Impact of a pulsed xenon disinfection system on hospital onset Clostridioides difficile infections in 48 hospitals over a 5-year period

The role of the environment in transmission of hospital acquired infections (HAI) is well established [1, 2]. Manual cleaning has been shown to inadequately remove pathogens from the environment [3], and these pathogens may persist on surfaces for months [4]. This inadequacy of manual cleaning can be directly linked to an increased risk of infection acquisition for subsequent patients, with a reported increased risk of 135% for patients in previous Clostridioides difficile isolation rooms [5‐7]. Hospital-onset C. difficile infection (HO-CDI) creates a large burden of disease, causing an estimated 223,900 infections and 12,800 deaths annually [8]. HO-CDIs contribute billions of dollars in direct costs to the US healthcare system annually [9].

Colonized or infected patients are able to heavily contaminate the environment by shedding organisms onto surfaces around them. In rooms of patients with active C. difficile infection, gloved hands become just as contaminated when contacting the patient’s environment as when contacting the patient’s skin [10]. Additionally, in a multi-center study assessing the distribution of C. difficile spores in the hospital environment, 33% of rooms not currently housing C. difficile patients were still culture positive for spores [11]. The authors speculate that the source of this contamination may originate from the previous occupant of the room, or the hands of healthcare workers. A direct observation of healthcare workers found a complex web of interaction between the hands of workers, the patient, and the patient environment [12]. A study using harmless DNA markers as a surrogate for pathogens found that the DNA could be recovered from more than half of sampled surfaces within a neonatal pod after inoculation of only one surface [13]. Additionally, the DNA markers could be recovered from 18% of surfaces in adjacent pods, demonstrating that transmission may not be contained to an individual unit or pod.

Numerous interventions, applied both individually and as bundles, have been implemented in an attempt to interrupt the environmental transmission of HO-CDIs. In the last decade, many hospitals have employed enhanced disinfection with ultraviolet light to combat increasing HO-CDI rates. A multi-center randomized controlled trial assessing a low-pressure mercury vapor UV device showed no change in the HO-CDI rate when the device was utilized for terminal cleaning of C. difficile isolation rooms (relative risk 1.00; 95% CI 0.57–1.75) [14].

Pulsed xenon UV devices (PX-UV) have been shown to reduce the environmental bioburden of C. difficile in rapid 5 min disinfection cycles [15]. These short cycles allow for more rooms per day to be disinfected, covering more square footage within a healthcare facility. A meta-analysis of studies on the use of PX-UV for enhanced disinfection showed a significant reduction in the rates of HO-CDI acquisition (incidence rate ratio 0.73; 95% CI 0.57–0.94) [16]. In many of the included studies, disinfection with PX-UV was intended for all patient rooms on specific targeted units, not only isolation rooms. When reviewing the studies individually, there appeared to be a relationship between the reported reduction in HO-CDI and to the frequency of use of the PX-UV devices. To further investigate the interaction between increased PX-UV utilization and HO-CDI infection rates, this study performs a retrospective data review of facilities that have implemented PX-UV disinfection programs.

Table 1 shows the descriptions of the facilities included. On average, the facilities were around 300 beds, located primarily in the South region. Table 2 shows the changes in the patient days, discharges and transfers, PX-UV system utilization, and HO-CDI rates and count across the years included in the analysis. From 2015 to 2019, the HO-CDI rate decreased from 11.12 to 3.55 CDI per 10,000 patient days, while PX-UV utilization increased from 33,835 to 205,342 cycles annually. Fig. 1 depicts the trend over time comparing monthly average HO-CDI counts and rooms disinfected. As intervention utilization increases, HO-CDI infection counts decline.

Table 3 shows the impact of compliance to protocol on the HO-CDI rate while controlling for the facility in order to prevent any facility from having undue leverage on the result using a negative binomial model with offset for patient days. The model was statistically significant with a p-value of < 0.001. Figure 2 shows with zero compliance (no PX-UV utilization), the monthly infection rate average would be 0.98 per 10,000 patient days across all facilities. At 25% compliance, the model predicts an infection rate of 0.67 per 10,000 patient days, or approximately a 31% infection reduction. At 50% compliance, the model predicts an infection rate of 0.45 per 10,000 patient days, or approximately a 54% infection rate reduction. At 75% compliance, this model predicts an infection rate of 0.31 per 10,000 patient days, which is an infection rate reduction of 68%. Finally, at 100% compliance, the model predicts an infection rate of 0.21 per 10,000 patient days, which is a 78% infection rate reduction. Variance explained in this model is approximately 45%, indicating that other factors (e.g. hand hygiene, manual cleaning, antimicrobial stewardship, etc.) outside of compliance rate and patient days may be key components in mitigation of infection risk. This model best demonstrates the dose–response relationship between increased utilization of the PX-UV device and decreases in HO-CDI rates. When increasing the compliance with disinfection of all discharges and transfers on targeted units, the infection rates on those units decreased.

Table 4 shows the impact of the total number of rooms disinfected on the HO-CDI rate while controlling for the facility in order to prevent any facility from having undue leverage on the result using a negative binomial model with offset for patient days. The model was statistically significant with a p-value of < 0.001. The maximum predicted infection rate reduction specific to number of rooms disinfected to protocol as a predictor was 71% assuming 2,026 rooms disinfected. Variance explained in this model was 45%. Figure 3 shows that with an increase in the number of rooms disinfected to protocol, there is a decrease in the number of reported HO-CDI.

Table 5 shows the impact of the total number of disinfection cycles on the HO-CDI rate while controlling for the facility in order to prevent any facility from having undue leverage on the result using a negative binomial model with offset for patient days. The model was statistically significant with a p-value of < 0.001. The maximum predicted infection rate reduction specific to PX-UV disinfection cycles as a predictor was 72% assuming 6,370 cycles. Variance explained in this model was similar at 45%. As with the previous analysis, Fig. 4 shows that with an increase in the number of PX-UV disinfection cycles, there is a decrease in the number of reported HO-CDI.

Figure 5 shows the KNN model depicting a lack of HO-CDI cases at higher PX-UV compliance to protocol levels as well as at higher numbers of PX-UV cycles. The data points for months where HO-CDI is present are more closely aggregated in the portion of the chart where total rooms disinfected and overall compliance are the lowest. However, the data points for months where there were no HO-CDI cases are more distributed across areas of higher total usage and compliance. These clustering patterns depict an inverse relationship between presence of HO-CDI and increased utilization of the PX-UV system, and supports the findings of the negative binomial regression analyses.

The aggregate analysis of 48 hospital’s use of the PX-UV system over a 5 year period shows a strong correlation between increased utilization of PX-UV disinfection and decreases in HO-CDI rates. Negative binomial regression models assessing the number of rooms disinfected to protocol, PX-UV disinfection cycles, and the compliance to disinfection protocol showed a statistically significant inverse association with infection rates as PX-UV device use increased. This is consistent with most prior publications on the use of PX-UV systems limited to individual facilities. Instances of PX-UV programs having limited or no impact on HO-CDI rates have been reported, and the authors of these articles state that these results may be attributed to relatively low utilization of the PX-UV device or changes in infection control practices (testing methods, antimicrobial stewardship, hand hygiene etc.) The analyzed dataset is large and robust and represents the aggregate experience of numerous hospitals. The results reflect a statistically significant trend in the decrease of HO-CDI over time associated with an increase in PX-UV system compliance across multiple facilities. These findings increase the likelihood of generalizability of the results to other facilities implementing similar PX-UV programs.

Based on the initial HO-CDI infection rate of 11.12 per 10,000 patient days, 470 infections would have been projected for the 2019 calendar year. Only 150 infections were reported from the included facilities, 320 fewer than projected. A recent meta-analysis showed that the attributable cost of a HO-CDI is $34,149, with an additional attributable length of stay of 7.8 days [9]. Using these estimates, the financial value of this reduced infection risk is approximately 10.9 million USD, and almost 2,500 bed days were made available. A review of studies reporting on 30-day attributable mortality of C. difficile infection showed that rates ranged from 5.7 to 6.9% [18]. Using the more conservative figure, an estimated 18 deaths would have been expected to occur among the reduced infections.

The study does not account for other potential changes in infection control (e.g. hand hygiene practices, manual cleaning, antimicrobial stewardship, laboratory testing protocols, etc.), patient risk factors, diagnostics, and/or screening of HO-CDI. The pseudo-R² for the regression models showed that the variance in use of the PX-UV system accounted for approximately 45% of the change in infection rates. This is consistent with the epidemiological understanding of the multi-factorial transmission pathways for HO-CDI. Publications on the trends of HO-CDI rates over long periods of time report that factors such as changes in antimicrobial stewardship practices, C. difficile testing practices, and pay-for-performance programs may be important drivers of infection rates [19]. By including a large subset of data where there was no utilization of the PX-UV device, result biases from these temporal changes from other factors may be reduced. Additionally, there is potential selection bias within the customer base to those facilities that had improved or inferior outcomes. The phenomenon of publication bias, wherein results that demonstrate an outcome (either positive or negative) are disseminated whereas results showing no outcome are ignored, may be at play in this data set. Facilities that did not experience changes in their infection rates may not have been motivated to report infection data for analysis. It is important to note not all facilities within this assessment were top tier programs following best practices. There were accounts that did not adopt best practice for implementing PX-UV disinfection, and therefore, did not see similar positive outcomes. Another limitation was the number of hospitals reporting data varied by year, leading to fluctuations in the total number of data points for different time periods in the analysis. This is due to differences in the amount of data reported by each facility over the study time period, leading to differences in each facility’s input in the statistical model. Facilities that reported more HO-CDI and discharge and transfer data could be more influential in the analysis. Finally, missing data for discharges and transfers was populated using predictive mean matching. The analysis would be more robust if discharge and transfer data had been 100% available.

The findings from this study support the use of enhanced disinfection for all patient rooms in healthcare facilities. Multiple models comparing the degree of utilization of a PX-UV disinfection system found an inverse relationship between increased utilization and HO-CDI rates. Each negative binomial regression model reflected similar results whether considering PX-UV cycles, rooms disinfected, or compliance as the predictor. Furthermore, the KNN model suggests that increased utilization of a PX-UV device is a reliable indicator of presence or absence of HO-CDI. Enhanced disinfection system programs require a strategic approach where the facility is meeting consistent compliance goals, which includes completing PX-UV disinfection cycles according to best practices and meeting daily patient demand to ensure opportunities for disinfection are not missed. Enhanced disinfection systems should be considered as part of a comprehensive approach to infection control programs.

These analyses demonstrate that increased use as measured by frequency of disinfection and compliance to protocol with a PX-UV disinfection program is associated with a decrease in HO-CDI rates. This information can be used by hospitals that are evaluating enhanced disinfection technology to determine best practices.