**3. Discussion**

Public health agencies and some guidance documents recommend monitoring temperature and water quality parameters as part of *Legionella* risk assessments and water managemen<sup>t</sup> programs. A reasonable assumption in following these recommendations is that there is some relationship to *Legionella* presence or absence.

However, few studies have been conducted to evaluate these relationships and to determine if any of these surrogate measurements can substitute for *Legionella* sampling in assessing risk or effectiveness of control measures [20–23] and there is a growing interest for a more efficient sampling approach for water system *Legionella* testing [20,26]. In this study we measured *Legionella* positivity, heterotrophic plate count bacteria, and other water physicochemical parameters, including temperature, free chlorine concentration, pH, calcium, magnesium, total organic carbon, iron, zinc, lead, manganese, and copper concentrations. We also evaluated the correlation between hot water return line *Legionella* positivity or these other water quality parameters to determine if any of these relationships were predictive of *Legionella* distal site positivity.

Our analysis showed that the concordance rate between hot water return line *Legionella* positivity and distal site *Legionella* positivity was 77.8%. This was similar to a previous report of 79% concordance between the hot water recirculation loop and distal sites [20]. In the current study, we further determined the sensitivity and specificity of using hot water return line *Legionella* positivity as a screening tool to predict distal site *Legionella* positivity. The low sensitivity (55%) indicates a low probability of finding distal site *Legionella* colonization based on the hot water return line *Legionella* positivity alone. In many cases, hot water return lines that yielded no *Legionella* had positive distal sites in that system.

Studies have linked the presence of *Legionella* in building water systems to water physicochemical parameters such as trace elements concentration, pH, and temperature. A significant association between *Legionella* presence and concentrations of Mn, Zn, and Fe was reported previously [22]. In another report, iron was significantly higher (average 1.43 mg/L) in *Legionella* positive public building water samples compared to *Legionella* negative samples (average 0.09 mg/L) [27]. This association was also seen with residential water systems [28]. In the present study, statistical analysis of the correlation between *Legionella* positivity and Fe as well as Pb concentration was not possible, because 49% and 77% of total samples had Fe and Pb concentrations lower than the lower detection limit (0.06 and 0.002 mg/L, respectively). Copper concentrations of > 0.05 mg/<sup>L</sup> have been associated with a lower risk of *Legionella* colonization [29].

We observed no statistically significant correlation between distal site *Legionella* positivity and HPC, temperature, pH, free chlorine (incoming cold water and distal site), Ca, Mg, Zn, Mn, Cu, and hot water return TOC. We found HPC concentration to be a poor predictor of *Legionella* positivity. HPC concentration was only able to explain 0.68% of the variation in *Legionella* distal site positivity. These results are consistent with a previous report that also showed the lack of correlation between total bacterial counts, measured by HPC, and *Legionella* colonization [23].

*Legionella* negative hot water systems trended towards higher temperature, Ca concentration, and lower hot water return TOC concentration, however, these were not statistically significant. In contrast to one other study, we found no correlation between *Legionella* colonization and manganese in building water systems [30]. *Legionella* positive systems trended towards higher Mn concentration on average, although this relationship was also not statistically significant.

From our experience, buildings > 10 stories high often have multiple centralized hot water systems installed to serve different building zones. The complexity of these centralized hot water systems lead to favorable environments for *Legionella* colonization, such as increased water age, favorable temperatures, and a lack of disinfectant residual [2,18,27,31,32]. Previous studies have shown a correlation between the building size and *Legionella* growth. These studies have indicated that larger buildings (>10 stories) and those with centralized hot water systems are more likely to support *Legionella* growth [17,18]. We did not see a statistical relationship in our study, between building size and *Legionella* positivity, however our data sugges<sup>t</sup> the need for further investigation with larger data sets.

The risk of acquiring Legionnaires' disease has been previously associated with high levels of distal site *Legionella* positivity (>30%) [1,9,13–15]. Localized *Legionella* colonization at the point-of-use such as faucets and shower heads had been frequently observed and linked to the risk of susceptible individuals [33–35] and would serve as a patient–water system interaction point. Monitoring a facility's water system for *Legionella* colonization involves sampling distal hot water outlets regularly, which may necessitate the collection of numerous samples, especially for large facilities with multiple hot water systems [6,20].

Based on our study, sampling and culturing only the hot water return lines for *Legionella* presence demonstrates a low sensitivity of identifying *Legionella* colonization and therefore Legionnaires' disease risk. Similarly, the measured water quality parameters were not predictors of *Legionella* distal site positivity. Hot water temperature or incoming cold-water chlorine thresholds, 124 ◦F and 0.5 mg/L, respectively, also did not serve as good screening tools for *Legionella* colonization. In facilities with high risk residents, such as hospitals and long-term care facilities, a more conservative approach of direct sampling of at least 10 distal sites is recommended [9,36,37]. Based on our results, we recommend that in lower risk facilities, such as commercial or administrative buildings, sampling at least three distal sites and the hot water return should be done for routine surveillance in each hot water system. If positive samples are found, a more thorough examination of the extent and location of colonization may be warranted especially in o ffice buildings where *Legionella* has been found to persist regardless of building age [2]. In ASHRAE Standard 188, an important part of any water managemen<sup>t</sup> program is to ensure that there is validation of the program's e fficacy. This is to ensure the water managemen<sup>t</sup> program is controlling identified hazardous conditions, specifically the risk of *Legionella* growth and spread. Our results demonstrate that these surrogate measurements cannot be used to validate the control of *Legionella* risk at a facility because they are not predictive of the presence or absence of *Legionella* species.

#### **4. Materials and Methods**

#### *4.1. Sample Collection and Onsite Water Quality Parameter Measurements*

Bulk water samples were collected from 28 buildings in New York City (25 buildings), San Francisco (two buildings), and New Jersey (one building) from March to September 2015. The sampled buildings included commercial buildings ranging from 5 to 57 floors. Samples were collected from the incoming cold-water per building, cold-water storage tank per cold water system, three hot water distal outlets per hot water system (near, mid and far), and the hot water return line from each hot water system.

Cold water and hot water return line samples were collected after a 1-min flush. A 1 L sample with sodium thiosulfate for microbiological analyses and a 250 mL sample preserved with nitric acid for metal analyses (iron (Fe), copper (Cu), lead (Pb), zinc (Zn), calcium (Ca), magnesium (Mg), and manganese (Mn)) were collected. Hot water distal outlet samples were collected and treated as above, but the 1 L microbiological sample was taken prior to flushing. Additionally, for hot water return line samples, two 50 mL vials with hydrochloric acid preservative were collected for total organic carbon (TOC) testing. Measurements for temperature, pH, and free chlorine were conducted onsite after sample collection.

Temperature, pH, and free chlorine residual concentration were measured on-site at the time of sample collection using a digital thermometer, portable Hach 900 colorimeter, and Oakton Acorn pH meter following the manufacturer's protocols. Samples for microbiological analyses were shipped on ice overnight to Special Pathogens Laboratory (Pittsburgh, PA, USA) and samples for metal analyses and total organic carbon testing were shipped to ALS Environmental (Middletown, PA, USA).
