Discussion

This report has demonstrated the detection of a sustained outbreak of C. difficile ST2 over a six-month period on a single ward using prospective MLST-based surveillance. This outbreak did not generate a clear rise above baseline variability in CDI incidence, so would almost certainly have gone unnoticed (or increased in size to become detectable) if surveillance was based on overall CDI incidence alone. The use of early warning algorithms based on incidence and epidemiological data have been proposed, however have been shown to perform poorly for detecting in-hospital transmission events of C. difficile. ^{Reference García-Fernández, Frentrup and Steglich4,Reference Edman-Wallér, Toepfer and Karp11,Reference Edman-Wallér and Werner12} In our approach we were able to combine basic low-discrimination genomic analysis with high-quality epidemiological data to detect the outbreak and initiate remedial IPC actions when there were only six suspected cases, and the outbreak was halted when it had reached eight cases.

The use of genomics is well-described for investigating hospital outbreaks of C. difficile, however most reported investigations were performed retrospectively, either over a defined time period, or instituted once an outbreak had reached sufficient size to generate a clear rise in CDI incidence. ^{Reference García-Fernández, Frentrup and Steglich4,Reference Berger, Gfrörer and Becker5,Reference Wen, Shen and Xia15,Reference Newcomer, Fishbein and Zhang25,Reference Miles-Jay, Snitkin and Lin26} Prospective PCR-ribotype-based or whole-genome sequencing surveillance is commonly performed at the reference laboratory level, however this is typically intended to monitor the wider genomic epidemiology of C. difficile, rather than to detect individual hospital outbreaks. ^{Reference Krutova, Kinross and Barbut13,Reference Persson, Nielsen and Coia14} Here we describe the potential value of a prospective genomic surveillance system, which can be implemented at the level of a medium-sized front-line clinical microbiology laboratory. This system was implemented at relatively low-cost and additional staffing requirement and without the need for specialist bioinformaticians to analyze data. ^{Reference Bloomfield, Hutton, Velasco, Burton, Benton and Storey16} Furthermore, we are not aware of prior reports where surveillance has been performed using solely ONT sequencing, which represents a potentially more accessible sequencing option for smaller hospital laboratories such as ours, due to low capital costs and laboratory space requirements. However, it should be noted that the requirement for a cultured isolate could represent a barrier for other laboratories wishing to adopt genomic surveillance for C. difficile. Most clinical laboratories use culture-independent methods for CDI diagnosis, ^{Reference Carroll and Mizusawa27} meaning culture is not part of the standard diagnostic pathway. Although our laboratory had experience with C. difficile culture prior to implementing genomic surveillance, we believe the approach of reflex culture on antigen- or PCR-positive stool samples would be relatively straightforward to implement from a workflow perspective for many diagnostic laboratories. We employ a simple 48-hour direct-plating culture method, without other more complicated steps such as alcohol/heat shock, ^{Reference Bloomfield, Hutton, Velasco, Burton, Benton and Storey16} and the anaerobic environment is created using disposable gas packs and small containers, which requires minimal capital expenditure compared to a dedicated anaerobic incubator.

A limitation of our approach is the relatively nondiscriminatory nature of MLST, which cannot be used to infer individual patient transmission events. ^{Reference Kamboj, McMillen and Syed28} This was demonstrated in the retrospective phylogenetic analysis, where two of the ST2 cases did not cluster with the other isolates. However, when combined with epidemiological data, the additional level of detail afforded by MLST allowed increases in incidence at the ST-level to be appreciated at a stage where increases in incidence at the organism-level were inapparent. Our data showed a high degree of genomic/ST variability within C. difficile detected in our hospital, which is consistent with prior reports. ^{Reference Eyre, Cule and Wilson3} This is why we were able to detect this outbreak at a relatively early stage using an MLST-based approach; there was a clear change at the ward-level from the expected variability in the STs detected. In jurisdictions where hospital CDI is dominated by few STs (if these exist), or where electronic capture of bed movements is not available, an MLST approach like this may be less effective.

ONT sequencing technology is constantly developing. Most sequence data for this investigation was generated using older chemistry and flow cells (R9.4.1) than are currently available. While a lower raw-read accuracy was achieved than with other platforms such as Illumina, ^{Reference Delahaye and Nicolas29} it was sufficient to show clustering of the isolates within the wider hospital phylogeny (Figure 3). However, we were not able to generate precise SNV distances. ONT accuracy has now been shown to perform well for higher-resolution typing, such as core-genome MLST using the newer Q20+ chemistry and duplex basecalling. ^{Reference Wagner, Dabernig-Heinz and Lipp30} Since the outbreak reported here, we have moved our surveillance program to Q20+ chemistry and are planning to move to a higher-resolution typing method, which we anticipate will enable outbreaks or patient-to-patient transmission events to be detected even earlier, and will enable early exclusion of unrelated cases with the same MLST.

Around one-third of CDI cases could not be typed, largely due to negative culture (Table 1). This is similar to another sequencing study using the same culture media, where 62% of cases could be typed. ^{Reference García-Fernández, Frentrup and Steglich4} This highlights a difficulty intrinsic to C. difficile genomic surveillance, and as a result, additional outbreak cases or transmission chains may have gone undetected. Furthermore, our surveillance did not test asymptomatic patients for C. difficile carriage, who could also contribute to transmission. ^{Reference Furuya-Kanamori, Marquess and Yakob31} These limitations may have delayed recognition of the outbreak or other transmission chains on the ward or within the hospital. Prior studies examining direct plating culture methods using chromID^TM C. difficile agar have found recovery rates above 90%, so our culture recovery rate of 75.4% can likely be improved on. ^{Reference Carson, Boseiwaqa and Thean32,Reference Yang, Nam and Kim33}

The early detection of this outbreak allowed our IPC team to implement a range of interventions which was followed by a decrease in ST2-associated cases. Our institution has low rates of CDI by international comparisons, however this outbreak occurred on our hemato-oncology ward, where vulnerable patients, broad-spectrum antibiotic use, diarrhea, and long/repeated admissions are common. ^{Reference Loo, Bourgault and Poirier34} CDI poses unique challenges in IPC. Asymptomatic colonization is common, and antimicrobial exposure can increase environmental shedding of C. difficile spores. ^{Reference Furuya-Kanamori, Marquess and Yakob31,Reference Kundrapu, Sunkesula, Tomas and Donskey35} Patients who develop CDI will shed spores for several weeks after their diarrhea resolves and spores can survive in the environment for months-years. ^{Reference Sethi, Al-Nassir, Nerandzic, Bobulsky and Donskey36,Reference Dancer37} Transmission has been shown to be higher in patients where prior room occupants had CDI despite room cleaning at discharge. ^{Reference Sood, Truelove and Dougherty10} Additionally, CDI risk is also higher where the prior room occupant had broad-spectrum antibiotics independent of CDI status, ^{Reference Freedberg, Salmasian, Cohen, Abrams and Larson38,Reference Root, Lindstrom, Xie, Mangino, Moffatt-Bruce and Hebert39} highlighting the contribution to environmental contamination from asymptomatically colonized patients. In this outbreak transmission events were not clearly linked to particular rooms, which suggested more widespread ward contamination and that transmission from staff or shared equipment were contributing factors.

Chemical disinfection with sporicidal agents is effective in controlling CDI in endemic and outbreak settings. ^{Reference Carling, Parry and Olmstead1} In an evaluation of four products, in both wipe and spray format, aHP showed the greatest reduction in viable spores and cleaning efficiency on surfaces for endemic and hypervirulent C. difficile strains. ^{Reference Kenters, Huijskens, De Wit, Sanders, van Rosmalen, Kuijper and Voss40} Implementation of aHP wipes for routine and equipment cleaning in addition to preexisting hypochlorite-based terminal cleaning was likely an important intervention in our setting. The aHP wipes were reported to be convenient and easy to use by staff performing cleaning tasks, particularly as they enabled a one-step process and were nontoxic to use. Similarly, the use of gloves for all room entries for patients with diarrhea (irrespective of cause) may have reduced C. difficile transmission more than would occur with improved hand hygiene alone. ^{Reference Kociolek, Gerding and Carrico41,Reference Gerding, Muto and Owens42} Given there were multiple components to the intervention it is not possible to determine which were more or less effective in this specific context. Furthermore, due to the low number of cases involved, we also cannot exclude a natural decline due to stochastic effects, however the sharp change in incidence after intervention supports practice improvements as the cause.

In this study, we have demonstrated the potential value of simplified prospective genomic surveillance for CDI using ONT sequencing, which detected an outbreak of CDI on a high-risk ward, which would have otherwise gone undetected. Whether these results can be applied to other settings and whether surveillance may be beneficial in lower-risk settings than a hemato-oncology ward remains to be determined, however, the additional level of detail afforded by sequencing allowed recognition of the outbreak, which in turn permitted effective intervention.