From May through June 2002, a total of 146 water samples were collected from private homes of six towns (Milan, Modena, Bologna, Rome, Naples, Bari) representative of different Italian regions (Northern, Central, and Southern Italy). A similar number of samples were taken from each town; selection was made on the basis of the water distribution systems inside the town and building and heater types in each area. After we identified each building, we asked a random family (in case of a condominium) to participate in the study, i.e., to complete our questionnaire and give informed consensus for water collection. Laboratory examinations were free, and at the end of the study each participating family received a letter with results of Legionella analysis.

Hot water samples were drawn from the bathroom outlets (shower heads or bathroom tap) in three sterile 1-L glass bottles after a brief flow time (to eliminate cold water inside the tap or flexible shower pipe). To neutralize residual free chlorine, sodium thiosulphate was added in sterile bottles for bacteriologic analysis, whereas acid-preserved glass bottles were used for chemical determinations. Collection bottles were returned to the laboratory immediately after sampling for bacteriologic and chemical-physical examination; if analyses would not begin within 24 hours, samples were kept at >4°C and processed within 48 hours of collection.

To detect Legionella spp., 2-L water samples were concentrated by membrane filtration (0.2-μm-pore–sized polyamide filter, Millipore, Billerica, Massachusetts, USA). The filter membrane was resuspended in 10 mL of original sample water and vortex-mixed for 10 min. To reduce contamination by other microorganisms, 5 mL of this suspension was heat-treated (50°C for 30 min in a water bath) (19). Two aliquots of 0.1 mL of the original and concentrated specimens (heat-treated and untreated, 1:10 diluted and undiluted) were each spread on duplicate plates of modified Wadowsky-Yee selective medium (Oxoid Ltd., Basingstoke, Hampshire, UK). The plates were incubated at 36°C in a humidified environment with at least 2.5% CO₂ for 10 days and read from day 5 at the dissecting microscope. Suspected colonies with a mottled surface or an iridescent and faceted cut-glass appearance, were counted from each sampling. All colonies from plates with <10 and 10–20 random colonies were subcultured on buffered charcoal yeast extract (BCYE) agar (with cysteine) and charcoal yeast extract agar (cysteine-free) media (Oxoid) for >2 days. Only colonies grown on BCYE were subsequently identified by an agglutination test (Legionella Latex Test, Oxoid). The test allows a separate identification of L. pneumophila serogroup 1 and serogroups 2–14 and detection of seven Legionella (polyvalent) species (other than L. pneumophila), which have been implicated in human disease: L. longbeachae, L. bozemanii 1 and 2, L. dumoffii, L. gormanii, L. jordanis, L. micdadei, L. anisa. Legionella-like bacteria serologically nonidentifiable (in total 4 colonies) are excluded from the account, awaiting a different confirmation by DNA sequence (polymerase chain reaction [PCR]-method). The results are expressed as CFU/L and the detection limit of the procedure was 25 CFU/L (mean value of two plates). All the research units participated in a quality control for Legionella detection in water that was organized by the National Health Institute, through a periodic distribution of water samples added with unknown Legionella species and concentration. The total microbial counts at 36°C and 22°C were obtained twice by the pour-plate method on plate count agar (Oxoid). The plates were incubated at 36°C for 48 h or at 22°C for 72 h.

To isolate Pseudomonas spp., 100-mL and 10-mL water samples were filtered through a 0.45-μm-pore–size membrane (Millipore). If the number of bacteria was high, suitable dilutions were made. The membranes were placed on Pseudomonas cetrimide fucidin cephalosporin (CFC) agar (Oxoid) and incubated at 30°C for 48 h. Each type of oxidase-positive colony was counted.

Water temperature and residual free chlorine (DPD method, colorimetric) were determined at the time of sample collection. Standard techniques were used to measure oxidizability and water hardness. Concentrations of calcium, magnesium, iron, manganese, copper, and zinc were measured by flame atomic absorption spectrophotometer (Perkin-Elmer, Wellesley, MA, mod 5000) on acidified samples (1% HNO₃) concentrated by boiling.

A detailed standardized questionnaire was developed to evaluate risk factors possibly associated with colonization. The first part collected information on family characteristics (number of components, age and sex, length of stay in residence) and on pneumonia events during their stay in the home. The second part was devoted to home data: type (flat, single house, villa), flats in the building, home floor, home rooms and bathrooms, building age, type of water supply, and disinfection systems used. The third part collected information on the heating system (central or independent, electric or gas heater), distance of the sample site from the water distribution point, existence of a tank and its volume, age of the system, service frequency, and existence and characteristics of a softening and water recycling systems. Water operating temperature (temperature at the distribution site) was also recorded.

All statistical calculations were made with SPSS/pc (SPSS Inc, Chicago, IL). Logarithmic transformations were used in statistical analyses to normalize the nonnormal distributions, and results are presented as geometric means. The bacteriologic data were converted into log₁₀ (x+1). When possible, variables were categorized into dichotomous ones. The results were analyzed by correlation analysis, t test, one-way analysis of variance (ANOVA), and by chi-square test. Odds ratios (OR) and 95% confidence intervals (CI) were calculated to assess categorical risk variables associated with microbial contamination. Variables that were significant in the univariate analysis were entered in a multiple logistic regression model. By using conditional logistic regression models, independent predictors of colonization were established. Variables were retained in the model if the likelihood ratio test was significant (p < 0.05).

Table 1 shows general characteristics of the examined water in terms of supply and distribution systems. Five heating systems were recognized, corresponding to those more frequently used at the domestic level, although with geographic differences, as the centralized systems were mainly adopted in northern Italy.

Table 2 shows chemical and microbiologic qualities of hot water samples. When samples were grouped according to their origin (mixture or groundwater), groundwater had significantly higher levels of calcium (105.7 ± 32.1 mg/L vs. 68.8 ± 31.3 mg/L, p < 0.001) and magnesium (21.5 ± 16.0 mg/L vs. 14.9 ± 6.0 mg/L, p < 0.01) and was harder (35.4 ± 13.0 vs. 22.1 ± 8.8°F, p < 0.001). Pseudomonas spp. were isolated from 56 of 146 (38.4%) samples, with levels ranging from 1 to 6.4 x 10⁴ CFU/100 mL; 85.7% of positive samples contained fewer than 10³ CFU/100 mL.

A total of 33 (22.6%) samples of 146 were contaminated by Legionella spp., and L. pneumophila (Table 3) was the most frequently isolated species (75.8% of isolates). In the positive samples, the mean number of legionellae was 1.17 x 10³ CFU/L (range 25 to 8.7 x 10⁴ CFU/L); three samples (9.2%) contained >10⁴ CFU/L, none of which were L. pneumophila serogroup 1. Although we examined colonies with different morphologic traits, the agglutination test did not reveal multiple species or serotypes in a single water sample.

The risk for microbial contamination according to the system characteristics was evaluated by applying a univariate logistic regression (Table 4). A central warm water system and distance of the water from the heating point >10 m were strongly associated with the risk for Legionella contamination (OR 9.24 and 8.10, respectively, p < 0.001). Other significant and positive associations were observed with tank volume, plant age, flooring in the home, and total number of flats in the building. Pseudomonas contamination was positively associated with heating system age and negatively with tank distance and water operating temperature. Pseudomonas was also associated with the particular water source, with 65.3% of groundwater colonized versus 12.3% of mixture water (OR 13.44; 95% CI 5.02 to 36.03, p < 0.001).

The univariate regression was then applied to study the association between microbiologic data and water chemical parameters (Table 5). Seven factors were independently protective against Legionella colonization: high levels of copper, hardness, oxidizability, and free chlorine and low concentrations of zinc, iron, and manganese. Lower levels of zinc and manganese were also associated with lower total count at 36°C and 22°C. Pseudomonas was positively associated with total hardness and iron <20 μg/L, whereas residual free chlorine significantly inhibited Pseudomonas. When water samples were grouped according to their trace element levels, water samples (n = 12) characterized by concentrations of zinc <100 μg/L, iron <20 μg/L, and copper >50 μg/L were all negative for Legionella. No other system or water parameters were associated with bacterial contamination of the examined samples.

The data were reanalyzed by means of multivariate conditional logistic regression models. A central heating system, distance from heating point >10 m, and a system >10 years old were each independently associated with higher risk of Legionella colonization, whereas water with levels of copper >50 μg/L and zinc <100 μg/L were predictive of no contamination (Table 6). For Pseudomonas, only groundwater remained highly predictive of colonization (OR 12.69; 95% CI 2.66 to 44.00, p < 0.001), whereas an operating temperature >50°C was predictive of noncontaminated samples (OR 0.25; 95% CI 0.11 to 0.98, p < 0.05).

The percentage distribution of Legionella species differed significantly according to the heater system (chi-square = 14.00, p < 0.05). Electric heaters were legionellae-free; the gas-heated independent systems had little contamination (10.0% of those with tank and 16.4% of those without) and were mainly colonized by either L. pneumophila serogroup 1 or non-pneumophila Legionella species. In the centralized heating systems of both single buildings and neighborhoods, Legionella colonization was higher (52.8% and 66.7%, respectively), and L. pneumophila serogroups 2–14 were the most frequently isolated serotypes. Germ concentration did not differ according to the heater type.

Table 7 shows that water temperature, level of free chlorine, and Pseudomonas contamination differed according to Legionella species (Table 7). Water samples contaminated by L. pneumophila serogroup 1 were characterized by significantly lower operating temperature compared to that of the other groups, whereas water samples positive for L. pneumophila serogroups 2–14 had lower residual chlorine and higher Pseudomonas count.

The reported frequency of pneumonia symptoms was double among persons living in the legionellae-positive homes compared to those living in legionellae-free buildings (8 cases in 95 residents vs. 15 cases of 333 residents), but the difference was not significant (OR 1.95; 95% CI 0.80 to 4.75). Results did not change by correcting for the duration of residence of each person in the examined house.