Survey of diagnostic and typing capacity for Clostridium difficile infection in Europe, 2011 and 2014

SM van Dorp 1 , DW Notermans 2 , J Alblas 2 , P Gastmeier 3 , S Mentula 4 , E Nagy 5 , P Spigaglia 6 , K Ivanova 7 , F Fitzpatrick 8 , F Barbut 9 , T Morris 10 , MH Wilcox 11 , P Kinross 12 , C Suetens 12 , EJ Kuijper 1 , for the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project on behalf of all participants 13 1. Department of Medical Microbiology, Leiden University Medical Centre, Leiden, the Netherlands 2. National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands 3. Institut für Hygiene und Umweltmedizin, Charité Universitätsmedizin Berlin, Berlin, Germany 4. National Institute for Health and Welfare (THL), Helsinki, Finland 5. Institute of Clinical Microbiology, University of Szeged, Szeged, Hungary 6. Istituto Superiore di Sanità, Rome, Italy 7. National Centre of Infectious and Parasitic Diseases, Sofia, Bulgaria 8. Health Protection Surveillance Centre and Beaumont Hospital, Dublin, Ireland 9. National Reference Laboratory for C. difficile, Paris, France 10. Public Health Wales, UK Anaerobe Reference Laboratory, Cardiff, United Kingdom 11. Leeds Teaching Hospitals NHS Trust, Leeds, United Kingdom 12. European Centre for Disease Prevention and Control, Stockholm, Sweden 13. Other members of the ECDIS-Net project are listed at the end of the article


Introduction
Since 2003, Europe has been affected by outbreaks of Clostridium difficile infection (CDI) associated with the emergence of PCR ribotype 027/NAP1 [1].A decade later, C. difficile was the microorganism responsible for 48% of healthcare-associated gastrointestinal infections in acute care hospitals across Europe [2].Despite being frequent, CDI remains underestimated in most European countries [3].Underdiagnosis mainly results from a lack of awareness among medical doctors of when to suspect that patients may have CDI and use of suboptimal diagnostic algorithms at local microbiological laboratories [3][4][5].Reference tests, i.e. toxigenic culture and cell culture cytotoxicity assay (CCA), are not suitable for routine application due to their complexity and long turnaround time [6,7].Rapid enzyme immunoassays (EIAs) to detect C. difficile toxins in faeces lack sensitivity [6,8].Highly sensitive tests such as EIA detecting glutamate dehydrogenase (GDH) -a C. difficile-specific enzyme [9] -or nucleic acid amplification tests (NAATs) have insufficient specificity [6,10].To overcome underdiagnosis and suboptimal performance of stand-alone tests, the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) has recommended since 2009 testing loose stools using two-step algorithms that have a highly sensitive test as the first screening step and a highly specific test as the second confirmatory test [6,11].The 'Bristol stool scores' [12] are commonly used to categorise stool consistencies and can be used to select samples for

Number of countries
a Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lichtenstein, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, Serbia, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey and the United Kingdom (data were analysed separately for England, Northern Ireland, Scotland and Wales, but counted as one country).No data were available for Iceland.
b Seven countries reported other changes in national laboratory diagnostics: Slovenia was developing new national guidelines for CDI at the time of the second survey; Romania started a national surveillance study in 2014; Spain published an opinion document on CDI [32]; Slovakia was in the process of implementing new diagnostic methods due to an increased interest in CDI; in Cyprus, the central diagnostic laboratory for C. difficile implemented a two-step diagnostic algorithm; in Finland, CDI diagnostics were subcontracted to laboratory consortia that applied nucleic acid amplification tests more often; and Hungary relocated its national reference laboratory to expand its laboratory capacity but still had limited resources.CDI testing.ESCMID recommended performing CDI testing not only upon request of a medical doctor, but also based on other indications such as the 'three-day rule', i.e. diarrhoea after three days of hospitalisation or when diarrhoea develops after antibiotic use [6,13].
The type of diagnostic algorithm applied influences not only clinical care [14], but also CDI surveillance's sensitivity and specificity [3,14,15].However, a consensus on when and how to test for CDI has not been established among reference and local laboratories.
Additionally, typing of C. difficile to understand its local or wider transmission remains non-standardised in Europe [16,17].Numerous typing methods have become available for routine use in the last 30 years.For C. difficile, these include methods that use restriction enzymes (e.g.restriction endonuclease analysis, pulsed-field gel electrophoresis (PFGE)), PCR amplification of housekeeping genes (e.g.multilocus sequence typing (MLST)), of repetitive elements (repetitive-element PCR, multilocus variable-number tandem repeat analysis (MLVA)), of the pathogenicity locus (e.g.toxinotyping) or of 16S-23S rRNA intergenic spacer regions (e.g.PCR ribotyping) [16,18].Whole genome sequencing, with its ultimate discriminatory power, can already be used for in-depth analysis of evolutionary patterns [19].Nevertheless, PCR ribotyping still remains the standard typing method in Europe as it involves relatively simple technology and its low costs permits widespread application [16,18]. In

Study design
The the survey, respectively (in 2011, Serbia did not participate in ECDIS-Net).All surveys are available online [21].

Selection
There was no European register of microbiology laboratories to use for random sampling.Therefore, ECDIS-Net national coordinators were requested to invite a representative sample of the local clinical microbiology laboratories (about 10%) in each country to participate in the survey.In Austria and Norway, the laboratories were selected by random sampling; all other countries used non-random convenience sampling [22].Selected laboratories were emailed an initial survey in October 2011: some laboratories replied in 2012.All respondents to the initial survey received a follow-up survey in June 2014.

Data collection
Data were collected through a centralised web-based system (Questback, New York, United States).In 2011, the initial survey contained questions on several aspects of local routine diagnostics, including indications for undertaking CDI diagnostics and methodologies.Laboratories were requested to report the type of screening test primarily used for CDI diagnostics and confirmatory test (if applicable).For both, they could report more than one test.In 2014, the follow-up survey listed 10 diagnostic algorithms each designated as either 'optimal', 'acceptable' or 'incomplete' (Table 1).Laboratories were requested to estimate the percentage of samples that had been tested according to each algorithm listed, or to describe their usual diagnostic algorithm and estimate the corresponding percentage.The categorisation of CDI diagnostic algorithms was made by some of the ECDIS-Net experts who were also involved in revising the ESCMID diagnostics guidelines for CDI [6].Algorithms designated as optimal had high sensitivity and specificity (not specifically defined), detection of free toxins in faeces and a rapid turnaround time [23].Acceptable algorithms met the same criteria but without detecting free toxins in faeces.Any other algorithm was designated as incomplete.The 2014 follow-up survey additionally contained questions on barriers to apply optimal or acceptable diagnostic algorithms and changes in the indications for sending samples for CDI diagnosis by medical doctors.

Data analysis
To allow comparison, data on diagnostics from the 2011 initial survey were distributed into the three categories of diagnostic algorithms defined in 2014.For each local laboratory, CDI diagnostics, i.e.CDI testing practices, were considered optimal if more than 80% of the samples followed an optimal diagnostic algorithm, and acceptable if more than 80% of the samples followed either an optimal or acceptable algorithm.CDI diagnostics of all other algorithms were considered incomplete.When a laboratory reported a three-step algorithm by applying a third diagnostic test when the screening and confirmatory tests were contradictory, this algorithm was allocated to the best-matching twostep algorithm.Changes in local laboratory diagnostic capacity were evaluated by the McNemar's test [24], and changes in the use of optimal, acceptable and incomplete algorithms in 2011 and 2014 were evaluated by a Bowker test for symmetry [24].A sensitivity analysis was performed using two assumptions on missing data in 2014, i.e.CDI diagnostics one category inferior (Table 1) than in 2011 and CDI diagnostics one category superior than in 2011.Data were analysed using IBM SPSS statistics 20 (SPSS Inc., Chicago, United States).

Data collection and analysis
All ECDIS-Net national coordinators received an initial survey in May 2011 and a follow-up survey in June 2014.Both surveys contained questions on national typing capacity (defined as any laboratory in the country performing typing) and on molecular typing methods, asking which were available in their country from a list of common methods [18].

Participants
Questionnaires on local diagnostic and typing capacity for CDI were completed by 126 (61%) of 206 laboratories in 2011-12 and by 84 (67%) of these same 126 laboratories in 2014 (Table 2).A total 124 (98%) of the 126 responding laboratories in 2011-12 provided microbiological services to hospitals, of which 103 (83% ) served at least one university, secondary or tertiary care hospital.In addition, 66 (53%) provided microbiological services to long-term care facilities, of which 45 provided services to nursing homes.Furthermore, 65/124 (52%; data were missing for two laboratories) provided medical services to other healthcare services (e.g.general practitioners).In 2011 and 2014, 120/126 (95%) and 83/84 laboratories (99%, among responders to both questionnaires; p = 0.50), respectively, reported that they performed CDI laboratory diagnostics.

Indications for Clostridium difficile infection diagnostics
The indications for CDI diagnostics reported in 2011 are listed in Figure 1.In 2014, a change of indications for sending samples for CDI diagnosis by medical doctors was observed; 16 (19%) of 83 laboratories reported that one or two changes had occurred since 2011.Several laboratories introduced the use of Bristol stool scores to assess stool consistency for sample selection (n = 5).Also, patient populations that were previously not monitored for CDI (e.g.outpatients, high-risk populations) were later explicitly included in protocols (n = 3) and awareness and recognition of CDI among clinicians had improved (n = 5).Other improvements of sample selection were also reported (n = 5), i.e. application of guidelines for sample selection (n = 3) and/ or the three-day rule, i.e. diarrhoea after three days of hospitalisation (n = 1), and unspecified attempts to improve sample selection (n = 1).

C. difficile infection diagnostics
In 2011, 17 (14%) of 120 laboratories had optimal CDI diagnostics, 12 (10%) acceptable diagnostics and 91 (76%) incomplete diagnostics (Table 3).Incomplete algorithms included use of EIA toxin detection for screening with or without a confirmatory test, or a combination of EIA GDH and toxin detection without other tests for confirmation.Among laboratories responding to both the 2011 and 2014 surveys and that performed CDI diagnostics at both time-points (n = 81), the percentage of laboratories with optimal CDI diagnostics increased from 19% to 46% and that with acceptable CDI diagnostics from 10% to 15% while the percentage of laboratories with incomplete CDI diagnostics decreased from 72% to 40% (p < 0.001; Table 3).Two laboratories without any diagnostics in 2011 had optimal and incomplete CDI diagnostics, respectively, in 2014.

Sensitivity analysis
Laboratories with optimal CDI diagnostics in 2011 were more likely to respond to the 2014 survey (15/17)  Barriers to optimal/acceptable diagnostics for C. difficile infection Barriers to applying optimal or acceptable algorithms were examined in 2014.Of the 33 laboratories with incomplete CDI diagnostics, 17 indicated that materials or tests were too costly, six indicated receiving insufficient reimbursement for tests from insurers and five had insufficient availability of trained staff.Of the 50 laboratories that had optimal or acceptable CDI diagnostics, 10 also indicated that materials or tests were too costly, seven indicated receiving insufficient reimbursement from insurers and five had insufficient availability of trained staff.Ten laboratories that responded in 2014 indicated that they disagreed with the project's designations of the CDI diagnostic algorithms as optimal, acceptable or incomplete.

Participating countries
The national coordinators of 31 and 32 countries responded to the national survey in 2011 and 2014, respectively (Table 2).Data were collected separately for the four countries within the United Kingdom (UK), i.e.England, Northern Ireland, Scotland and Wales, but the UK was counted as one country.

Changes in national diagnostic capacity
In 2014, eight of the 32 responding countries (France, Germany, Greece, Latvia, Luxembourg, Sweden, Switzerland, Turkey) reported no change in national/ subnational laboratory diagnostics for CDI.Conversely, 24 countries reported one or more changes in national/ subnational laboratory diagnostics for CDI since 2011 (Figure 2).Specifically, 16 countries had experienced a change in availability of commercial diagnostic tests (Bulgaria, Czech Republic, Estonia, Hungary, Ireland, Italy, Lichtenstein, Lithuania, the Netherlands, Norway, Poland, Portugal, Romania, Serbia, Slovenia, UK), 10 countries had new or revised guidelines for CDI diagnostics (Austria, Czech Republic, Denmark, Estonia, Hungary, Ireland [25], Italy, Lithuania, Romania, UK) and three countries had changes in relevant legislation (Hungary, Poland, Romania).Three countries (Belgium, Croatia, Czech Republic) had implemented changes in reimbursement policies for diagnostic tests.Greece had limited access to and reimbursement of materials in both 2011 and 2014.In 2012, the UK implemented 'harmonised' diagnostics using GDH screening (or NAAT) and EIA toxin detection (or CCA) in all its laboratories [26].

C. difficile national typing methods
The  [27].The percentage of laboratories with optimal CDI diagnostics increased from 19% to 46%, and that with acceptable diagnostics increased from 10% to 15%.Importantly, the ESCMID-recommended twostep diagnostic algorithm [6] became more common.Nevertheless, we still observed a considerable variation in CDI diagnostics within and between European countries, in line with another European study with 482 participating hospitals in 2011-13 [3].This variation in diagnostics can substantially affect CDI incidence rates obtained by surveillance [15,28].Our survey showed that suboptimal CDI diagnostics may result from, for example, financial restrictions or limited availability of trained staff.As a consequence of the disagreement by a sizable minority of laboratories with the designation of diagnostic algorithms, the ESGCD undertook to revise its diagnostic guidelines [6] and propose an algorithm that can also be implemented in laboratories with limited numbers of trained staff and limited financial resources.These revised guidelines will be published in 2016 on behalf of ESCMID.
Among countries having national guidelines available, the UK was the only one that had succeeded in harmonising CDI diagnostics, by recommending a single two-test diagnostic algorithm ('comprising a GDH EIA (or NAAT/PCR) followed by a sensitive toxin EIA') [3,26].The recommendations in the UK Department of Health guidance were supported by local study data and inclusion of frequently asked questions to allay objections of the laboratories to implementing the proposed diagnostic algorithms [26].Furthermore, the diagnostics guidance was one of many C. difficilerelated activities in the UK, for example, implementation of mandatory CDI reduction targets with financial penalties for national health services [29].There probably are two possible ways to optimise testing: either to promote one national diagnostic algorithm or to promote the use of optimal testing strategies by local laboratories.However, the proposed algorithm in the UK was not fully compliant with the designation of diagnostic algorithms as optimal in this survey, highlighting the need for further discussion among experts to reach a consensus.Another example is Spain, where several national studies and meetings were organised [30,31] that resulted in an opinion document to enhance optimal diagnostics for CDI [32].We hope that the national reference laboratories that participated in the ECDIS-Net project will follow these examples and promote optimal diagnostics for CDI and its implementation in local laboratories.Switzerland).b CDI diagnostics were considered 'optimal' if > 80% of the samples followed an 'optimal' testing algorithm, and 'acceptable' if > 80% of the samples followed either an 'optimal' or 'acceptable' testing algorithm.CDI diagnostics of all other laboratories were considered 'incomplete'.The diagnostic algorithms are described in Table 1.c Two laboratories that did not perform CDI laboratory diagnostics in 2011 were not included.These laboratories indicated in the 2014 questionnaire that they used optimal and incomplete CDI diagnostics, respectively.d The percentages in this column do not add up to 100 due to rounding.method's suitability for standardised CDI surveillance [17].

Typing capacity
We assume that ECDIS-Net activities during 2012-14, including a training programme for C. difficile PCR ribotyping, contributed to the increased PCR ribotyping capacity.For example, Romania joined the training programme in 2012 and received a set of reference strains from the ECDIS-Net project and is now able to apply PCR ribotyping in their national surveillance.Poland reported having started their first national surveillance programme, stimulated by ECDIS-Net activities in 2012 [33].A few countries (Hungary, Italy, Slovenia) had national surveillance under development at time of the 2014 survey.Despite these positive trends, our study also indicates that some European reference and local laboratories are affected by limited resources and budget reductions, which hamper implementation and technical improvements of molecular typing methods.

Limitations
This study has several limitations including the small, non-random selection of local laboratories for both surveys and the moderate response rate, limiting the degree to which conclusions can be extrapolated to all European microbiological laboratories.The representativeness of the invited and participating laboratories could not be assessed due to the absence of a suitably complete European register.Laboratories with better CDI diagnostics may have been more likely to participate in the original and follow-up surveys, leading to an overestimation of the number of laboratories with optimal CDI diagnostics in Europe.Additionally, the categorisation of CDI diagnostic algorithms into three levels, although made through a series of consultations with a team of international experts from the ECDIS-Net project, was based on expert opinion and some subjectivity cannot be excluded.Also, although the 2014 questionnaire for local laboratories requested quantitative data on the percentage of tests that followed each algorithm on a provided list, as the list had the subheadings 'optimal', 'acceptable' and 'incomplete', it is possible that those responding overestimated the proportion of desirable answers.We estimate that this reporting bias was minimal as for almost all laboratories, just one algorithm was used.

Conclusions
We conclude that the ECDIS-Net project laid the foundations for Europe-wide surveillance of CDI, although increased use of optimal diagnostic algorithms should be promoted, taking into consideration the limited resources and budget cuts in several European countries.The ESGCD revised the ESCMID diagnostics guidelines for CDI, which, once published, should contribute to standardisation of CDI diagnostics at local and national level in Europe.Typing capacity for CDI in Europe was acceptable overall; however, an internationally standardised capillary PCR ribotyping protocol is now available [17] and requires further implementation in European countries.We would recommend that these important steps are considered as part of the integration of C. difficile molecular typing data in The European Surveillance System (TESSy), within the ECDC-coordinated Europe-wide CDI surveillance (since 1 January 2016) [34].
the European Study Group on C. difficile (ESGCD).The authors would like to thank Marjolein P Hensgens for advice regarding data analyses.

Figure 1
Figure 1Criteria for selection of faecal samples tested for Clostridium difficile among responding local laboratories that participated in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011 (n = 120) a

Figure 2
Figure 2Reported changes affecting national/subnational laboratory diagnostic capacity for Clostridium difficile infection between 2011 and 2014 in participating European countries (n = 32) a Laboratories in 31 countries responded to the 2011 survey: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lichtenstein, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Spain, Sweden, Switzerland, Turkey and the United Kingdom (not including Wales).Serbia did not participate in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011.No laboratories in Slovakia and Wales were invited to participate by ECDIS-Net national coordinators in 2011. a

Figure 3
Clostridium difficile typing methods available in countries that participated in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011 (n = 31) a and 2014 (n = 32) a Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, Switzerland, Turkey and the United Kingdom (not including Wales).No data were available for Iceland.In 2011, Serbia did not participate in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project.In 2014, Serbia participated in the ECDIS-Net project and responded to the 2014 questionnaire, as did Wales, and so the number of responding countries in 2014 was 32.

Table 1
Criteria for categorisation of Clostridium difficile infection diagnostic algorithms, survey of European countries participating in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project, 2011 (n = 31) a and 2014 (n = 26) a Laboratories in 31 countries responded to the 2011 survey: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lichtenstein, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Spain, Sweden, Switzerland, Turkey and the United Kingdom (not including Wales).Serbia did not participate in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011.No laboratories in Slovakia and Wales were invited to participate by ECDIS-Net national coordinators in 2011.Laboratories in 26 countries responded in 2014 (no data from laboratories in Croatia, Iceland, Latvia, Slovenia and Switzerland).b Categorisation of CDI diagnostic algorithms in the second survey, in 2014 [21].c Corresponding CDI diagnostic algorithms in the second survey, in 2014 [21].
Dutch National Reference Laboratory for C. difficile (Leiden University Medical Centre, Leiden, and the National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands) coordinated data collection in 2011 and in 2014 by cross-sectional surveys among two target groups: (i) local microbiology laboratories, in order to evaluate changes in routine laboratory diagnostics; and (ii) national coordinators, i.e. representatives of national or regional reference laboratories nominated by competent bodies for surveillance on the request of ECDC, to evaluate national changes in diagnostic and typing capacity for C. difficile.In 2011 and 2014, 32 and 33 countries participating in the ECDIS-Net project were invited to take part in CDI: Clostridium difficile infection; EIA: enzyme immunoassay; GDH: glutamate dehydrogenase; NAAT: nucleic acid amplification test.a

Table 2
Response of participating European countries to local laboratory (n = 31 and n = 26, respectively) and national/subnational surveys (n = 31 and n = 32, respectively) on Clostridium difficile infection diagnostic and typing capacity, 2011 and 2014 For the UK, data were analysed separately for England, Northern Ireland, Scotland and Wales, but the UK was counted as one country.
a b Serbia did not participate in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011.c No laboratories in Slovakia and Wales were invited to participate by ECDIS-Net national coordinators.
capacity for various C. difficile typing methods in participating countries in 2011 and 2014 is depicted in Figure 3.The number of countries able to perform any method of typing increased from 22/31 countries in 2011 to 26/32 countries in 2014.Only six countries (Croatia, Cyprus, Estonia, Lichtenstein, Lithuania, Serbia) reported that they did not have any national typing capacity in 2014 (none of these countries had typing capacity in 2011); however, Lichtenstein sent samples to another country (Austria) for typing.

Table 3
Laboratories participating in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project according to their diagnostics category, 2011 (n = 120)a and 2014 (n = 81) a Laboratories in 31 countries responded to the 2011 survey: Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lichtenstein, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, Slovenia, Spain, Sweden, Switzerland, Turkey and the United Kingdom (not including Wales).Serbia did not participate in the European Clostridium difficile Infection Surveillance Network (ECDIS-Net) project in 2011.No laboratories in Slovakia and Wales were invited to participate by ECDIS-Net national coordinators in 2011.Laboratories in 26 countries responded in 2014 (no data from laboratories in Croatia, Iceland, Latvia, Slovenia and a