumophila serogroup 1.J Clin Microbiol 1994;32:2692–97.

17 Belkum AV, Struelens M, Quint W. Typing of Legionella pnuemophilastrains by polymerase chain reaction mediated DNA fingerprinting. J Clin Microbiol 1993;31:2198–200.

18 Garbe PL, Davis BJ, Weisfeld JS et al. Nosocomial Legionnaires’disease, epidemiologic demonstration of cooling-towers as a source.JAMA 1985;254:521–24.

19 Bhopal RS, Fallon RJ, Buist EC, Black RJ, Urquhart JD. Proximity ofthe home to a cooling tower and risk of non-outbreak legionnaires’disease. Br Med J 1991;302:378–83.

20 Plouffe JF, File TM Jr, Breiman R et al. Reevaluation of the definitionof Legionnaires’ disease: use of the urinary antigen assay. CommunityBased Pneumonia Incidence Study Group. Clin Infect Dis 1995;20:1286–91.

21 Dondero TJ, Rendtorff RC, Mallison GF et al. An outbreak ofLegionnaires’ disease associated with a contaminated air-conditioningcooling-tower. N Engl J Med 1980;302:365–70.

22 Maesaki S, Kohno S, Koga H et al. An outbreak of Legionnaires’disease in a nursing home. Intern Med 1992;31:508–12.

23 Marston BJ, Plouffe JF, File TM et al. Incidence of community-acquired pneumonia requiring hospitalization: population-based activesurveillance in Ohio. Arch Intern Med 1997;157:1709–18.

24 Marston BJ, Lipman HB, Breiman RF. Surveillance for Legionnaires’disease. Risk factors for morbidity and mortality. Arch Intern Med1994;154:2417–22.

25 Bentham RH, Broadbent CR. A model for autumn outbreaks ofLegionnaires’ disease associated with cooling-towers, linked to systemoperation and size. Epidemiol Infect 1993;111:287–95.


Documents

A community outbreak of Legionnaires' disease linked to hospital