Japan nosocomial infections surveillance (JANIS): a model of sustainable national antimicrobial resistance surveillance based on hospital diagnostic microbiology laboratories

Antimicrobial resistance (AMR) is now recognized as a major threat to public health that limits patients’ treatment options for bacterial infections. In 2015, the 68th World Health Assembly adopted the Global Action Plan on Antimicrobial Resistance [1]. One of the five strategic objectives of the Global Action Plan is to strengthen knowledge through surveillance and research [2].

In Japan, methicillin-resistant Staphylococcus aureus (MRSA) was one of the first pathogen that made health professionals realize that AMR is a real-world threat. MRSA was prevalent in healthcare settings in Japan in the 1980s, especially in surgical wards, where it causes post-surgery nosocomial infections [3]. Consecutively, multidrug-resistant Pseudomonas aeruginosa also emerged and spread in hospitals across Japan. This led the Ministry of Health, Labour and Welfare (MHLW) to establish a surveillance system for nosocomial infections and AMR, the Japan Nosocomial Infections Surveillance (JANIS).

Common types of infectious disease surveillance are usually patient-based, with case definitions consisting of clinical diagnoses based on the symptoms and signs of the patient, often accompanied by diagnostic test results [4]. By contrast, AMR surveillance must focus on the isolated pathogen rather than the patient. Microbiological test results of bacterial identification and antimicrobial susceptibility testing (AST) are the primary components of AMR surveillance [5]. Therefore, many existing AMR surveillance systems rely on data generated by microbiology laboratories [6‐9].

The main AMR surveillance component of JANIS is its Clinical Laboratory (CL) division, which collects data from diagnostic microbiology laboratories of hospitals. However, the outstanding characteristic of the Japanese medical system is that it has one of the largest number of hospitals in the world [10]. In 2014, there were 8493 Japanese hospitals, by far exceeding 5627 hospitals in the United States [11]. The large number of hospitals results in dispersed healthcare resources, and more than 1000 hospitals run their own microbiology laboratories. Therefore, a practical surveillance system was required to continuously accumulate data from those laboratories to achieve national surveillance. In this study, we aimed to propose a model of sustainable national AMR surveillance which overcame those barriers and creates not only national reports but also benchmarking reports for each hospital to facilitate infection control practices.

From the beginning, JANIS was designed to collect comprehensive bacterial data from hospital diagnostic microbiology laboratories even though it was initially limited to blood and CSF samples. This seemed to be a promising approach because both automated microbiology test systems and laboratory information systems became widely used in hospitals during early 2000’s. Hospitals already have the electronic data from their routine bacteriological tests, which are supposed to be submitted monthly to JANIS.

Data collection focusing on blood and CSF samples minimizes heterogeneity and improves comparability between reporting laboratories but information would be limited to invasive infection [15]. European Antimicrobial Resistance Surveillance Network (EARS-Net) and Central Asian and Eastern European Surveillance of Antimicrobial Resistance focus on blood and CSF samples [8, 16]. In UK, antimicrobial resistance surveillance on routinely generated laboratory results has started in 1989, including the mandatory reporting of MRSA bacteremia to the Public Health England and its predecessors [7, 17, 18]. However, WHO Global Antimicrobial Resistance Surveillance System (GLASS) recommends comprehensive data collection at the national level to include other species and specimen types beyond their priority list [2].

During the first several years after the JANIS CL division was launched, half of the enrolled hospitals did not submit surveillance data. This failure may have been due to the insufficient announcement of surveillance methods, delayed or poor-quality Feedback Reports which were printed out and sent by mail 6 to 12 months after data submission, and limited internet access during that time period.

The revision of JANIS in 2007 was intended to settle these issues, and it was possible partly because most of the hospitals had improved their internet access by the middle of the 2000s. As a result of those systemic revisions and annual recruitment of participants, the number of member hospitals increased year by year, even though JANIS is a voluntary surveillance system. The inclusion of participation in the JANIS CL division as a prerequisite for receiving additional reimbursement for infection control demonstrates that JANIS has been established as a sustainable national surveillance system.

Although the number of participating hospitals has been increased, the majority of participants are large hospitals, and only about 4% of small hospitals with less than 200 beds were covered by the JANIS CL division in 2015. Thus, JANIS CL division data may over-represent AMR epidemiology of large hospitals in Japan. This may be because small hospitals usually do not have their own microbiology laboratories and their subscribing commercial laboratories may not provide surveillance data. In addition, in the absence of their own microbiology laboratories, these hospitals are unlikely to employ professionals who can interpret and utilize feedback from the JANIS CL division. How to collect and analyze surveillance data from those small hospitals is an important issue for JANIS to address in the future, and these questions are also critical for AMR surveillance systems in general. The EARS-Net also pointed out the over-representation of large hospitals, and also noted that tertiary care hospitals are much more likely to participate in this surveillance than smaller hospitals [19].

In addition, even if the coverage of small hospitals were increased, interpretation of surveillance data requires careful consideration of the denominator due to variability in the frequency of culture sampling. The Society for Healthcare Epidemiology and the Healthcare Infection Control Practices Advisory Committee suggests calculating prevalence of AMR per 100 patients or 1000 patient-days admitted to the hospital [20]. However, because AMR can only be detected by specimen submission, JANIS calculates the prevalence of AMR by defining the denominator as the number of patients for whom specimens were submitted. Moreover, one of the metrics to measure the occurrence of AMR proposed by WHO GLASS is to have the total number of sampled patients per specimen type as the denominator. To our knowledge, JANIS is the only national surveillance that truly collects and analyzes comprehensive data, including active surveillance testing and culture-negative results for all specimens [21]. Data for all specimens can provide information about how samples were submitted from each hospital. These data revealed that the number of samples per 100 beds for small hospitals is half of that of large hospitals, and more sterile samples such as blood cultures are obtained in large hospitals, whereas more easy-to-obtain samples (such as respiratory and urine samples) are collected in small hospitals. Inversely, the proportion of culture-positive specimens of small hospitals is higher than that of large hospitals, indicating minimal bacterial investigation in small hospitals, which may overestimate the prevalence of AMR. The higher prevalence of MRSA in the western part of Japan may partially be explained by that there are more small hospitals per 100,000 capita in that region, and the sampling frequency is low in that type of hospital.

Quality assurance of diagnostic microbiology laboratories and data validation is fundamental to AMR surveillance, especially for a system that collects data but not bacterial isolates to be tested in a central reference laboratory. One reason that JANIS CL division initially targeted only large hospitals was to ensure data reliability, as those hospitals are expected to employ clinical microbiology professionals. The diffusion of automated microbiology test systems among hospital laboratories in Japan guarantees the basic quality of microbiological tests because the manufacturers of those systems provide quality assurance services to the custom hospitals. There are a few domestic external quality assessment (EQA) programs in which most of the JANIS member hospitals with in-house laboratories may be enrolled. Large commercial laboratories are also enrolled in internationally approved EQA, but those programs are difficult to afford for hospitals and small commercial laboratories. Accordingly, a national scheme for laboratory quality assurance is required. Data validation for Unusual AMR by the JANIS CL division could play a certain role in quality assurance because it would inform hospitals what type of microbiology test results requires confirmation.

The accumulation of nationwide surveillance data over a decade enables us to describe the long-term trends in AMR prevalence. There are encouraging findings that AMR of major clinically important bacterial species in Japan is decreasing or maintained at a low level. For instance, the prevalence of MRSA and multidrug-resistant Pseudomonas aeruginosa has been decreasing since 2008, to around half (or less) of the previous levels. It is also notable that multidrug-resistant Acinetobacter spp. (MDRA), which are prevalent in many Asian countries, are still well controlled in Japan [22]. On the other hand, antimicrobial resistance among Escherichia coli keeps increasing as many other countries.

Even though the principal purpose of JANIS is to create national AMR profiles, it was initiated to support infection prevention and control in medical facilities. Feedback Reports to each participating hospital are created to raise awareness of infection control practices and to motivate data submission through the box plot chart, which intuitively benchmarks the prevalence of AMR of each facility against the entire distribution. To enable this benchmarking, JANIS collects MIC values rather than interpreted RIS because there are different standards and frequent updates for the categorization of RIS. In addition, JANIS applies a unified de-duplication program to exclude repeated isolates from the same patients.

Stratification by hospital bed counts and prefectures revealed different distributions of AMR prevalence among hospital types and geographical regions. This indicates that the goals of AMR containment should be set in consideration of local factors, and that regional cooperation is essential. In fact, several regional networks share JANIS files to create their local antibiograms.

There are several limitations of the JANIS CL division, suggesting opportunities for improvement. WHO GLASS requires data on AMR combined with patient and microbiological information. All of them are included as data elements in JANIS data format but currently clinical data are rarely reported because of the different data sources. If integration of patient and microbiological information in each facility became more feasible, it may become possible to fulfil GLASS requirement.

Inclusion of data from small hospitals is challenging, but collection of such data has been gradually increasing; in 2017, 8.5% of 5800 small hospitals are participating in JANIS, almost double the number from 2015. Population-based AMR surveillance, which captures AMR data from an entire predefined population or a representative sample, is considered to be the gold standard for surveillance [4]. However, laboratory-based surveillance covering a small number of sentinel sites (fewer than 20% of national hospitals) is considered to be sufficient to provide relatively robust estimates of national AMR prevalence, assuming that the sampling is constant and the sample is representative of the target population [15]. Therefore, increasing sampling frequency, especially prior to treatment with antibiotics, may increase the representativeness of the JANIS Open Report for small hospitals more effectively than recruiting additional hospitals. Moreover, data validation is a challenge for small hospitals without microbiology staff. Accordingly, support of regional hospital networks is required for purposes of education.

JANIS achieved sustainability as a national AMR surveillance by initially focusing on large hospitals with diagnostic microbiology laboratories in order to establish the management scheme of the surveillance. Comprehensive data for all specimens promotes understanding of the sampling frequency and prevalence of AMR. Because JANIS is a well-established system for providing rich information to guide action both locally and nationally, it may also be utilized for sharing AMR data globally.