Introduction

The World Health Assembly adopted the Global Action Plan on Antimicrobial Resistance (AMR) in May 2015 [1]. In Japan, the National Action Plan on Antimicrobial Resistance was adopted in April 2016 and included two goals [2], “implementing appropriate infection prevention and control” and “appropriate use of antimicrobials.” These goals are of particular relevance to healthcare facilities in terms of preventing the spread of antimicrobial-resistant organisms. Specifically, at the field level, linking efforts between existing infection control team (ICT) and antimicrobial stewardship (AMS) programs was suggested to be important [2].

Previous studies reported that human resources (expressed as full-time equivalents (FTEs) of infection control practitioners) and FTE-to-bed ratios were related to improvements in AMS [3–5]. However, the definition of improvement varied from study to study. For example, one study used an increase in the number of implemented AMS programs [3] to evaluate the performance of AMS, while another study examined the effectiveness of each program [4].

The purpose of the present study was two-fold: to report the results of nationwide multicenter questionnaire surveys on countermeasures against AMR and infections in Japanese teaching hospitals, and to identify latent statuses, which might imply underlying subgroups of hospitals with similar achievement levels of AMS, and examine the effects of FTE-to-bed ratios of ICT members on the latent statuses.

Methods

We posted questionnaires to all teaching hospitals in Japan (n = 1017 as of 2015). To examine changes over a period of roughly 1 year, surveys were conducted twice in November 2016 and February 2018 (see the English translation of the questionnaires in Additional file 1). No intervention was provided by our study team between the two surveys. The contents of the questionnaire included basic information, such as the number of beds (1st survey only), questions divided into sections 1 to 12 for countermeasures against AMR based from a previous study [6] and a guide published by the Japanese Ministry of Health, Labour and Welfare [7], and section 13 for results of bacterial cultures, as follows: 1. Organizational structure for nosocomial infection control; 2. Activities of ICT; 3. Preventive measures by the route of infections; 4. Maintenance of medical equipment; 5. Standard precautions; 6. Ward; 7. Intensive care unit (ICU); 8. Operating room; 9. Prevention of postoperative infections; 10. Management of food hygiene in hospitals; 11. Management of medical waste; 12. Cleaning, disinfection, and sterilization of instruments; and 13. Antimicrobial-resistant organisms. The questions were answered (1) numerically (e.g., number of physicians) or by choosing (2) either “yes” or “no” or (3) one among three to five options in order (e.g., “in approximately 100%/80%/50%/20%/0% of relevant cases”).

We analyzed valid responses, which included hospital information to link the 1st and 2nd surveys. We excluded duplicate responses to the same survey by the same hospital. Answers to questions in sections 1 to 12 are presented as medians and interquartile ranges, calculated after excluding missing values. For single-choice questions, we presented the proportions of “yes” or the most favorable option (e.g., “approximately 100% of relevant cases”). For these types of questions, we created a “missing” category. Student’s t-tests or Satterthwaite tests were used for continuous variables, and Cochran-Mantel-Haenszel tests for categorical variables, to compare results from the 1st and 2nd surveys.

The answers to questions in section 13 were the results of surveillance in 2015 for the 1st survey and 2016 for the 2nd survey, which were 1 year before each survey. We calculated the proportions of isolated microorganisms and antimicrobial-resistant organisms for each hospital that responded to both surveys. Wilcoxon signed-rank tests were used, assuming that the results from the 1st and 2nd surveys regarding section 13 were paired data.

To study the achievement level for AMS programs of hospitals, we performed a latent transition analysis (LTA), which is a longitudinal extension of latent class analysis [8]. Latent class analysis identifies underlying subgroups in a population, but the characteristics of these underlying subgroups are hard to observe directly; these are indicated by several observed variables [8]. While latent class analysis identifies underlying (unobservable) subgroups within a population as “classes,” LTA refers to the subgroups as “statuses” to reflect the fact that membership in the subgroups can change over time [8]. In this study, we performed LTA using data from hospitals that responded to both surveys, and time periods 1 and 2 for LTA were defined as those of 1st and 2nd surveys, respectively. Questions for which the proportion of the most favorable answer was less than 80% in the 2nd survey were used to classify hospitals into subgroups, latent statuses, with similar sets of answers to these questions. We excluded questions regarding handwashing sinks in ICUs, for which the proportion of the most favorable answer was less than 80%, given the lack of established guidelines. We also reduced the multiple categories in each question to two (most favorable/others) to improve the precision of estimates [9]. FTEs of ICT members were selected as a covariate that might affect the membership probabilities for time period 1. We determined the number of statuses by considering interpretability and fit statistics, and presented the fit statistics, status membership probabilities, transition probabilities, item-response probabilities, and estimated odds ratios for covariates. The domains, which consisted of several questions, were determined empirically according to the LTA results.

SAS® software version 9.4 (SAS Institute Inc., Cary, NC, USA) was used for all analyses, and PROC LTA (version 1.3.2) was used for the LTA [10]. A two-tailed significance level of 0.05 was used for all tests.

Results

Among 1017 teaching hospitals, 683 and 563 hospitals responded to the 1st and 2nd surveys, respectively. The numbers of valid responses were 678 for the 1st survey (response rate: 66.7%) and 559 for the 2nd survey (response rate: 55.0%) after excluding duplicated responses and those with missing hospital information. The number of hospitals that responded to both surveys was 437 (response rate: 43.0%).

The mean number of hospital beds was 434 (median, 389: 675 responses). Table 1 presents the results of the two surveys for all hospitals with valid responses and hospitals that responded to both surveys (see Tables S1 and S2 in Additional file 2 for more details). More than 99% of hospitals reported having active ICTs, with a median of 10 to 11 ICT members. Both crude numbers and FTEs of ICT members did not differ significantly between the 1st and 2nd surveys.

Table 1 Results of the 1st and 2nd questionnaire surveys
Full size table

More than 90% of hospitals had weekly ICT meetings, although proportions of specific activities differed from hospital to hospital (section 2): 79.9% (1st survey) and 66.7% (2nd survey) of hospitals had an antimicrobial stewardship team. More than 90% of hospitals indicated that they monitored and intervened to assure appropriate use of antibiotics, but only 70% had established intervention criteria. The proportions of hospitals with intervention criteria for patients administered anti-methicillin-resistant Staphylococcus aureus (MRSA) antibiotics and carbapenems were approximately 60 and 50%, respectively. The proportions of hospitals that performed surveillance varied by the types of infections: catheter-associated urinary tract infections and ventilator-associated pneumonia were monitored less frequently compared to surgical site infections and central line-associated bloodstream infections.

With regard to the maintenance of medical equipment (section 4), less than 50% of hospitals indicated that they used maximal barrier precautions for central line catheter insertion and prepared intravenous hyperalimentation admixtures on clean benches.

For standard precautions (section 5), approximately 50% of hospitals held practical hand hygiene training sessions for new employees regardless of professions; the remaining hospitals trained new employees of selected professions only. Training regarding personal protective equipment for all new employees was held in about 80% of hospitals, although less than 20% of hospitals held these training sessions every year.

Regarding the ICU (section 7), the proportion of hospitals that answered “yes” to “We have handwashing sinks at the entrance of ICU” was lower than the other questions in this section. Roughly 60% of hospitals had handwashing sinks at the ICU entrance, whereas approximately 80% of hospitals answered “yes” for other questions.

Less than 70% of hospitals responded that their staff members do not change their shoes when entering the operating room, and less than 50% had manuals regarding the duration of prophylactic antibiotics available in all departments (section 8). The proportion was lower than 80% even when hospitals that had manuals in selected departments were included (Tables S1 and S2 in Additional file 2).

Table 2 presents the proportions of isolated microorganisms and antimicrobial-resistant microorganisms. Among antimicrobial-resistant microorganisms, only the proportion of those belonging to the family Enterobacteriaceae decreased in 2016 compared with 2015. The proportions of antimicrobial-resistant Streptococcus pneumoniae and Escherichia coli increased during this period.

Table 2 Isolation proportions of microorganisms and antimicrobial-resistant microorganisms
Full size table

Tables 3, 4, 5, 6, 7, and Table S3 and Figure S1 in Additional file 2 show the results of the LTA. Five statutes, from the most favorable (status 1) to the least favorable (status 5), were identified (Table 3). Latent status 4 showed the highest status membership probabilities for both time periods (Table 4). As for transition probabilities, members of statuses 1, 2, 4, and 5 were stable in their status membership (Table 5). On the other hand, members of status 3 showed the lowest probability of remaining in the same status (42.7%, Table 5), with 32.7% moving to status 5 and 14.8% moving to status 4 (Table 5). We assigned five domains according to the item-response probabilities for each question (Tables 3, 6 and Figure S1 in Additional file 2): “antimicrobial stewardship” (domain 1); “surveillance” (domain 2); “medical and hospital equipment” (domain 3); “ICT activities regarding vaccinations and education of employees” (domain 4); and “acknowledgment of updating relevant guidelines” (domain 5). Compared to status 1, status 2 showed lower probabilities of having criteria for anti-MRSA antibiotic use and broad-spectrum antibiotic use, whereas status 3 had lower probabilities of performing surveillance. Status 4 had only one domain (i.e., domain 3) with higher probabilities for questions in it. Status 5 had no domain that showed higher probabilities compared to other statuses. In the analysis using the number of ICT members (FTE per 100 beds) as a covariate, the odds ratio of status 3 versus status 5 was 1.32, whereas odds ratios were 0.55 and 0.61 for statuses 1 and 2 versus status 5, respectively (p = 0.027, Table 7).

Table 3 Item-response probabilities for each question by identified latent statuses
Full size table
Table 4 Status membership probabilities for the 1st and 2nd time periods
Full size table
Table 5 Transition probabilities of each status from the 1st to 2nd time periods
Full size table
Table 6 Characteristics of each latent status
Full size table
Table 7 Odds ratio estimates of covariates
Full size table

Discussion

We conducted two surveys on AMR and infections in teaching hospitals in Japan, with an interval of approximately 1 year between the surveys. Most hospitals had activities of ICTs, however, actual activities differed among hospitals. The results of LTA suggested that there were five subgroups of hospitals, which were considered indicating similar achievement levels of AMS. The presence of local (i.e. hospital-level) guidelines for using anti-MRSA and broad-spectrum antibiotics, and the range of surveillance activities of each hospital were identified as two major determinants of the membership in each subgroup.

The proportion of hospitals with antimicrobial stewardship teams decreased during the study period. In fiscal year 2018 (after the 2nd survey), a fee for antimicrobial stewardship teams was introduced by the National Fee Schedule. To claim this fee, hospitals must fulfill requirements such as having at least one full-time staff member who is a physician with more than 3 years of experience in infectious disease treatment, a nurse with more than 5 years of experience working in a hospital, or a pharmacist or a laboratory technologist with more than 3 years of experience working in a hospital. Our results suggest that hospitals not fulfilling this requirement might have changed their answers to this question from “having an antimicrobial stewardship team” to “not having an antimicrobial stewardship team.”

The proportion of hospitals with preauthorization and/or restriction systems for the use of anti-MRSA antibiotics and broad-spectrum antibiotics decreased during the study period. Preauthorization and/or prospective audit and feedback interventions by AMS programs are strongly recommended [11]. Although more than 90% of hospitals in our study responded that they carried out monitoring and intervention activities, roughly 70% had established intervention criteria, and less than 40% had preauthorization and/or restriction systems for anti-MRSA antibiotics and broad-spectrum antibiotics. These proportions also decreased throughout the study period. The use of restricted antibiotics lists has been reported to reduce antimicrobial resistance rates and costs [12]. Thus, hospitals should consider introducing preauthorization and/or restriction systems for relevant antibiotics to enhance their AMS programs.

The proportions of hospitals with surveillance for ventilator-associated pneumonia and catheter-associated urinary tract infections increased slightly, but remained under 60%. Given that these infections are considered major healthcare-associated infections along with surgical site infections and central line-associated bloodstream infections, surveillance is recommended [13–15]. The proportion of hospitals performing active surveillance cultures was roughly 65%. However, active surveillance cultures for MRSA and vancomycin-resistant enterococci for all inpatients except for high-risk patients are not recommended [16]. The WHO Guidelines Development Group strongly recommends surveillance cultures for asymptomatic carbapenem-resistant Enterobacteriaceae and surveillance for carbapenem-resistant Acinetobacter baumannii and Pseudomonas aeruginosa despite a very low quality of evidence [17]. Further studies will be needed to determine the targets for active surveillance cultures and their efficacy.

For all questions regarding the ICU, the proportions of hospitals with the most favorable answers were less than 80%. This might be due to the fact that hospitals without an ICU were also included in this study. However, the proportion of hospitals that answered “yes” to the question about handwashing sinks at the ICU entrance was considerably lower (less than 60%) compared to those of hospitals that answered “yes” to the other questions. A Japanese guideline (2002) that recommended hospitals to place handwashing sinks at the ICU entrance [18] was revised to allow for the location to be based on staff accessibility [19]. However, since recent studies have suggested that sinks in the ICU might be a source of infections [20,21,22], further investigations will be needed on appropriate locations and specifications of sinks in the ICU.

The LTA identified five statuses. There was a slight increase in the most favorable status (status 1) over the course of the study period (23.6 to 25.3%). However, the least favorable status (status 5) also showed an increase (26.3 to 31.8%), which was mainly due to a decrease in status 3 (18.0 to 8.8%). Previous studies have reported that human resources (FTEs of infection control practitioners) and FTE-to-bed ratios were related to improvements in AMS [3–5], defined as an increase in the number of implemented AMS programs [3] or effectiveness of AMS programs [4]. However, improvements in AMS may not correlate with the number of implemented programs, considering that the weight of each program is unlikely to be equal. In fact, the results of the LTA do not support a relationship between increasing the number of FTEs per bed and being in more favorable latent statuses. The odds ratio of status 3 versus status 5 was 1.32, indicating that more infection control practitioners might be required to improve domain 1, “antimicrobial stewardship,” whereas improvement in domains 2, 3, and 5 could not be fully explained by an increase in human resources alone. However, since previous studies, as well as our study, did not account for patient-level variations, further studies will be needed to identify factors associated with AMS other than human resources.

This study had some limitations. First, response rates were 55.0% for all hospitals with valid responses and 43.0% for those that responded to both surveys. There may have been selection bias in these hospitals. Second, hospitals participating in our study may have different profiles of cases and individual risks. To address these issues, we plan to link administrative data and the data of this study for further analyses.

Conclusion

The present nationwide surveys revealed the need for more comprehensive AMS programs; specifically, hospitals should consider introducing preauthorization and/or restriction systems for anti-MRSA antibiotics and broad-spectrum antibiotics. Our results also suggest that surveillance activities for ventilator-associated pneumonia and catheter-associated urinary tract infections need to be increased.