Introduction
Healthcare-associated infections are among the most frequent and costly complications in healthcare, with around nine million cases reported annually in European hospitals and long-term care facilities, resulting in 25 million additional hospital days and costs of billions of euros. 1 Environmental contamination, including surfaces and indoor air, significantly contributes to HAI transmission by serving as reservoirs for pathogens and enabling cross-contamination with patients. 2,Reference Dancer3 To improve infection control, researchers have explored more effective environmental disinfection methods. One such technique involves using ozone (O3), a high-energy, triatomic form of oxygen. Christian Friedrich Schönbein first identified ozone in 1839. Reference Rubin4 Ozone has a high oxidative potential (+2.07 V), which enables it to break down a variety of organic and inorganic molecules. This property makes it an effective bactericidal, fungicidal, and virucidal agent. It generates reactive oxygen species (ROS) that trigger lipid peroxidation, damaging cell membranes. Reference Pagès, Kleiber and Violleau5 The damage spreads to proteins and DNA, thus leading to structural changes and loss of function. Reference Rangel, Cabral and Lechuga6 In addition to its efficacy, ozone offers practical advantages. It breaks down rapidly into oxygen, leaves no harmful by-products like chlorine-based disinfectants, and reduces the need for repeated chemical purchases and storage. Reference Remondino and Valdenassi7 Healthcare settings now use ozone in several ways: for drinking and wastewater treatment, Reference Mahmoodi and Pishbin8 food sanitation, Reference Xue, Macleod and Blaxland9 air and surface disinfection—either alone Reference Moccia, De Caro and Pironti10 or with other technologies Reference Sottani, Favorido Barraza and Frigerio11 —and sterilization of medical devices, as an alternative to ethylene oxide. Reference Epelle, Macfarlane and Cusack12 Some studies have also investigated its potential for clinical therapeutic use. Reference Travagli and Iorio13
However, ozone can also harm human health. Many studies have linked it to adverse effects from both acute and chronic exposure. It can damage the respiratory system, increase epithelial permeability, and reduce lung function. It can also worsen asthma and bronchitis symptoms, and cause eye irritation, headaches, and fatigue. Reference Nuvolone, Petri and Voller14,Reference Wu, Liu, Xu, Li and Wei15 For this reason, its use requires strict safety protocols and ongoing monitoring of worker exposure.
This study aims to assess the effectiveness of gaseous ozone in reducing microbial contamination in the air and on surfaces within hospital environments. The focus is on risk areas where patients frequently interact with healthcare staff, undergo procedures, and receive visitors.
Materials and methods
Procedures and data collection
Ozone has a short half-life and cannot be stored. For this reason, it must be generated on-site at the time of use. Ozonators produce ozone by converting ambient oxygen through electrical discharges. Reference Epelle, Macfarlane and Cusack16 The Sany-Plus ozonator was used (SANITY SYSTEM ITALIA Srl, Verona, Italy), which relies on a corona discharge generator that is able to increase ozone concentration in indoor environments to 2 ppm, utilizing a preset program of 37 minutes as recommended by the manufacturer for medium-sized indoor environments. The “Safe System” in the ozonator catalyzes residual ozone at the end of each cycle, converting it back into oxygen. This feature allowed the personnel to resume the operations immediately after each treatment, thus minimizing downtime and ensuring safety for both patients and personnel. Although the device operates independently of user input, company representatives provided training to the personnel assigned to carrying out the task, specifying preset time cycles, ventilation grilles management and safety handling.
The research team handled all experimental phases except routine cleaning and device operation. All samplings were conducted according to the relevant technical standard 17 and involved inpatient wards that were identified through meetings with the hospital management and the relevant healthcare staff. We selected hospital rooms that were either vacant due to patient turnover or temporarily cleared by relocating in patients. Each patient room had its own bathroom that was included in the sampling. During ozone disinfection, the door between the room and the bathroom was kept open to assess the ozonator’s effective range. The environments included in this study were two-bed inpatient rooms and they had a surface area of less than 30 square meters.
The studied hospital environments are managed as medium risk settings. In this context, as defined in the healthcare organization’s standard operating procedures, floor and high-contact surfaces are washed on a daily basis. Washing activities consists of low-level disinfection with professional cleaning chemicals containing sodium hypochlorite (for steel surfaces, floors and sanitary facilities) and acidic cleaners (for various surfaces).
Three inpatient rooms from different wards were selected and two experimental sessions were carried out, differing in the number of sampling timepoints. With regard to each ward, the same room underwent both experimental sessions (three-timepoint and two-timepoint).
In the first session (three-timepoint, 3T), the collected samples were:
-
• 3T.T0: after routine patient activity;
-
• 3T.T1: after routine cleaning by staff;
-
• 3T.T2: after ozone disinfection.
In the second session (two-timepoint, 2T), no cleaning phase was performed. The samples regarded:
-
• 2T.T0: after routine patient activity;
-
• 2T.T1: after ozone disinfection.
By removing the “preozone cleaning” step, the aim was to more accurately detect the disinfection effect of the ozonator on microbial load. The sampling activities were always performed in conjunction with patient discharge, in an empty room, and always in the morning between 09:00 and 12:00. Therefore, the last routine cleaning in the room had taken place 18 to 22 hours earlier.
At each time point and for each session, the performed sampling were:
-
Active air sampling using the Surface Air System® (SAS, VWR International Srl, Milan, Italy) was always conducted first to minimize contamination from operators.
-
Surface sampling at ten predetermined sites by using RODAC plates (Replicate Organism Direct Agar Contact).
The SAS device collects airborne contaminants such as spores, bacteria, molds, and yeasts, by aspirating air at a constant flow rate through a perforated head. This airflow directs particles onto the surface of a RODAC plate positioned below. The SAS DUO 360 model was used, which features two sampling heads to allow simultaneous duplicate collection under identical conditions. Each head drew air at 180 L/min. During preliminary testing, we evaluated two volumes (200 L and 500 L) in similarly contaminated hospital rooms. Based on the colony counts, with a selected volume of 500 L, which consistently yielded countable colony-forming units (CFUs) at baseline and enabled clear quantitative comparisons across timepoints. After each sampling session, the SAS heads were sterilized by autoclaving and/or flaming. Raw colony counts were corrected using the manufacturer’s conversion factor. This adjustment allowed us to express airborne microbial loads as CFU per cubic meter (CFU/m3) while minimizing the risk of underestimating overlapping colonies. For each inpatient room one air sampling was performed for each sampling session (3 sessions for the 3T protocol and 2 sessions for the 2T protocol). The sampler was positioned in the center of the room, at one meter from possible obstacles or furniture, as a minimum. The instrument’s sampling flow rate was found to be adequate and consistent with the expectations established during the calibration performed by the manufacturer (calibration certificate number 090NDP24, VWR International s.r.l.).
For surface sampling, the convex agar surface of each RODAC plate was pressed directly onto the area of interest, by applying uniform pressure for at least 10 s. The results were expressed as CFU per plate (CFU/plate) based on visible viable colonies after incubation. For each room 10 possible high-touch surfaces were selected (e.g. handles, bed control-panels, faucet handles); due to this, 10 surface samples were collected for each sampling session (3 sessions for the 3T protocol and 2 sessions for the 2T protocol). For each inpatient room, the 10 surfaces selected for contamination monitoring remained unchanged across the different sessions of the 3T and 2T protocols. With some variations due to the specific characteristics of each room, the monitored surfaces included: floors, tables, handles, bedside tables, sinks, toilets, bed control panels, bed rails, and chairs.
Figure 1 shows in detail the sampling strategy for surfaces and air.
Description of the sampling sessions and identification of the sampling protocols. Details on the number of samples for each session and the sequence of activities.
Two microbiological parameters were analyzed: total bacterial count and the presence of yeasts and molds. For both, prefilled RODAC plates were used (VWR International Srl, Milan, Italy) with suitable culture media. All media, incubation times, and temperatures followed the applicable reference standards. 18,19
Specifically:
-
Plate Count Agar (PCA) to quantify bacterial load at 30°C (±1°C) over 72 hours;
-
Dichloran Glycerol Agar Base (DG-18) to detect yeasts and molds, incubating the plates at 25°C for 5–7 days.
With the aim to prevent the results from being influenced, the cleaning staff was not informed about our investigations, both in the 3T than in the 2T protocols.
Statistical analysis
Surface and air samples were analyzed separately for bacteria and for molds/yeasts. For surface samples, ten high-touch surfaces were sampled in each room. Each sampling surface was treated as a repeated observational unit across timepoints, and statistical comparisons were performed using paired CFU values obtained from the same sampling surfaces. For the three-timepoint protocol (3T), the Friedman test, a nonparametric test that compares ranks rather than actual values, was used to assess overall differences across sampling times. For pairwise comparisons (T0 vs T1 and T1 vs T2), the Wilcoxon signed-rank test, a nonparametric test based on signed ranks, with a Bonferroni correction was applied to account for multiple testing. For the two-timepoint protocol (2T), the Wilcoxon signed-rank test was utilized to compare microbial loads before and after ozone treatment. Results are presented as median values with interquartile ranges (IQRs), and P-values<.05 were considered statistically significant. All analyses were performed with STATA 18. Reference StataCorp20
Results
First sampling session (three timepoint sampling)
As shown in Table 1, the differences in bacterial and fungal contamination of surfaces were analyzed between the 3T sampling activities by testing the statistical significance of the results obtained from surface sampling with RODAC plates with PCA and DG-18. The Friedman test, applied to the overall analysis across all wards, showed a statistically significant difference between the three sampling times (P = .0020), thus indicating a reduction in bacterial load over time. In the ward-specific analysis, the effect was significant in Ward 3 (P = .0057) and in Ward 2 (P = .0478), while no statistically significant difference was observed in Ward 1 (P = .2497). For molds as well, the statistical significance of the results obtained from surface sampling with RODAC plates with DG-18 was tested. The Friedman test showed a significant overall reduction in fungal load across all wards (P = .0026). When analyzing individual wards, a statistically significant decrease was observed only in Ward 1 (P = .0332).
Differences between wards in bacterial and fungal contamination for surface and air sampling time 0, 1 and 2 (T0, T1, T2). Time 0 represents the sampling activities performed after routine patient activity, time 1 after routine cleaning by staff and time 2 after ozone disinfection. Results are expressed as medians and interquartile ranges of CFU for surface and for the total values regarding air. Results reported for air sampling regarding wards 1, 2 and 3 only refer to single count of CFU. *Friedman test
Regarding air contamination, the differences were analyzed in bacterial and fungal air contamination between the 3T sampling activities. Table 1 shows that no statistically significant reduction was observed across the three sampling times for airborne contamination by bacteria and molds.
Figure 2 shows that surface bacterial contamination markedly decreased from time 0 to time 1, followed by a further, though smaller, decline from time 1 to time 2. The distributions show progressively lower mean values across the time points of the 3T protocol, along with an overall reduction in variability. These trends refer to all wards combined. This figure also highlights that fungal surface contamination decreased between time 0 and time 1. However, values between time 1 and time 2 across all wards remained essentially unchanged, with no evidence of a further reduction.
Box plots of surface bacterial and fungal contamination across all wards (3T protocol). Contamination loads are expressed as CFU on the y-axis.
To assess differences between sampling times, we performed pairwise comparisons (T0 vs T1 and T1 vs T2) using the Wilcoxon test, both for each ward and for the overall data set. Because multiple comparisons were carried out, P-values were corrected using the Bonferroni method to reduce the risk of type I errors and increase the robustness of the results. After Bonferroni correction, the aggregated analysis across all wards showed a statistically significant decrease in microbial contamination between T0 and T1, for both bacteria (P = .0024) and molds/yeasts (P = .0060). Detailed pairwise comparisons are reported in Table 2. At the ward level, the only statistically significant comparison was the diminution in bacterial load between T0 and T1 in Ward 2 (P = .0312). No other comparisons were significant after correction. Pairwise comparisons across the three air sampling times confirmed the absence of statistically significant differences for both bacterial and fungal loads.
Pairwise comparison between timepoint samplings for bacterial and fungal surface and air contamination. Time 0 represents the sampling activities performed after routine patient activity, time 1 after routine cleaning by staff and time 2 after ozone disinfection Results are expressed as P-values. *Wilcoxon test; **Bonferroni-adjusted P-values
Second sampling session (two timepoint sampling)
In the 2T sampling protocol no cleaning activity was performed and, unlike the 3T sampling protocol, the overall analysis across all wards did not show statistically significant differences in surface bacterial load (P = .0766). Detailed results for the 2T protocol are presented in Table 3. When considering wards individually, we only observed significant variation in Ward 1 (P = .0039). Regarding molds and yeasts, the analysis of surface fungal contamination using the Wilcoxon test showed a statistically significant reduction between T0 and T1 when considering all wards together (P = .0058). Figure 3 illustrates the distribution of bacterial and fungal surface contamination across all wards in the 2T protocol. In the ward-stratified analysis, no statistically significant differences were observed. In the 2T sampling protocol, air samples did not show statistically significant differences in bacterial or fungal contamination between the two sampling points. Ward-specific analyses also did not reveal significant changes. These results confirm the pattern observed in the 3T sampling protocol, where air contamination remained stable across sampling times.
In the 2T sampling protocol, air samples did not show statistically significant differences in bacterial or fungal contamination between the two sampling points. Ward-specific analyses also did not reveal significant changes. These results confirm the pattern observed in the 3T sampling protocol, where air contamination remained stable across sampling times.
Differences between wards in bacterial and fungal contamination for surface and air sampling time 0 and 1 (T0, T1). Time 0 represents the sampling activities performed after routine patient activity and time 1 after ozone disinfection. Results are expressed as medians and interquartile ranges of CFU for surface and for the total values regarding air. Results reported for air sampling regarding wards 1, 2 and 3 only refer to single count of CFU. *Wilcoxon test
Box plot of surface bacterial and fungal contamination across all wards (2T protocol). Contamination loads are expressed as CFU on the y-axis.
Discussion
The present study evaluated the effectiveness of gaseous ozone in reducing microbial contamination in hospital environments, focusing on both air and surfaces across different inpatient wards. In the three-timepoint protocol (3T), a significant overall reduction in surface bacterial and fungal contamination was observed over time. The most pronounced decrease occurred after routine cleaning, while the additional effect of ozone was limited and not statistically significant after correction. Air contamination showed a decreasing trend but without statistically significant differences. In the two-timepoint protocol (2T), where ozone was applied without prior cleaning, no significant reduction in surface bacterial load was found overall, although fungal contamination showed a significant decrease. Ward-specific analyses revealed inconsistent effects. As in the 3T protocol, no significant changes were observed in airborne contamination.
Overall, the results indicate that routine cleaning plays a major role in reducing surface microbial contamination, while ozone contributes only marginally, particularly for fungi. No clear evidence of ozone effectiveness was found for air decontamination.
There is abundant literature on the oxidant effect of ozone on microbes. Ozone is used in reducing microbial load in matrices such as water and food, for the disinfection of medical equipment Reference Epelle, Macfarlane and Cusack16 or common objects. Reference Gupta, Taylor, Wang, Cooper and Saunte21 It is proved that, in specific experimental settings, gaseous ozone is very effective against fungi Reference Gupta, Taylor, Wang, Cooper and Saunte21 and viruses. Reference Tizaoui, Arshad, Iqbal, Farooq, Awan and Farooqi22 However, experimental studies on the effectiveness of gaseous ozone for healthcare indoor environment disinfection are lacking, both in healthcare and in other sectors. Regarding environmental decontamination with gaseous ozone, Martinelli et al. Reference Martinelli, Giovannangeli and Rotunno23 tested the efficacy of an ozone-generator system in reducing aerial microbial load. They found a modest effect in airborne bacterial load decrease at 36°C, while they found no reduction in contamination caused by psychrophile bacteria at 22°C. Their findings partially align with ours because, in our two-timepoint setting that excluded the cleaning activities, any significant airborne contamination decrease under the operating conditions described was not observed.
In a study comparable to ours, Moccia et al. Reference Moccia, De Caro and Pironti10 applied the Sany-Plus ozonator for healthcare environment decontamination. In their setting, ozone concentration reached 3 ppm levels and sampling was performed at two different times, before and after disinfection. Cleaning activities were performed by the cleaning staff and disinfection was performed with the ozonator. By applying such device, they found a relevant reduction in fungal and microbial decontamination both in the air and on the surfaces. Conversely, our findings cast doubts on the efficacy of the Sany-Plus ozonator in reducing microbial load in the air and on surfaces. With regard to surfaces, in our three timepoint sampling setting, a clear and significant diminution in bacterial and fungal contamination between time 0 and time 2 was observed. Such findings seem to fully match with the findings of Moccia et al. Reference Moccia, De Caro and Pironti10 but the pairwise comparison suggests that, at the overall level, the adjusted microbial load decrease is due to the cleaning activities, not to the use of the ozonator that was performed according to the ozonating cycle setting suggested by the manufacturer (significant load decrease for T0 vs T1, both for bacteria than for fungi). Regarding air contamination, although in some wards a decreasing trend was observed, neither the triple sampling protocol nor the double sampling protocol showed a statistically significant decrease.
Limited to air disinfection, a recent study from Bhushan et al. Reference Bhushan, Harshitha, Sumana and Prashanth24 demonstrated the gaseous ozone effectiveness in the reduction of airborne microorganisms. The authors used an ozonator in a 100 m2 conference room, where 0.5 to 5 ppm of ozone concentration were reached during the experiment. They assessed airborne bacteria contamination at 0.5, 2.5, and 5 ppm ozone concentration. Their findings highlight a significative decrease in airborne microbial load by analyzing it before and after the ozonization process. These results do not align with ours; in fact, we observed no significant decrease in air contamination. A relevant difference in the experimental setting is represented by the ozone concentrations. In our setting, the Sany-Plus ozonator can reach 2 ppm in ozone concentrations. Such information is provided by the manufacturer and confirmed by airborne ozone monitoring performed by Moccia et al. Reference Moccia, De Caro and Pironti10 during its use. Bhushan et al. Reference Bhushan, Harshitha, Sumana and Prashanth24 ozonator reached a maximum of 5 ppm instead. This could explain the difference in the microbial load diminution and in disinfection effectiveness. In fact, when used in the gaseous form, O3 concentration is a factor of major importance for the efficacy of the environmental sanitization process. Reference Travagli and Iorio13
To better explain the outcomes of our study, we highlight that in the three-timepoint protocol (3T), surface sampling showed a clear and significant drop between T0 and T1 for both bacteria and molds/yeasts. No further reduction between T1 and T2 was observed. Patterns at the ward level were not consistent, and after adjusting for multiple comparisons, the most consistent effect remained at the overall level. Air samples did not show significant changes for either type of microorganism. In the two-timepoint protocol (2T), surface sampling showed a significant overall drop for molds/yeasts, while bacterial reduction was not significant. A significant bacterial reduction was observed only in Ward 1. Air results showed no significant change.
The marked decrease from T0 to T1 in the 3T protocol, followed by stable values from T1 to T2, suggests that routine cleaning was the principal cause of the contamination reduction on surfaces. In the 2T protocol, removing the intermediate step reduced variability and made the antifungal effect clearer (despite only in an overall view), while the antibacterial effect remained limited and observed only in one ward. Ozone may provide an added benefit in certain situations, especially against molds/yeasts. However, it should be used as a complementary measure rather than a replacement for standard cleaning and disinfection. In both protocols, air sampling did not show significant changes for bacteria or molds/yeasts. This may be due to factors such as room ventilation or uneven gas distribution.
Conclusion
Overall, our results suggest that most of the reduction in both bacterial and fungal contamination on surfaces comes from cleaning practices. Limited to this aspect, the application of ozonization in the described operating conditions failed to disinfect the studied surfaces. Moving from three to two timepoints did not reveal an overall antibacterial effect on surfaces, but it made the overall antifungal effect easier to detect when we consider data from all wards. No gaseous ozone substantial effect in air contamination was observed in our real-life healthcare environment.
Routine cleaning should remain the primary strategy for reducing environmental contamination. Ozone can be considered as an extra measure, particularly where fungal control is important, but it should not replace standard cleaning practices.
Limitation
As previously stated, some phases of the study were not under the control of the researchers. Specifically, health professionals responsible for operating the instrument did not close the ventilation grilles. If left open, these grilles can cause ozone dispersion, making it difficult to achieve the concentration required for microbicidal effect. It was not possible to monitor the ozone concentration in the air, unlike other authors Reference Moccia, De Caro and Pironti10,Reference Bhushan, Harshitha, Sumana and Prashanth24 and this aspect represents a limitation. With regard to statistical power, the small sample size and the performed nonparametric tests may yield biased results and for this reason they must be considered as exploratory. Moreover, the unknown colonization status of the patients and possible other differences may also affect our results.
Acknowledgments
None.
Financial support
This research received no external funding.
Competing interests
All authors report no conflicts of interest relevant to this article.
Open access
