Concordance between European and US case definitions of healthcare-associated infections

Implementation of surveillance of healthcare-associated infections (HAI) has been shown to result in decreasing HAI rates and contributes to the prevention of HAI[1‐3]. Feedback of data on HAI rates to clinical staff has been shown to be a key factor reducing these rates[4‐6].

Comparing HAI rates of one’s own institution with reference data seems to be particularly successful. In the 1970s, the US Centers for Disease Control and Prevention (CDC) created the National Nosocomial Infection Surveillance System (NNIS) and published uniform surveillance definitions for nosocomial infections[7‐9]. These definitions have been updated gradually for surgical site infection (SSI)[10], for ventilator-associated pneumonia (VAP)[11], primary bloodstream infection (BSI)[12] and in 2010 for urinary tract infection (UTI)[13]. Key terms such as “device-associated infection” or “intensive care unit (ICU)-associated infection” were also defined[14]. This system is now integrated as part of the National Healthcare Safety Network (NHSN)[12].

In the 1980s and 1990s, many European countries performed national prevalence studies of HAI and established national surveillance systems using CDC definitions or a modified version of these definitions[15], while other countries developed their own surveillance definitions that better reflected European diagnostic practices. The first harmonization of national surveillance activities in Europe was performed by the Hospitals in Europe Link for Infection Control through Surveillance (HELICS) project, which was funded by the European Commission in the context of Decision 2119/98/EC of the European Parliament and of the Council on communicable disease surveillance and control in EU Member States[16].

The HELICS project (2000–2004) developed case definitions for surgical site infection (SSI), pneumonia (PN), bloodstream infection (BSI), catheter-related infection (CRI) and urinary tract infection (UTI) and recommended their use in EU Member States[17, 18]. The work of HELICS was continued as a component of the European Commission-funded Improving Patient Safety in Europe (IPSE) network (2005–2008). The IPSE network aimed at contributing to European surveillance of HAI by describing HAI epidemiology, improving the understanding of inter-country variation of HAI rates and facilitating quality-of-care improvements in a multi-centre setting. In July 2008, the IPSE network was transferred to the European Centre for Disease Prevention and Control (ECDC)[19, 20]. Since this date, HAI surveillance activities in Europe are coordinated by ECDC and the network was re-named the Healthcare-Associated Infections surveillance Network (HAI-Net). HAI-Net adopted the European (HELICS/IPSE) definitions for its HAI surveillance modules and for the ECDC point prevalence survey of HAI in European acute care hospitals. Comparisons of HAI rates between countries are essential to raising awareness about HAI and their prevention and control, but require a standardized methodology, including uniform definitions. Because they were implemented independently, national HAI surveillance systems in European countries decided to use either the US (CDC/NHSN) definitions[12‐14] or the European (HELICS/IPSE) definitions[17, 18] and questions have been raised about whether comparisons of HAI rates between national networks were indeed appropriate. While adoption of the European definitions is mandatory for newly implemented national HAI surveillance systems, changing definitions could interrupt continuity of reference data and require reorganization for an existing national surveillance system.

The present study was conducted to assess the concordance between US (CDC/NHSN) definitions and European (HELICS/IPSE) definitions of HAI for inter-country comparison of HAI rates. The study was initiated and sponsored by ECDC through a specific service contract (ECD.1781) with the Institute for Hygiene and Environmental Medicine, Charité – University Medicine Berlin, Germany.

HAI surveillance methods vary across Europe. Some countries use European definitions while other countries use US definitions. As a contribution to further harmonization of Europe-wide surveillance of HAI, this study assessed the concordance between US and European definitions of BSI and of PN, two major types of HAI that are partly preventable[27] and are under surveillance in most European countries. The recommendations of Landis and Koch for evaluating the strength of an agreement were used[28]. Overall, an “almost perfect” agreement was found for PN (κ = 0.99). This was different when PN cases were subdivided into clinically and microbiologically defined PN. More PN cases were classified as microbiologically-defined PN following the European definitions than the US definitions. This was still the case when Europe-defined PN2 cases which are based on the criteria “non-protective sample and quantitative culture” were considered as clinically defined PN. This difference was no longer evident when all US-defined PNU3 (“PN in immunocompromised patients”) cases were reclassified into the US-defined categories PNU1 or PNU2. Since all 26 PNU3 cases could be classified as either PNU1 cases (n = 2) or PNU2 cases (n = 24), the results of this study suggest that the PNU3 subcategory may not be essential when performing surveillance of PN in immunocompromised patients.

As expected, concordance of BSI definitions was only “substantial” according to Landis and Koch[28]. Since one major criterion of US definitions, i.e. signs and symptoms of BSI must not be related to an infection at another site, is not included in the European definitions, 51 (42%) BSI cases were not identified with the US definition. With the European BSI definition, which includes the specification of the origin of the BSI, these 51 BSI cases were reported as “secondary to another infection site”.

European definitions provide two more categories, i.e. “catheter” and “unknown”, for origin of BSI[18]. The origin “catheter” was reported in 37 (30%) BSI cases and the origin “unknown” was reported in 35 (29%) BSI cases, which correspond to the “primary BSI” of the US definitions. All 72 of these BSI cases were also defined as LCBI cases with the US definitions. Thus for a potential comparison, US-defined LCBI cases should only be related to Europe-defined BSI cases with either “catheter” or “unknown” origin.

Definitions of the key term “ICU-acquired HAI” varied between the U.S. and Europe. According to European definitions 97% of HAI cases were defined as ICU-acquired (i.e. HAI occurring later than 48 hours after admission in the ICU). By contrast, according to the US definitions, all HAI cases were defined as ICU-acquired (i.e., no evidence that the infection was present or incubating at the time of admission to the ICU). Since US definitions do not specify a time period between admission to the ICU and onset of symptoms, it is easy to explain why a few more HAI were recorded as ICU-acquired according to the US definitions. Nevertheless, agreement for the key term “ICU-acquired HAI” was still “almost perfect” according to Landis and Koch[28].

A strength of our study is that it was performed in seven European countries with different diagnostic methods and habits reflecting the variety of diagnostic practices in Europe. A limitation of the study is that the time of admission to the ICU and the time of onset of the HAI were recorded less precisely (in “days” instead of “hours”) than in the original definition because the findings of the study pre-test revealed major difficulties in collecting time of onset data in “hours”. As a consequence, all HAI occurring on or after the third day of ICU stay (rather than after 48 hours according to the European definitions), were defined as ICU-acquired. A further limitation was that specific patient groups, such as paediatric and immunocompromised patients, were likely to be underrepresented in the study.

In conclusion, countries using either US or European definitions for HAI surveillance can compare PN and primary BSI rates as long as the following points are taken into account. First, data should of course be valid and be collected following the original US and European definitions since country-specific modification of definitions may result in additional differences[29]. Second, PN data should always be compared in total, but not as subcategories of clinically defined and microbiologically defined PN. Third, for BSI the source should always be reported since all BSI cases with the origin “secondary to another site” according to European definitions should be excluded when making comparisons with US-defined BSI. Fourth, only Europe-defined BSI cases with either “catheter” or “unknown” origin should be compared to US-defined LCBI cases. Fifth, there are differences between US and European surveillance protocols, other than just case definitions of HAI, and these differences should be taken into account before performing comparison of HAI rates. Finally, comparisons are valid as long as US and European definitions do not change. Indeed, changing US definitions for PN to ventilator-associated complications under the influence of public reporting[30, 31] would certainly affect the current good concordance between US and European definitions of PN.