Mitigating Legionella spp. risk in an Australian healthcare facility using on-site electrochemical water disinfection Open Access

In developed settings, microbial contamination of potable water distribution with biofilm-associated environmental pathogens, such as Legionella pneumophila, nontuberculous mycobacteria, and Pseudomonas aeruginosa, is a persistent challenge and public health threat (Cassini et al. 2019; Collier et al. 2021). These waterborne pathogens are particularly problematic in hospitals and aged care facilities, where complex pipe networks and vulnerable populations increase the risk of infections, particularly Legionnaires' disease, characterised by severe pneumonia, and Pontiac fever, a milder form of legionellosis (NAS 2020). Legionnaires' disease is extremely serious in vulnerable patients, such as the immunocompromised, elderly, and newborn, with mortality rates as high as 40% in healthcare settings (WHO 2022), and vulnerability to legionellosis is generally increasing due to factors such as aging demographics, climate change and travel (LeChevallier et al. 2024).

Legionella spp. was first identified after a pneumonia outbreak in Philadelphia in 1976 and has since caused significant outbreaks, although over 75% of cases are thought to be sporadic (NAS 2020). The first Australian case was described in 1978, and significant outbreaks soon followed, including 108 cases in South Australia from 1979 to 1988, 110 cases at the Melbourne Aquarium in 2000 and persistence (from 2011 to 2013) in the plumbing of a Brisbane-based hospital (Collier et al. 2021). Recently, Sydney reported seven hospitalisations, Queensland saw a spike with 88 cases, more than triple the average for this time of year, and Melbourne had at least 114 cases as of August 2024, 10-times the normal rate for that time of year (Health-Vic 2024).

A combination of plumbing construction and operational and behavioural factors, including those driven by energy conservation and patient scald risk management, contribute to Legionella spp. exposures in healthcare settings (Heida et al. 2021). Particular attention is needed to address building extensions and renovations that may result in lengths of pipe being cut-off or capped, creating ‘dead legs’ in pipe networks where reduced or no water flow is favourable for biofilm growth and Legionella spp. colonisation. The commissioning of thermal mixers to decrease scalding risks may create persistent areas of warm water (28–42 °C) suited to Legionella spp. growth (NAS 2020). Additionally, water softeners, intended to improve water quality, may inadvertently remove chlorine or other disinfection residuals from treated water, allowing opportunistic pathogen growth (Oliveira et al. 2024).

The highly variable usage of facilities adds further complications. For instance, ensuite showers may remain unused for extended periods if patients are bedridden, leading to no or low flow conditions that can significantly impact water quality (Jamshidi et al. 2020). Pipe biofilms increase legionellosis risk by providing food for free-living protozoa that host Legionella spp., which can lead to explosive growth cycles and release cells to the flowing water (Ashbolt 2023). The propensity for problematic biofilm formation depends on many factors, such as the type of pipe material (influence of the substratum), water flow conditions and rate (no flow/slow flow/high flow), water temperature (low growth <20 °C; fast growth 28–42 °C; no growth >50 °C), and the content of chlorine or other disinfection residuals in the water (Mancini et al. 2015; Busch et al. 2025).

To address the risk of waterborne infections in the elderly and immunocompromised, site-specific water quality management plans are increasingly implemented in hospitals and aged care facilities. Guidelines emphasize the importance of operational controls, such as the commissioning of continuous disinfection systems and regular flushing of outlets, to limit human opportunistic pathogen counts and other microbiological risks (Australian Government 2015). Contamination of water sources with high concentrations of Legionella spp. can occur when changes in water flow or pressure disrupt biofilms, releasing large amounts of bacteria into the surrounding water. Biofilms, therefore, significantly contribute to microbiological risks, and minimising their development is vital for maintaining water quality (Zhao et al. 2023).

Various treatments are used to control the growth of Legionella spp. in engineered environments, including physical, chemical, and thermal methods. However, these methods alone are not fully effective, and regrowth and recolonization often occur after a lag period (Sciuto et al. 2021). Ongoing effort is required. Hyperchlorination is effective for planktonic Legionella spp. but less effective for sessile communities, and regular exposure to chlorine can lead to higher tolerance and promote disinfection resistance (Assaidi et al. 2020). Chlorination resistance is influenced by community diversity and interspecies relationships, necessitating an examination of disinfectant decay and efficacy.

An emerging alternative to on-site chlorination, which is not easy to control, is to use electrolysed oxidising water (EOW), which is an established technology in the food sector. In short, EOW is generated by passing an electric current through a dilute salt solution. To avoid corrosion problems that may occur when using acidic EOW, most advanced devices produce a solution with neutral pH, sometimes referred to as neutral electrolysed oxidising water or as neutral anolyte (a disinfectant solution generated at the anode compartment during electrolysis and containing active chlorine mainly in the form of HOCl) (Iram et al. 2021).

The benefits of anolyte as a potable water infection control strategy for healthcare facilities include the capability for continuous dosing to existing chlorinated drinking water supplies. Anolyte may offer several advantages: such as enabling energy savings by allowing for lower hot water heating temperatures (e.g., 50 °C vs. 60 °C), reducing infrastructure stress, eliminating the need for hyperchlorination, minimising patient disruption, reducing contamination in ice machines and toilets, and enhancing patient safety by lowering the risk of waterborne nosocomial infections. Ultraviolet (UV) disinfection (alone) is not effective as it does not provide any residual effect (Iram et al. 2021).

In hospital and clinical settings, the recommended first-line therapy for Legionnaires' disease in infected patients is empirical pharmacological treatment using a fluoroquinolone (levofloxacin or moxifloxacin) or a macrolide (azithromycin preferred). However, there is limited information available regarding adverse events, complications, the duration of antibiotic therapy, and its association with clinical outcomes. Therefore, prevention and infection control are crucial for managing Legionella spp. infections (EU-OSHA 2021).

There is limited data on the long-term efficacy of anolyte on-premise plumbing disinfection and biofilm control (Ecas4 Australia 2019), and a lack of studies evaluating anolyte systems against other disinfection technologies used in healthcare settings. Such comparisons are necessary to establish the relative advantages and limitations and to ensure the system's overall safety and efficacy under field conditions.

Another challenge in controlling Legionella spp. in engineered water systems is the uncertainty tied to standard detection techniques (Whiley & Taylor 2016; Nisar et al. 2023). Current methods include culture-based approaches that identify the subset of strains that can grow under the culture conditions, quantitative PCR (qPCR) assays for estimating bacterial genomic load of both live and dead bacteria, and flow cytometric analysis utilized for characterising non-culturable (VBNC) populations in environmental samples.

This study evaluated the effectiveness of anolyte for on-site water disinfection over a 145-day period at a hospital and aged care facility in Adelaide. The site had previously experienced persistent Legionella spp. contamination in both its hot and cold-water distribution systems, despite the application of conventional control measures. Microbiological water quality was monitored throughout the study using both culture-based methods and qPCR to track Legionella levels. The primary objective was to determine whether this disinfection technology could provide a dependable long-term strategy for maintaining water safety in healthcare environments.

The study site was a community-owned, 60-bed private hospital established in the early 1970s with an adjoining 84-resident aged care facility. The facility was organised over three stories with the hot water system comprising four circuits and two risers. Pipes were copper throughout the building up to the connection points to mixers, where braided stainless-steel pipes with a crosslinked polyethylene (PEX) inner tube were used. Average annual water consumption was ∼8,000 m³. Building works have resulted in significant structural changes and additions, with increased complexity of the drinking water premise plumbing system. Water quality issues were first identified three years prior to the commissioning of the Ecas4 water disinfection system, with elevated Legionella spp. colony counts and heterotrophic colony counts (HCCs) detected during routine testing in the aged care facility. Rigorous monitoring and management followed. The hot water was continually heated to 80 °C, reaching the thermostatic mixing valves prior to point of use at approximately 70 °C, and cleaning staff were instructed to conduct daily flushing of taps and shower heads. Despite the high energy costs, this intervention proved insufficient as Legionella spp. and HCC counts above potable water quality guideline limits continued to be periodically recorded as the placement of thermostatic mixing valves resulted in sections of the pipe prior to the point of delivery frequently holding warm water suitable for Legionella spp. growth. A chemical disinfection residual booster was thus installed on-site. On commencement of this study, a plumbing investigation revealed the presence of corrosion and biofilms in the pipe network and revealed that hot- and cold-water mixing takes place several metres before outlets, creating a warm water zone that is likely suitable for Legionella spp. growth. It also means the system cannot effectively be heat treated, as part of the pipe network can never reach the required temperature.

The Ecas4 system generates a low amount of anolyte, a diluted, slightly saline, and pH-neutral solution that contains HOCl (ECAS4 Australia 2019). The synthesis of anolyte is carried out through a patented reactor equipped with four chambers (two anodic and two cathodic, separated by ion-exchange membranes); the anodes are coated with a proprietary catalytic mixture of ceramic metal oxides. A minimal dose of anolyte is injected into the cold-water distribution systems to provide continuous disinfection without altering the potability of the building's drinking water supply. The system installed at the study site produces up to 80 L/h of anolyte (with a concentration of 300–350 mg/L of free chlorine). Measurements of free chlorine and total chlorine (by adding a couple of drops of 0.5% KI solution to the sample) were made at the dosing and sampling points using Hanna Instrument devices (either HI96771 or HI711); all measurement were undertaken on-site. The anolyte was dosed into the cold-water system to achieve a free chlorine concentration between 0.9 and 1.5 mg/L, using a metering pump connected to a flow meter. Dosing was based on flow rate; specifically, the dosing pump was activated following the arrival of a pulse sent by the flow meter. The Elster H4000 flow meter, equipped with an Elster PR7 inductive pulse unit, sent a pulse every time it measured 10 L of water. The dosing pump was an Ecolab EMP III, capable of dosing approximately 7.5 mL per stroke. To obtain a 0.5% dose (∼1.5 ppm), approximately 50 mL of the 300 mg/L solution was added to 10 L of water, with each pulse triggering six injections of 7.5 mL each. To increase sensitivity, the piston stroke (which provided the injected volume) was reduced to 50%, so as to require 12 injections of ∼4 mL each per pulse received.

The chlorine concentration in the system was checked manually at sampling time. The concentration fluctuated for several reasons: the injected solution was diluted in a volume between 10 and 19 L of water (when the 20th L was reached, a new pulse was sent to the pump); the anolyte solution had a nominal concentration of 300 mg/L, but in reality it varied depending upon the temperature of the water (which changes the yield of the electrochemical reaction producing the active agent) and the age of the electrodes (which deteriorate with use); finally, the chlorine concentration measured downstream also depends on the time spent in the pipes and the presence of biofilms. We did not evaluate chlorine decay because the goal was to measure between 0.9 and 1.5 mg/L active chlorine at the point of use, not at the point of injection.

The water disinfection system was initially set up to dose both the cold water and hot water supplies in the hot water system. However, the high temperature to which the water was heated (80 °C) caused a rapid decrease in free chlorine and the dosing of the hot water was discontinued. However, as the cold water feeding the hot water system was still dosed with anolyte (like all the cold water entering the main building), the temperature of the hot water circuits could be lowered from 80 to 62 °C and a free chlorine residual of 0.2–0.5 mg L⁻¹ was maintained.

Samples were analysed for relevant physicochemical parameters such as temperature, pH, redox potential, electrical conductivity, total organic carbon (TOC), and free and total. The water temperature and pH were respectively recorded using a temperature probe and an Ag/AgCl single-junction electrode connected to a portable digital pH meter equipped with automatic temperature compensation (Ecoscan pH6, EUTECH Instruments, USA). The electrical conductivity (EC) and the oxidation-reduction potential (Eh) were measured using a K = 1.0 stainless-steel cell and a platinum electrode, respectively, connected to an Ecoscan Con6 and Ecoscan Ion6, both featuring automatic temperature compensation (EUTECH Instruments, USA). These measurements were taken on account of the strong correlation between Eh and pH values and chlorine speciation, and hence the relevance of these measures to water treatment optimisation and disinfection efficacy (James et al. 2004; Copeland & Lytle 2014). Free chlorine and total chlorine concentrations were determined on-site using ready-made reagents in a calibrated photometer (HI 96734, HANNA Instruments, USA) according to the standard DPD (diethyl-p-phenylene diamine) colorimetric method (USEPA method 330.5; Standard Methods for the Examination of Water and Wastewater 4500-Cl G). Dosing adjustments were made according to the Australian Drinking Water Guidelines (NHMRC 2023), which recommend maintaining free chlorine levels between 0.2 and 0.5 mg/L at tap outlets, to ensure adequate residuals within the chlorinated drinking water distribution network.

Microbiological water quality was assessed using multiple methods. Culturable bacterial (HCCs) and Legionella spp. colony counts (L. pneumophila serogroups, and other Legionella species) were undertaken by the National Association of Testing Authorities (NATA)-accredited laboratory used by the hospital for regular monitoring (using the Australian Standard methods AS/ANZ 4276.3.2:2003 and AS/NZS 3896, respectively). Diagnostic and verification monitoring using DNA-based quantitative polymerase chain reaction (qPCR) analysis was also undertaken to provide estimations of the total microbial load and Legionella spp. as described below.

Water samples (approximately 500 mL) were vacuum filtered through a 47 mm diameter, sterile plain cellulose nitrate membrane, 0.1 μm pore size (Whatman, GE Healthcare Life Science, Parramatta, NSW, Australia) immediately on arrival at the laboratory. Filtering was stopped once it slowed significantly, and the volume of each filtered sample was recorded for subsequent calculations. Genomic DNA was isolated from each filter (DNeasy PowerWater™ DNA isolation) according to the manufacturer's specifications (Qiagen, NSW, Australia). The extracts were quantified using a Quant-iT™ HS ds-DNA assay kit in a Qubit™ fluorometer (Invitrogen, Carlsbad, CA), and quality checked by 1.5% agarose gel electrophoresis. For the analyses of the biofilms at selected points in the distribution system, the withdrawn sections of the copper and flexi-pipes were cut open longitudinally. Biofilm samples were collected with sterile swabs of an area of a few cm² on the internal surface of the sampled pipes. The biofilm was then extracted from the swabs in phosphate buffer saline [pH 7.4, (1X)] by triplicate cycles of sonication and vortexing. The resulting suspensions were further centrifuged for 5 min at 14,000 g and processed for genomic quantifications similar to the water samples immediately after removing the swabs and part of the supernatants.

Genomic DNA was extracted as described above. Real-time qPCR was performed using the QX200™ ddPCR™ EvaGreen Supermix (Biorad, Australia) on the Bio-Rad CFX96™ Real-Time system (Bio-Rad Laboratories, Hercules, CA, USA) to quantify total 16S rRNA gene presence, all samples were sequenced following amplification using the primers 331F/518R (Denman & McSweeney 2006), with Legionella spp. estimated as described previously (Jian et al. 2020) and a 55 °C annealing temperature. Legionella spp. quantification was performed with the iQ-Check^® Quanti Legionella spp. Real-Time PCR Quantification Kit (Bio-Rad, Australia) according to the manufacturer's specifications, except for the PCR reaction volume, which was reduced from 50 to 20 μL. Standards were made from full-length PCR amplified E. coli ATCC11775 DNA extracted with 100 μL of lysozyme (500 μg/mL) for 5 min at room temperature. The PCR product was purified from deoxynucleotide triphosphate (dNTP), primer dimers, polymerases, unused primers and PCR mix buffer using the Agencourt^® AMPure^® XP kit (Beckman Coulter, Australia). After purifying the PCR product, the numbers of 16S rRNA SSU copies were calculated by assuming average molecular masses of 340 Da for one nucleotide of single-stranded DNA following the equation: copies per nanogram = (n × MW)/(NL × 10), where n is the length of the standard in base pairs or nucleotides, MW is the molecular weight per bp or nucleotide, and NL is the Avogadro constant (6.022 × 10²³ molecules per mol). The 16S rRNA copy number in the samples was determined from the standard curves and subsequently standardized to copy numbers per mL of water. In all runs, standard curves and the amplification efficiency were calculated using the Bio-Rad CFX96™ Real-Time software (Bio-Rad Laboratories, Hercules, CA, USA). All standards were assessed in at least two different runs to confirm the reproducibility of the quantification.

Microbial diversity analysis was performed using amplicon sequencing on: (i) 16S rRNA gene amplicons, resulting in a total of 2,565,040 high-quality sequence read-pairs for the bacterial dataset, with 15,088 read-pairs (±507 standard error – SE) generated per sample; (ii) 18S rRNA gene amplicons, resulting in a total of 125,808 high-quality sequence read-pairs, with 740 read-pairs (±69 SE) per sample. High-quality reads were considered to be those where: (i) maximum expected errors based on Phred Q values did not exceed 2 per read of each pair; (ii) the minimum size was 150 bp for quality trimmed reads after trimming from the error-prone read-end at a first incidence of minimum Phred Q of 10; (iii) read-pairs merged towards the reconstruction of the amplicon of origin without mismatches, ensuring second strand verification; (iv) putative chimeric amplicon sequences were removed; and (v) off-target amplicon sequences were removed (e.g., unclassified, mitochondrial/chloroplast sequences were removed from the bacterial dataset and non-protist sequences were removed from the eukaryotic dataset). The 515f/806r primer set was used for amplifying the prokaryotic 16S rRNA gene and the 1931F/EukBr for amplifying the eukaryotic 18S rRNA gene (Gilbert et al. 2018). The PCR products were prepared in two reactions: a 28-cycle reaction for amplifying the target sequences (using primers for the target microbial group) and an 8-cycle reaction to index the amplified sequences, with the same primers containing the indexing and linker 5′ overhangs (‘Primers_indexes.xls’ file including eukaryotic primers) generated with the Barcrawl v.100310 software. The selected two-step strategy helps reduce index-associated biases during amplification. The 20 μL first step reactions contained 2 μL of DNA extract template normalized to between 5 and 10 ng/μL concentrations, Phusion Flash High-Fidelity PCR Master Mix (Finnzymes/Thermo Fisher Scientific, Inc., Waltham, MA), 0.5 μM of each primer, 0.4 μg/μL UltraPure^TM BSA (Life Technologies/ThermoFisher Scientific, Inc., Waltham, MA, USA) to mitigate PCR reaction inhibition, and made up to 20 μL final volume with ultrapure PCR grade water. The second step reactions included the same reagents except for the BSA, using 2 μL of the first step PCR products as templates. The cycling conditions were: 3 min at 95 °C for polymerase activation and template denaturation; 28 cycles for the first step reactions and eight cycles for the second step reactions of 30 s at 95 °C for denaturation, 30 s at 50 and 57 °C for the bacterial and the eukaryotic primer annealing, respectively, and 30 s at 72 °C for elongation; 10 min at 72 °C for a final elongation. DNA template and PCR products were quantified with the Qubit^® dsDNA HS fluorometric assays (Life Technologies/ThermoFisher Scientific, Inc., Waltham, MA, USA). The indexed PCR products were equimolarly multiplexed, and the PCR product pool was purified using the Agencourt AMPure XP solid phase reversible immobilization method (Beckman Coulter, Brea, CA, USA), and sequenced using a HiSeq2500 instrument (Illumina, San Diego, CA, USA) in rapid mode giving 2 × 250 bp sequence reads at the Brigham Young University DNA Sequencing Center (Provo, UT, USA).

Cluster, base-calling, and Illumina standards quality control were performed with the Illumina HiSeq control software v.2.2.58, and the real-time analysis v.1.18.54 combined with the bcl2fastq2 v.2.17.1.14 Illumina proprietary software (Illumina, San Diego, CA, USA). Demultiplexing of sequences to their original samples was performed with the Flexbar v.2.5 software allowing no mismatches with the index-linker-primer sequence constructs. Sequence quality control, amplicon reconstruction, amplicon sequence variant calling, chimeric sequence screening, and sequence classification were performed with dada2 v1.22.0 (Callahan et al. 2016) using the SILVA v.138 reference database (Yilmaz et al. 2014) for the prokaryotic 16S rRNA gene annotation and the PR2 v.5.0.0 database for the 18S rRNA gene annotation (Vaulot et al. 2022), in the R v.4.1.3 programming and statistical environment.

To obtain the absolute abundance of individual prokaryotic taxa, we multiplied the relative abundances with the absolute 16S rRNA gene counts as previously suggested (Jian et al. 2020).

Post-commissioning monitoring revealed mean values (and standard deviations) for key physicochemical parameters as follows: pH 7.1 (±0.4), redox potential (Eh) 277 mV (±86), EC 652 μS/cm (±222), and TOC 13.6 mg/L (±4.1). Most parameters were stable over time; however, TOC increased by 19.5% over the 8 days following the commissioning of the water disinfection system. Total and free chlorine levels were monitored at selected sampling points to optimise dosing (Supplementary Table S1).

L. pneumophila serogroups 1–14 were not detected by colony counts before or after the commissioning of the water disinfection system. However, other Legionella species were identified at multiple sampling points during pre-commissioning testing and within the first 8 days post-commissioning (Supplementary Table S2). By the third post-commissioning sampling event (day 22), following additional tap water flushing implemented by hospital staff during the initial 8 days to facilitate the removal of released biofilm and enhance the mixing of anolyte solution, Legionella spp. colony counts indicated a consistent improvement in water quality throughout the system, with no samples testing positive for Legionella spp. thereafter. In contrast, HCCs displayed significant variability and a tendency to increase over the monitoring period (Supplementary Table S3).

Despite achieving a significant reduction in Legionella spp. concentrations throughout the network and consistently low mean cell counts within 2–3 weeks of continuous treatment, several rooms exhibited a resurgence in Legionella spp. concentrations during subsequent sampling events (Figure 2).

Among all pairwise comparisons, the Legionella GU counts showed strong associations with the total 16S rRNA gene counts (Figure 4(c)), whereas a very weak and insignificant association was observed between HCC and total 16S rRNA gene counts (Figure 4(d)).

Analysis of pipe surfaces sampled prior to the initiation of the disinfection treatment showed a substantial presence of Legionella spp. within the biofilms. High levels of Legionella spp. were present in both hot and cold flexi-pipes feeding sinks (Table 1). In contrast, the section of the copper pipe main that was sampled showed no detectable Legionella spp. While Legionella spp. was still detected in the flexi-pipes nearly 3 weeks after the commissioning of the Ecas4 system, the levels were low. Notably, no Legionella was detected in the pipes that had been replaced just before the commissioning of the system, indicating limited biofilm growth during that period.

Third, other microbial groups, such as Burkolderiales (i.e., TRA3-20, Burkholderiales), Commamonadaceae, and Xanthobacteracea, showed increased relative abundance post-installation (Figure 5), although this was more evident for particular rooms and sampling times. Subsequent calculations indicated a decreasing absolute abundance of these taxa post-system installation (Figure 6). The analysis also showed that 3 weeks after starting the anolyte disinfection regime, microbial 16S rRNA gene counts in the water sampled from most rooms had dropped by 2–3 orders of magnitude (Figure 6). Specifically, examination of taxonomic groups, particularly Legionellaceae, demonstrated a significant decrease in Legionella load in the potable water subsequent to deploying the Ecas4 system.

On day 0, Stramenopiles (flagellated protozoans) accounted for 25% of the Eukaryota reads, primarily represented by oomycetes and Blastocystis. The remaining 75% were classified as Opisthokonta, a major eukaryotic supergroup that includes choanoflagellates, sponges, and fungi. Between days 1 and 8 post-disinfections, the proportion of Stramenopiles decreased by approximately 5%, with a corresponding increase in Archaeplastida. Opisthokonta completely disappeared and was replaced by Amoebozoa, which included 18% Tubulinea. Notably, 100% of the Tubulinea reads were attributed to Vermamoebidae. Following the initiation of anolyte treatment, Amoebozoa peaked at 75% of the Eukaryota community between Days 1 and 8, before declining to 25% between Days 22 and 137.

Maintenance of a disinfectant residual in premise plumbing systems, supported by a flushing programme that prevents water stagnation and ensures the disinfectant reaches distal parts of the system, is recognised as an important component of legionellosis management (Grimard-Conea et al. 2024). However, the relative performance of free chlorine versus monochloramine or other on-site disinfectants is dependent on the specific plumbing materials in use, the nature of corrosion products and other factors (Triantafyllidou et al. 2016). In this work, chlorine residual was introduced and maintained by an on-site electrochemical chlorine system (Ecas4) along with a routine flushing programme, which significantly improved the hospital building water quality by reducing Legionella spp. occurrence within 22 days post-commissioning. Monitoring before and after system commissioning showed this to be a robust solution for maintaining high standards of water safety in hot and cold-water services in the healthcare facility although some further challenging effects of the pipe biofilms on water quality were also observed. We observed a transient increase in TOC immediately post-commissioning, potentially due to biofilm dislodgement within the distribution pipes. This increase was associated with initially higher HCCs, underscoring the substantial effect of biofilms on microbial community dynamics and water quality maintenance (Thi Nguyen et al. 2024). These findings align with previous research on chlorine disinfection in warm water plumbing systems in the USA, which showed that sequential chlorine dosing (0, 0.1, 0.25, and 0.5 mg/L over 8 weeks) resulted in elevated TOC levels (>1.0 mg/L) (Martin et al. 2020).

We used a DNA-based qPCR technique for increased sensitivity and decision support. Consistent with previous work (Lee et al. 2011), our data showed that qPCR monitoring methods were more sensitive than standard culture-based methods as Legionella spp. detectable at low cell numbers by qPCR were frequently reported as negatives via the standard plating method. In practice, this means that a gradual increase in microbial burden, e.g., due to decreased water flow when a patient is absent or on bed rest and not using the shower/basin or other hydraulic events, can be detected much earlier, allowing for timely management.

The qPCR method also includes data relating to viable but not culturable (VBNC) Legionella spp. These VBNC L. pneumophila are missed by standard plate culture approaches (enHealth 2022) but are important as they can persist as dormant cells for long periods in biofilms, particularly within amoebal cysts or trophozoites at ambient temperatures (Shaheen et al. 2019), and lab research shows that they can be resuscitated by amoeba co-culture (Dey et al. 2019). However, as the development of unculturable cell states caused by copper pipe corrosion products (and other stressors) is complex (Song et al. 2024), and qPCR is unable to distinguish between DNA from VBNC cells and free nucleotides released by dead cells, further investigation is necessary to explore this complex relationship in detail, e.g., through the use of a modified qPCR method using propidium monoazide (Taylor et al. 2014) or integration of vital staining and flow cytometry to accurately identify and quantify VBNC cells (Nisar et al. 2023).

Future investigations could also incorporate RNA-based screening methods to differentiate between active and dormant taxa. This approach, along with metagenomic analysis, can provide a clearer understanding of microbial dynamics and their implications for water quality management.

Although the current study was initiated due to the detection of Legionella spp. in the hospital plumbing system, our initial assessments post-commissioning of the water disinfection system using culture-based methods revealed no presence of L. pneumophila serogroup 1 or serogroups 2–14. Given the caveat of VBNC legionellae possibly present, and confirmation of no L. pneumophila by qPCR, this suggests either effective control measures or their absence during the sampling periods. Nonetheless, sporadic identification of other Legionella species was observed both pre-commissioning and within the first 8 days post-commissioning of the water disinfection system, highlighting the dynamic nature of Legionella spp. colonisation and biofilm-release in water distribution systems (Bavari et al. 2021). Similarly, some microorganisms may persist due to their resistance to oxidative disinfection or protection within biofilm structures, potentially acting as reservoirs for opportunistic pathogens. Biofilms in water distribution systems act as barriers to disinfectants and can shelter VBNC cells, which go undetected using standard monitoring techniques. Under certain conditions – such as water stagnation, temperature shifts, or reduced disinfectant levels – these biofilms can enable the re-establishment and growth of opportunistic pathogens. Thus, while the Ecas4 system effectively lowers planktonic bacterial and Legionella levels, sustained surveillance and system maintenance are crucial to prevent biofilm-driven recontamination and health risks over time. Some bacteria may live inside the biofilm/free-living protozoa attached to the surfaces of the pipes; over time, this biofilm may be removed by the continuous disinfection treatment, thus allowing a rebound of planktonic cells but diminished habitat for premise plumbing pathogens.

Overall, the absence of L. pneumophila serogroup 1, a primary pathogenic serogroup associated with Legionnaires' disease (NAS 2020), provides encouraging results regarding the effectiveness of the water disinfection system tested. As challenges remain in fully eradicating Legionella spp. from complex network environments (Sciuto et al. 2021), key goals in its management should centre on limiting biofilm growth, plumbing dead-ends and water stagnation, facilitated by an ongoing flushing programme.

Our study investigated microbial diversity over a 145-day period using 16S rRNA gene amplicons to analyse microbial community dynamics within the hospital's water distribution network. We observed significant variability in microbial composition across different rooms and sampling times. Notably, there was a decrease in the relative proportion of unclassified species, indicating a shift towards a higher prevalence of known microbial species. Previous research has consistently shown that Proteobacteria, particularly Alpha- and Beta-proteobacteria, dominate drinking water bacterial communities, regardless of system's origin or type of disinfectant residual (Santos et al. 2016).

Maintaining a disinfectant residual affects the relative abundance of key drinking water bacterial genera, such as Legionella and Pseudomonas spp. Routine sequencing of negative controls is essential to improve data quality in drinking water studies, though standardizing sample processing protocols remains challenging (Nisar et al. 2023). Our analysis revealed that, despite a 2–3 log reduction in overall microbial load, genera such as Sphingomonas did not decrease at the same rate. Instead, they consistently increased in relative abundance across all monitored rooms, possibly indicating their ability to adapt to the persistent, nutrient-poor, chlorine residual conditions typical of drinking water environments (Sorouri et al. 2023).

In addition, we observed diverse patterns in the relative abundance of other genera across different rooms. For example, the genus Ralstonia showed room-specific fluctuations, likely influenced by localized factors such as pipe biofilm composition, temperature variations, or water flow patterns (Lowe-Power et al. 2018).

These findings highlight the dynamic nature of microbial communities within different rooms, shaped by complex interactions between environmental variables and microbial dynamics. Continued research is crucial to deepen our understanding of these dynamics and their implications for water quality management across various environmental contexts. Advances in techniques for isolating and sequencing DNA from mixed microbial communities, such as metagenomics, have provided valuable insights into microbial diversity and evolution (Zhang et al. 2021).

This study offers valuable insights into the intricate interactions between microbial communities and water quality in a healthcare hot and cold drinking water system and the positive impact of an on-site chlorine disinfection system. It demonstrated a significant reduction in the risk of exposure to Legionella spp., with biofilm reduction and overall enhanced water quality management across the site's hot and cold plumbing systems. Nonetheless, it underscored the necessity for further research into protozoan diversity and their interactions with pathogens like Legionella spp. to further develop effective waterborne disease mitigation strategies. Ongoing investigation into microbial diversity, biofilm dynamics, and the various pathogen–protist interactions is essential for refining these strategies. Future studies employing advanced RNA-based screening methods will provide deeper insights into microbial activity and viability, ultimately leading to improved microbial water quality management practices in critical healthcare environments.

This paper summarizes the findings of a project supported by Ecas4 Australia and the South Australia Government through the Office of International Coordination. Assistance from Scott Williams, Daniel Walker, Sharon Piro, Simon Crabb and Daniel Vallelonga is gratefully acknowledged.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.