Effects of infectious disease consultation and antimicrobial stewardship program at a Japanese cancer center: An interrupted time-series analysis

In cancer patients, appropriate diagnosis and management of infection are frequently challenging owing to subtle or atypical presentation. We investigated the effectiveness of infectious disease (ID) consultations and the Antimicrobial Stewardship Program (ASP) in a Japanese cancer center. This 36-month-period, single-institution, interrupted time series analysis was retrospectively conducted during April 1, 2018–March 31, 2021, to evaluate a two-phase intervention: Phase 1 (notification of antimicrobials by the infection control team) and Phase 2 (establishing an ID consultation service and implementing ASP). Among 32,202 patients hospitalized, 22,096 and 10,106 hospitalizations occurred at baseline and during intervention period, respectively. The Antimicrobial Stewardship Team (AST) provided feedback on specific broad-spectrum antimicrobials in 913 instances (347 appropriate [38%]; 566 inappropriate [62%]), and 440 ID consultations were completed, with a 75% overall acceptance rate for AST suggestions. In Phase 2, monthly carbapenem days of therapy (CAR-DOT) decreased significantly, and narrow-spectrum antibiotic usage increased significantly in both trend and level; monthly DOT of antipseudomonal agents decreased significantly in trend. The results of these analyses of antimicrobial use are consistent with the DOT-based data based on antimicrobial use density (AUD). The total number of inpatient specimens increased significantly; the trend of multidrug-resistant Pseudomonas aeruginosa and methicillin-resistant Staphylococcus aureus infections decreased, without changes in the incidence of other resistant organisms, all-cause in-hospital mortality, and length of stay. Actual and adjusted CAR purchase costs per patient-day decreased without significant changes in the actual and adjusted purchase cost per patient-day for all intravenous antimicrobials. Combining ID consultation and ASP reduced carbapenem use without negative patient outcomes. Their implementation could facilitate establishment of safe cancer treatment facilities in Japan and improve prognosis in cancer patients.


Introduction
In recent years, antimicrobial resistance (AMR) has become a global concern and the implementation of AMR-prevention measures has become an urgent requirement for medical institutions [1]. Although carbapenem-resistant Enterobacteriaceae (CRE) and carbapenemresistant Pseudomonas aeruginosa, which are increasing worldwide, account for <1% of drugresistant bacteria in Japan, the resistance rates of Escherichia coli to third generation cephalosporins and fluoroquinolones are showing an upward trend, and the rate of methicillin-resistant Staphylococcus aureus (MRSA) remains high [2]. Patients with cancer, especially those with hematologic malignancies or severe neutropenia and those who undergo hematopoietic stem cell transplantation, frequently develop serious infections and receive multiple rounds of antimicrobial therapy during the progression of their underlying disease [3,4]. In patients with cancer, appropriate diagnosis and management of infection are frequently challenging due to the subtle or atypical presentation of infections [5,6]; therefore, substantial amounts of antimicrobials are used in these patients [7]. Carbapenems (CAR) are commonly prescribed for infections in patients with malignancies. Physicians are reluctant to discontinue or de-escalate antimicrobial therapy within the optimal treatment period for these patients [4,8]. There is a consensus that antimicrobial use places significant antimicrobial pressure not only on the patient's normal flora, but also potentially on the surrounding environment, which significantly contributes to the development of resistant bacteria [9,10]. For example, patients with CRE infections are at high risk for adverse outcomes from invasive infections and can spread resistant genes within healthcare facilities. Carbapenemase-producing Enterobacteriaceae (CPE) outbreaks have been reported in cancer centers across Japan, complicating infection control and treatment of infected patients [11]. Most studies on the Antimicrobial Stewardship Program (ASP) were conducted in the United States or European countries, and few studies have evaluated the effectiveness of ASP in other regions [12]. While there are insufficient studies globally that have addressed the role of ASPs specifically in patients with cancer [7,8], there have clearly been more opportunities to provide guidance on appropriate antimicrobial prescription in patients with cancer than in most other patient populations and settings. However, there are no previous reports on ASP at cancer centers in Japan. Akazawa et al. [12] and Matono et al. [13] reported that an ASP intervention in a tertiary care facility in Japan reduced the use of broad-spectrum antimicrobials without negative outcomes. Thus, we conducted a retrospective study to assess the impact of the introduction of the Antimicrobial Stewardship Team (AST) and infectious disease (ID) consultation in a Japanese cancer center.

Secondary outcome measures
DOT for antipseudomonal agents. The total DOT per month per 100 patient-days was calculated for three antipseudomonal agents: piperacillin-tazobactam, cefepime, and cefozopran. This was because the aforementioned antibiotics are broad-spectrum antimicrobials, similar to CAR. The DOT for antipseudomonal agents was evaluated to assess whether CAR was simply being replaced with other broad-spectrum antipseudomonal agents.
DOT for ampicillin, ampicillin-sulbactam, cefazolin, and cefmetazole. The total DOT per month per 100 patient-days was calculated for four antimicrobial agents, namely ampicillin, ampicillin-sulbactam, cefazolin, and cefmetazole, as these antibiotics are narrow-spectrum antimicrobials after de-escalation. This parameter was assessed to confirm whether the intervention resulted in appropriate de-escalation.
Incidence of hospital-acquired resistant microorganisms/Clostridioides difficile infection (CDI), and candidemia. We measured the annual incidence of antibiotic-resistant bacterial infections, CDI, and candidemia per 1000 patient-days between April 2018 and March 2021, the period for which these study data were available, as an indicator of the outcome of the ASP. The resistant microorganisms included CPE, CRE, multidrug-resistant P. aeruginosa (MRPA), MRSA, and extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae. Hospital-acquired microorganisms were defined as those that were identified >72 h after D De-escalation Broad-spectrum antimicrobials were used based on the clinical characteristics, culture results, and local factors, and these can be changed to a narrow-spectrum antimicrobial.

E Escalation
The antimicrobials do not provide adequate coverage of the target microorganisms; therefore, the spectrum needs to be broadened or the antimicrobial should be changed.

F Fitting dose
The dose and method of administering the antimicrobials are inappropriate due to the renal function or other factors; thus, adjustments are necessary.

H Halt
Discontinuation is necessary because the purpose of antimicrobial administration has been achieved, further use is unnecessary, or there is a risk of allergy.

I Indication document
The purpose of use and the target microorganisms of the antimicrobials are not described in the medical record and, therefore, cannot be evaluated; additional descriptions are needed.
Time out T Time out Notify physician that culture results are available (3-5 days after initiation of the antimicrobial therapy) or that it is time to reconsider the termination of antimicrobial therapy (10-14 days after initiation of the antimicrobial therapy).
admission [15]. To exclude duplicates, when the same resistant microorganisms were detected more than once in the same patient, only the first specimen obtained in each month was included in the analysis [15]. However, if the resistant microorganism was detected in blood samples, this infection was defined as a new event if the same resistant bacteria had not been detected in blood samples from the same patient within the past 2 weeks [15]. Only clinical specimens of resistant microorganisms were included, and specimens for surveillance culture and negative confirmation were excluded. CDI was defined as the number of patients with positivity of CD toxin (C. DIFF QUIK CHEK COMPLETE; Alere Medical Co., Tokyo, Japan). Moreover, ESBL-producing Enterobacteriaceae, including E. coli, Klebsiella pneumoniae, Klebsiella oxytoca, and Proteus mirabilis were identified using the Cica β-test (Kanto Chemical Co., Tokyo, Japan) and disc diffusion method. CPE were identified using the modified carbapenem inactivation method, which was performed according to CLSI M100-S27 [16]. CRE was defined as either resistance to meropenem [minimum inhibitory concentrations (MIC) �2 μg/ mL] or resistance to both imipenem (MIC �2 μg/mL) and cefmetazole (MIC �64 μg/mL) [17]. MRPA was defined as P. aeruginosa that is resistant to CAR, aminoglycosides, and fluoroquinolones according to the standard set by the Clinical and Laboratory Standards Institute's for antimicrobial susceptibility tests, regardless of whether carbapenemase was produced or not [12]. Furthermore, we included the incidence of candidemia, although the blood samples intended for confirming negative candidemia were excluded from this analysis. Total number of inpatient specimens. To assess the impact of ID consultation and application of the ASP, we evaluated the total number of inpatient samples from April 2018 to March 2021.
Cost of CAR and all intravenous antimicrobials. To assess the economic impact of ASPs, we assessed the cost of purchasing antimicrobials per patient-day each year from April 2018 to March 2021. The adjusted purchase costs were calculated based on the actual purchase costs (considering the cost of switching from branded to generic products and changes in drug prices) and prices of branded agents in April 2020. The exchange rate of 1 USD to 108 JPY was used for the calculations in March 2021.
Assessment and acceptance rates of proposals by the AST. The evaluation (Table 1) and acceptance rates by the AST were calculated. Acceptance rate was the sum of the accepted and partially accepted suggestions divided by the total number of suggestions.
All-cause in-hospital mortality and length of hospital stay. Data for all-cause in-hospital mortality and length of hospital stay were extracted and included in the analysis.

Statistical analysis
To demonstrate the effect of each intervention on CAR-DOT, CAR-AUD, the DOT, and AUD for the three antipseudomonal agents (piperacillin-tazobactam, cefepime, and cefozopran) and four narrow-spectrum antibiotics (ampicillin, ampicillin-sulbactam, cefazolin, and cefmetazole) were ascertained. Furthermore, we carried out a segmented regression analysis of interrupted time-series studies for the two periods (Phase 1: April 1, 2018 to March 31, 2020; Phase 2: April 1, 2020 to March 31, 2021), when the DOT data were available. Trends and changes in the levels of the incidence of antibiotic-resistant bacterial infection, CDI, candidemia, mortality, and length of hospital stay were evaluated using a segmented regression analysis of interrupted time-series studies for the same periods. We adopted the most standard model based on the linear regression model, including level change and slope change. Autocorrelation (confirmed by the p-value of the Ljung-Box test and visualization of autocorrelation) was observed in the DOT for the antipseudomonal agents, incidence of CRE, MRPA, MRSA, candidemia, in-hospital mortality, and length of hospital stay. However, we did not adjust for autocorrelation since the number of time points was small (36 before and after the intervention) and we wanted to discuss the same model consistently for the primary outcome without autocorrelation issues. Therefore, this study was set as an exploratory (hypothesis-generating) research. A bivariate analysis for the cost of antimicrobials and the total number of inpatient specimens was carried out using the Mann−Whitney U test (continuous variables), with p<0.05 regarded as statistically significant. R software, version 4.0.2. (The R Foundation for Statistical Computing, Vienna, Austria) was used for all analyses.

Ethical considerations
This study was approved by the Institutional Review Board of the ACC hospital (approval number: 2020-1-640) and conducted according to the principles of the Declaration of Helsinki. The requirement of informed consent was waived because this study only used data that were collected in clinical practice.

Incidence of antimicrobial-resistant microorganisms, CDI, and candidemia
During the study period, CPE was not detected. The change in the level of the monthly incidence of MRPA did not decrease (coefficient: 0.09; 95% CI: −0.015 to 0.19, p = 0.10); however,

PLOS ONE
Effects of infectious disease consultation and antimicrobial stewardship program at a Japanese cancer center there was a significant reduction in the MRPA trend (coefficient: −0.02; 95% CI: −0.02 to − 0.004, p = 0.02); Fig 4). The level of the monthly incidence of MRSA did not decrease

Total number of inpatient specimens
Regarding the total number of patients, the number of samples increased significantly after the intervention (median: 574.5 vs. 669.5, p<0.001). Table 2 shows the actual and adjusted CAR purchase costs for all intravenous antimicrobials in each year. The actual CAR purchase cost per patient-days significantly decreased after the initiation of Phase 2 of this program (p<0.001). However, there was no significant change in the actual cost per patient-days of all intravenous antimicrobials that were purchased (p = 0.83). The adjusted CAR purchase cost per patient-days significantly decreased after the

Assessment and acceptance rates of proposals by the AST
Based on the evaluation by the AST, we determined that there were 347 appropriate (38%) and 566 inappropriate (62%) instances of antimicrobial use (Table 3). "De-escalation" was the commonest type of inappropriate use (42%). The overall acceptance rate of the AST suggestions was 76%.

Discussion
This is the first study to report on the effects of ID consultation and ASP interventions in a cancer center in Japan. Cancer centers constitute challenging facilities for conducting antibiotic stewardship interventions due to the complexity of patients conditions, on account of the underlying cancers and hematologic malignancies, neutropenia, the various departments that are involved in the multidisciplinary patients' management, and the varying departmental The actual cost includes the cost of switching to generic drugs and factors in the changes in the drug prices.
The adjusted cost is calculated based on the drug price in April 2020.
https://doi.org/10.1371/journal.pone.0263095.t002 Table 3. Content and acceptance rate of feedback by the AST with regard to specific broad-spectrum antimicrobial usage. structures and internal guidelines [7]. However, our intervention was effective in reducing CAR consumption without increasing the consumption of alternative broad-spectrum antimicrobials or resulting in other negative outcomes. In our study, the level and trend of CAR-DOT was significantly decreased, and only the trend of the DOT for the three antipseudomonal agents was significantly decreased. The levels of the three antipseudomonal agents did not decrease significantly, but the results suggest that a decrease can be expected in the future if significant decrease in trend is sustained. There was no significant decrease in the level of the three antipseudomonal agents, suggesting that a decrease might be expected with long-term observation. This finding might have been due to the routine use of aggressive chemotherapy with febrile neutropenia in cancer centers [18]. This study also evaluated monthly DOT trends for the narrow-spectrum antimicrobials to confirm de-escalation from broad-spectrum antimicrobial usage. The use of these drugs has not been previously investigated [8,12]. In our study, the level and trend of CAR-DOT and the trend of DOT for antipseudomonal agents was decreased, whereas the level and trend of the narrow-spectrum antimicrobials was increased. The results of the analysis between antimicrobial use using AUD and DOT were consistent. Although there were no significant differences, the length of hospital stay and mortality rates tended to decrease and are expected to decrease with long-term observation. These results of no increase in mortality and hospital stay, suggest that antimicrobial agents were being used appropriately. For more accurate assessment in future, it is thought that mortality due to ID should also be evaluated, but individual patient's data were not evaluated in this study. Discontinuation or de-escalation of antimicrobial agents in patients with malignancy is often avoided [4,8]; however, with ID consultation and ASP, they can be implemented safely without negative patient outcomes.

Number of evaluations (n) Suggestions (n) Accepted (n) Partially accepted (n) Not accepted (n) Acceptance rate a (%)
The main concept of ASPs in any patient population, including cancer patients, is to facilitate antimicrobial use such that each patient receives the most effective and safest antimicrobials for treating their infection while minimizing the ecological impact of the antimicrobials used [1]. Given the high frequency of infection in cancer patients and the need for both therapeutic and prophylactic use of antimicrobials, the magnitude of antimicrobial usage is substantial. This imposes a significant antimicrobial pressure not only on the patient's normal microbiota, but also potentially on the surrounding environment. An indirect consequence of increased antimicrobial use, for example, is a higher prevalence of CDI in this population [19]. Moreover, broad-spectrum antimicrobial use increases the risk for candidemia [20]. Furthermore, the greatest concern associated with increased antimicrobial usage is the emergence of infections caused by drug-resistant bacteria [7]. In our study, there was a significant decrease in the trend of MRPA and MRSA, although there was no change in the number of infections caused by other drug-resistant bacteria. There was no significant difference between candidemia and CDI; however, a decreasing trend was observed. This might have been due to the short observation period of our study, which implied that a marked decrease can be expected with long-term observation in future evaluations [8,13]. In particular, as the trend of MRPA and MRSA incidence was significantly reduced in our study, additional observation for a short-term period is expected to further reduce the level of MRPA and MRSA incidence. Moreover, the total number of inpatient specimens were significantly increased during the study period. Although there were no significant differences between CRE and ESBL-producing Enterobacteriaceae, an increasing trend was observed, which might have been influenced by an increase in the total number of samples. Especially in Japan, the incidence of cefotaxime resistant E. coli and K. pneumoniae, which are suspected to be ESBL-producing, has been increasing every year [2]. No in-hospital outbreaks of CRE or ESBL-producing Enterobacteriaceae were observed during the study period. Therefore, the detection rate of resistant bacteria that was possibly underestimated before the intervention might have edged closer to the true detection rate, and this factor might have affected our results.
Our interventions began in April 2020, but changes have been occurring in the use of carbapenems and the three anti-pseudomonal agents since late 2019. This was partly due to the impact of a one-hour lecture on the appropriate use of antimicrobials to healthcare providers by an external ID expert on February 19, 2019.
In our study, the total number of samples was increased after the intervention. This is an important finding for the promotion of diagnostic stewardship. Diagnostic stewardship refers to the performing of the appropriate test on the appropriate patient at the appropriate time, to provide optimal clinical care [21]; it is an integral part of ASP [22]. At the beginning of our intervention, the ASP was impeded by the insufficient number of specimens submitted; however, as the total number of specimens increased, the ASP could be conducted appropriately. Conversely, without active ASP intervention, the diagnostic stewardship program may be ineffective [23], and a strategy that combines both is needed to optimize the care of patients with cancer.
Our results showed that the decrease in CAR-DOT due to ID consultation and ASP intervention led to a significant reduction in the cost of CAR. While the use of alternative antimicrobials was increased, the overall cost of antimicrobials did not change significantly after adjustment, indicating that our intervention did not impose an economic burden on the hospital.
The acceptance rate for our ASP was as high as 76% in this study. Other studies have reported similarly high acceptance rates of ASP (76-87%) [8,13]. The high compliance rate suggests that the results of our study are likely positive. Moreover, the high compliance rate could be attributed to the approval by the hospital director who informed all the institution's medical staff about the importance of the intervention and about antimicrobial resistance, before the intervention started. Additionally, this may be because the ASP and ID consultations resulted in appropriate interventions and management of cases with difficult diagnoses and treatments. Most cancer centers in Japan do not have cross-organ departments such as the Department of General Internal Medicine; the Department of ID plays a part in this role [5]. It is possible that these activities increased the satisfaction of the oncologists; consequently, they became more compliant with ASP.
There is scarcity of ID physicians in Japan. Ideally, Japan needs 3,000-4,000 ID physicians; however, as of August 2021, there were only 1622 ID physicians (i.e., 1.29 per 100,000 population in Japan vs. 2.78 per 100,000 in the United States) [24][25][26]. Furthermore, the training of ID physicians, pharmacists, clinical laboratory technicians, and infection control nurses in cancer centers is often time-consuming, since the complex background of patients with cancer requires sufficient experience and knowledge of ID and oncology. Therefore, the lack of appropriate human resources at specialized facilities such as cancer centers is one of the major barriers to implementing ASP. However, our study showed a positive effect of the ASP with the involvement of fewer human resources, which suggests that in cancer centers in Japan, the ASP would be a highly effective activity that can be sufficiently effective even when few specialists are involved.
This study has several limitations. First, as this was a single-center study at a Japanese cancer center, it is unclear whether the hospital-based ASP strategy is generalizable. Second, the impact of broad-spectrum antimicrobial prescribing for patients with coronavirus disease (COVID-19) and for hospitalized patients suspected to have COVID-19 during the COVID-19 pandemic needs to be considered [27], since broad-spectrum antimicrobials have been used both prophylactically and therapeutically for such patients [27,28]. Although our institution was affected by the COVID-19 pandemic, the trends of broad-spectrum antimicrobial prescribing declined. Third, in 2019, a nationwide cefazolin shortage occurred in Japan due to manufacturing-related problems at a major production facility [29]. Although the cefazolin shortage led to a significant increase in the use of alternative antimicrobials at several facilities in Japan [29], our data trends from 2018 to 2019 showed a decline in broad-spectrum antimicrobial use in 2020. Additionally, there was a supply restriction of cefepime from June 2018 to October 2019; nonetheless, broad-spectrum antimicrobial use declined in 2020 compared to the usage trends in 2018 and 2019. To resolve these limitations, a long-term multicenter study including other Japanese cancer centers is warranted in the future.
In conclusion, the combination of ID consultation and ASP in this study reduced the use of CAR without negative impacts on the outcome-related parameters, such as mortality rates and length of hospital stay. The introduction of these systems may lead to the establishment of safer cancer treatment and improve the overall prognosis for patients with cancer in Japan. Investigation: Naoya Itoh.