The San Francisco Public Utilities Commission provides an average of 250 million gallons (950 million liters) of water per day to ≈2.4 million residents in northern California, including 750,000 in the city and county of San Francisco. Surface water makes up >99% of the water supply. Chlorine was added to kill microorganisms present in source water (primary disinfection) throughout the study period. Chlorine concentrations are monitored at several locations throughout the distribution system. Chlorine used for residual (or secondary) disinfection was replaced with monochloramine on February 2, 2004.

Buildings with >3 stories in San Francisco were identified from lists of commercial customers of the San Francisco Public Utilities Commission and real property owned by the city and county of San Francisco. Building managers and owners gave permission for sample collection inside the buildings for the duration of the study. Results of Legionella cultures were provided only at the completion of the study. Standardized questionnaires were administered to building engineers and facilities managers to obtain information on the age of the building, capacity of water heaters and hot water storage tanks, type of water heating system (boiler, heat exchanger, or instantaneous heaters), and type of pipe material used throughout most of the building. At the completion of the study, building engineers were surveyed about routine maintenance plans for the potable hot water system, standard procedures for flushing outlets after a disruption of water service, and knowledge of industry guidelines for controlling Legionella growth in building water systems (16).

Samples from each building were collected 6 times during the 2-year period, 3 times before and 3 times after conversion to monochloramine disinfection. Each round of sampling lasted 8–10 weeks. Preconversion and postconversion rounds of sampling were conducted at corresponding seasonal intervals.

Nine samples were collected from each building during sampling rounds, including a 1-L water sample from a water heater or heat exchanger, four 1-L samples of hot water, and 4 swabs of biofilm at point-of-use outlets (faucets or shower heads). Water samples were collected in sterile, 1-L plastic bottles (Nalge Nunc International, Rochester, NY, USA) containing 0.5 mL 0.1 N sodium thiosulfate solution to neutralize free chlorine and chloramines. Water heater samples were drawn from the drain valve, pressure relief valve, or from the closest outlet to heat exchangers. Point-of-use outlets were selected at farthest points from water heaters when possible. Biofilm samples were collected from shower outlets and faucets by inserting a sterile, polyester-tipped applicator swab (Falcon, Becton Dickinson and Company, Sparks, MD, USA) and rotating it firmly against the interior surface. Biofilm swabs were placed in sterile, screw-capped test tubes containing 0.1 mL sodium thiosulfate solution in 5 mL of water from the same site. Hot water taps were run until the temperature reached a maximum for collection of water samples. The same locations were sampled in each round. When sampling could not be performed at the selected site, the nearest substitute site was sampled; however, only samples collected from the same site before and after monochloramine conversion were included in analyses.

Water temperature, pH, and free (disassociated) and total chlorine concentrations were measured in a separate sample bottle. Total chlorine includes free chlorine plus monochloramine. Temperature was measured with a hand-held thermometer. pH was measured with a digital meter (pHep 3, Hanna Instruments, Leighton Buzzard, UK). Free and total chlorine residuals were measured by using the N, N-diethyl-p-phenylenediamine method with a colorimeter and test kit (Model DR/890, Hach Chemical Co., Loveland, CO, USA). Building engineers were asked about any interruptions in water service affecting the building or specific sites in the 3 months preceding the sampling date.

All culturing for Legionella species and amebae was performed in the Legionella Laboratory at the Centers for Disease Control and Prevention in Atlanta, Georgia, following standard procedures (17). Legionella organisms were speciated or serogrouped by macroscopic slide agglutination with a panel of polyclonal rabbit antisera against Legionella species and L. pneumophila serogroups (18). Laboratorians were blinded to the identity of buildings from which samples were obtained, and buildings were assigned different identification numbers in each round. Samples were transported at ambient temperature and processed a mean (± standard deviation) of 3 (±2) days after collection. Water samples from point-of-use outlets were concentrated 100-fold by filtration through a 0.2-μm polycarbonate filter (Nucleopore, Pleasanton, CA, USA). Biofilm swab samples were placed on a lawn of Escherichia coli for detection of ameba and treated with diluted acid (0.1 mol/L KCl, 0.005 mol/L HCl) to reduce the number of non-Legionella bacteria before plating.

Concentrations of Legionella spp. in water samples are expressed as CFU/mL based on plate counts of Legionella colonies grown from a known volume of original sample. Concentrations determined by this method are approximate. The upper and lower limits of detection were 0.05 and 25 CFU/mL for point-of-use outlets and 10 and 5,000 CFU/mL for water heaters. Plate counts were not determined for samples overgrown with non-Legionella organisms.

Active, laboratory-based surveillance for culture-confirmed Legionella infections in San Francisco residents was conducted from January 1, 2003, through December 31, 2004, through the Active Bacterial Core surveillance activity of the California Emerging Infections Program (19). We reviewed legionellosis case report forms from the national passive surveillance system for cases among San Francisco residents or persons with a history of travel to San Francisco during the incubation period. Surveys were sent to infection control departments at all San Francisco hospitals to identify cases of probable or confirmed Legionnaires' disease during 2003 and 2004. Information was solicited from hospitals about environmental testing for Legionella spp. in water systems, and measures taken to reduce microbial contamination of water systems during 2003 and 2004.

Data were entered into Access version 2002 (Microsoft, Redmond, WA, USA) and analyzed by using SAS for Windows version 9.0 (SAS Institute, Cary, NC, USA). We conducted building- and site-specific analyses of the prevalence of Legionella colonization. A building was considered colonized at a timepoint if Legionella spp. were cultured from any site. We considered a point-of-use outlet colonized if Legionella spp. were cultured from either a water sample or biofilm swab. Wilcoxon rank sum test was used to analyze differences in the proportions of positive sites or concentrations of Legionella. We also calculated adjusted prevalence ratios (PRs) and 95% confidence intervals (CIs) or p values by using PROC GENMOD (SAS Institute) for the clustered nature of sites within buildings. Preconversion and postconversion sampling rounds were considered repeated measures. Multivariable models investigated associations between Legionella colonization and water measurements or building characteristics.

The conversion to monochloramine provided higher concentrations of total chlorine (which includes both free chlorine and monochloramine) and lower concentrations of trihalomethane compounds, the principal disinfection byproducts in treated water entering the distribution system (Table 1). The conversion to monochloramine also resulted in an ≈10-fold increase in total chlorine concentrations measured in building hot water systems. Average temperature and pH measured in building water samples did not change significantly.

Prospective Legionella testing was performed in 53 buildings, including 24 public and 29 commercial buildings. When chlorine was the residual disinfectant in municipal drinking water, Legionella spp. were cultured from building water systems on 96 (60%) of 159 occasions, and 37 (70%) of 53 buildings were colonized with Legionella spp. in >1 of the 3 sampling rounds (Table 2). After conversion to monochloramine, Legionella spp. were found on 7 (4%) of 159 occasions in 5 (9%) of 53 buildings. These 5 buildings had been colonized at multiple sites before disinfection with monochloramine. Conversion to monochloramine resulted in a 93% reduction in the prevalence of Legionella colonization in building water systems (PR 0.07, 95% CI 0.03–0.16). Colonized water systems were no more likely than Legionella-free systems to include hot water storage tanks or material other than copper for hot water plumbing, although sample size limited building-level analyses. Legionella spp. were recovered from 12 (60%) of 20 buildings for which engineers reported maintaining water systems according to standard practices, such as maintaining backflow prevention and flushing outlets after interruption of water service, versus 18 (75%) of 24 buildings for which no standard maintenance of water systems was reported (p = 0.28).

A total of 364 (13%) of 2,822 water and biofilm samples yielded Legionella spp: 352 (25%) of 1,405 samples collected before conversion and 12 (<1%) of 1,417 samples collected after conversion to monochloramine. Five Legionella species and 7 serogroups of L. pneumophila were identified (Figure). L. pneumophila serogroup 1 accounted for >60% of all Legionella organisms. The same species of Legionella and serogroups of L. pneumophila were repeatedly cultured from individual sites (Figure).

Legionella spp. were cultured from 46 (15%) of 316 water samples from building water heaters: 45 (29%) of 157 samples collected before conversion versus 1 (<1%) of 159 after conversion to monochloramine (p<0.001). When we controlled for water heater temperature, building height, and interruptions in water service, monochloramine use decreased the prevalence of Legionella colonization in water heaters by 96% (Table 3). Colonization of water heaters was more prevalent in buildings with >10 stories and in which water service had been interrupted in the past 3 months. Water temperatures >50°C were associated with the lowest prevalence of colonization, and Legionella spp. were not detected when the temperature exceeded 60°C (140°F). Over the 2-year study period, temperatures of water in building water heaters were >50°C at 88 (28%) of 318 sampling timepoints. In buildings in which engineers reported familiarity with industry guidelines for controlling Legionella growth in water systems, water heater temperatures were >50°C on 26 (21%) of 125 occasions versus 46 (32%) of 144 occasions in buildings in which engineers were not familiar with industry guidelines (p = 0.04).

At point-of-use outlets, Legionella spp. were cultured from 247 (20%) of 1,252 water samples and 70 (6%) of 1,254 biofilm swab samples. Combining culture results from the water samples and biofilm swabs from each site, Legionella spp. were cultured from 246 (39%) of 624 paired samples before conversion to monochloramine versus 9 (1%) of 622 paired samples after conversion (p<0.001). Median concentrations of Legionella spp. at colonized outlets were significantly lower after conversion to monochloramine (Table 2). Legionella were cultured from both the water and biofilm samples on 59 (24%) of 246 occasions before conversion versus 2 (22%) of 9 occasions after conversion. The same Legionella species and serogroup was cultured from biofilm swabs and water samples on 56 (92%) of 61 occasions when both were positive. When we controlled for Legionella spp. in the sampled water heater, water temperature at the point of use, building height, and interruptions in water service, monochloramine use decreased the prevalence of Legionella colonization at point-of-use outlets by 96% (Table 4). Legionella colonization at point-of-use outlets was independently associated with Legionella spp. in the sampled water heater, building height, and interruptions in water service. Legionella spp. were cultured only from point-of-use outlets and not from water heater samples on 72 (61%) of 118 occasions when building water systems were colonized, including 6 (86%) of 7 occasions after monochloramine conversion.

Amebae at sampled sites were associated with Legionella spp. colonization only when chlorine was used for residual disinfection. Legionella spp. were cultured from 61 (36%) of 169 samples in which amebae were present versus 291 (24%) of 1,236 samples without amebae (p = 0.01). After conversion to monochloramine, Legionella were found in 1 (1%) of 78 samples containing amebae and 8 (1%) of 866 samples without amebae (p = 0.75). During disinfection with chlorine, Legionella concentration was higher in samples containing amebae (median 9.0 CFU/mL, range 0.1–25.0) compared with those without amebae (median 1.5 CFU/mL, range 0.05–25.0, p<0.001). The prevalence of amebae decreased from 169 (12%) of 1,405 samples when chlorine was the residual disinfectant to 78 (8%) of 944 samples collected in the first 2 rounds after conversion to monochloramine (p = 0.006). Results of ameba cultures from the final round of sampling were discarded after amebae were found in negative control water samples.

Active, population-based surveillance for Legionella infections identified 1 confirmed case of Legionnaires' disease in a San Francisco resident in November 2004, who traveled to Mexico within 2–10 days of symptom onset. Infection control departments in 7 (70%) of 10 hospitals in San Francisco, including the 3 largest hospitals, reported no hospitalized patients meeting the definition of a probable or confirmed case of Legionnaires' disease (20) during the study period. Review of case report forms from the national passive surveillance system did not identify any cases of Legionnaires' disease in persons with history of travel to San Francisco during the incubation period of their illness.

No environmental testing for Legionella spp. in hospital water systems was conducted during 2003 and 2004 in the San Francisco hospitals that responded to the survey. Two hospitals added supplemental chlorine to their water system to prevent microbial contamination before monochloramine conversion; supplemental chlorination was discontinued after monochloramine was added to municipal drinking water.