Impact of an Infectious Disease Specialist-Led Postprescription Carbapenem Intervention on Prescribing Behavior in a Japanese Tertiary Hospital: a Before-After Study

Background: There are a few infectious disease (ID) specialists in Japan. A solution to promote the implementation of antimicrobial stewardship activities under these resource-limited settings is required. Methods: We compared carbapenem consumption between a 24-month baseline and 12-month intervention periods. During the intervention period, an ID specialist provided daily advises through a chart review and ID consultation service to prescribers against all prolonged carbapenem use ( ≥ 14 days). Additionally, we send an aggregated table containing the weekly point prevalence amount of each departmental carbapenem use for 7–13 and ≥ 14 days through e-mail to all doctors. Results: Among the 1,241 carbapenem courses during the intervention period, the ID specialist provided a total of 96 instances of feedback regarding carbapenem use for ≥ 14 days, and the feedback acceptance rate was 76%. After the initiation of the intervention, the trend in monthly carbapenem consumption changed (coecient: -0.62; 95% CI: -1.15 to -0.087, p = 0.024), and its consumption decreased (coecient: -0.098; 95% CI: -0.16 to -0.039, p = 0.002) without an increase in the consumption of broad-spectrum antimicrobials or in-hospital mortality. Interestingly, the monthly number of carbapenem courses, but not the duration of carbapenem use, signicantly decreased (coecient: -3.02; 95% CI: -4.63 to -1.42, p = 0.001). The carbapenem-related annual estimated savings after the intervention was $83,745, with a 22% cost reduction. Conclusions: Our ID specialist-led daily intervention with weekly feedback regarding long-term carbapenem use was effective in reducing antimicrobial may be resource-limited


Background
Antimicrobial resistance is a serious global health threat, and the annual antimicrobial resistance-related death is predicted to be 10 million people in 2050 [1]. Given this threatening substantial issue, reducing the use of inappropriate antimicrobials is an urgent task. In acute care hospital, approximately 20-50% of overall antimicrobials are estimated to be unnecessary or inappropriate [2], and antimicrobial stewardship programs (ASPs) are advocated as key elements in optimizing antimicrobial use in healthcare settings [3]. Among various ASPs, the importance of infectious disease (ID) specialist-led ASPs has been highlighted in recent years [4], involving preprescription authorization (PPA) and postprescription review with feedback (PPRF) as core strategies of ASPs [3,5]. However, in Japan, there are only a few board-certi ed ID specialists (1.18 per 100,000 people in Japan versus 2.56 per 100,000 people in the USA) [6][7][8]; therefore, a lack of human resources is one of the major barriers for the implementation of these evidence-based ASPs. Further, although the Japanese government issued a national action plan on antimicrobial resistance in 2016, only 14.6% of overall Japanese medical institutions, including 41% of hospitals with > 500 beds, had an approved formal antimicrobial stewardship team (AST) in 2018, and most staff members were not full-time equivalent (FTE) [9]. Thus, at present, the Japanese guidance for implementing ASPs has recommended the PPRF strategy instead of the time-consuming PPA [10].
In Japan, third-generation cephalosporin-resistant Enterobacteriaceae has increased in national surveillance data [11]. In our hospital, third-generation cephalosporin-resistant Escherichia coli has been recently identi ed in approximately 30% of total E. coli cases; therefore, carbapenem consumption increased from a median of 2.19 de ned daily doses (DDDs) per 100 patient-days in 2012 to 3.73 DDDs per 100 patient-days in 2016. Despite this alarming increase, we have insu cient protected time to devote to the implementation of PPA and early PPRF, and thus, there is a need for a solution. Hence, the aim of this study was to assess the effectiveness of an ID specialist-led, hospital-based ASP against prolonged carbapenem use in resource-limited settings using weekly feedback and daily intervention.

Study design and setting
This uncontrolled before-after study was designed to compare carbapenem consumption in a 24-month baseline period (from August 2015 to July 2017) to that in a 12-month intervention period (from August 2017 to July 2018). The study was conducted at the Iizuka Hospital (AIH), a tertiary care hospital with 1048 inpatient beds and an adult/neonatal intensive care unit in Fukuoka, Japan. The data for this study were obtained via a medical database at the AIH linking a microorganism database in the microbiology laboratory, a prescription database in the pharmacy department, a patient database from the AST, and a chart review. This study was approved by the Institutional Review Board at the AIH (approval no. 17183) and conducted according to the principles expressed in the Declaration of Helsinki. The need for informed consent was waived because this study only used data collected from clinical practice.

Feedback and intervention
AST members included a physician, clinical pharmacist, nurse, and microbiology laboratory technician who performed ASP duties since April 2015. We restricted antimicrobial prescription of daptomycin, linezolid, and tedizolid as PPA but did not restrict against carbapenem prescription. Every weekday during the baseline period, two ID-trained pharmacists performed a prospective audit for all carbapenem prescription from the initiation to discontinuation of the prescription with a FTE of 1.1. They predominantly provided dosing optimization feedback to carbapenem prescribers through the patient charts. Moreover, they recorded a template document in the patient charts on the day of carbapenem prescription initiation and after use for 7 days, recommending the proper use of antimicrobials. Indication and duration of carbapenem use was up to the prescribing doctors because their feedback contained neither an alternative prescription choice nor a suggestion for discontinuation.
During the intervention period, we additionally initiated an intervention by a board-certi ed ID specialist against all carbapenem use for ≥ 14 days. The ID specialist provided daily ID consultation through faceto-face or telephone conversations with prescribers and recorded the recommendation in the patient's chart. In addition, we reported an aggregated table involving the weekly point prevalence surveillance indicating the amount of departmental carbapenem use for 7-13 days and ≥ 14 days through e-mail to all medical doctors at the AIH.

Outcome measure and de nition
The primary endpoint of this study was the trend in monthly carbapenem consumption. Secondary endpoints were 1) change in each intravenous antimicrobial consumption which was available at the AIH, and 2) the trend in the detection of hospital-acquired microorganisms/infections including methicillinresistant Staphylococcus aureus (MRSA), extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae, carbapenem-resistant organisms (CROs), Candida spp. bloodstream infections, and Clostridioides di cile infections (CDI). Additionally, we extracted data for all-cause in-hospital mortality and length of hospital stay from the medical database at the AIH.
Monthly aggregated hospital antimicrobial consumption was measured as days of therapy (DOTs) per 100 patient-days and DDDs per 100 patient-days. The incidence of microorganism detection was calculated as the number of clusters per 1000 patient-days. A microorganism recovered from a previously enrolled patient (e.g., continuous bacteremia) within the same month was excluded. Hospital-acquired microorganism/infection was de ned as a microorganism/infection identi ed > 48 hours after admission. ESBL-producing Enterobacteriaceae, including Escherichia coli, Klebsiella pneumoniae, Klebsiella oxytoca, and Proteus mirabilis was identi ed using the Cica β-test (Kanto Chemical Co., Tokyo, Japan) and the disc diffusion method. CRO was de ned as an imipenem-resistant organism based on the Clinical Laboratory Standard Institute (CLSI) M100-S29. CDI was de ned as a patient with diarrhea and a stool specimen positive for glutamate dehydrogenase plus toxin assay (C. DIFF QUIK CHEK COMPLETE from 2015-2017; Alere Medical Co., Tokyo, Japan, and GE Test Immunochromato-CD GDH/TOX from 2017-2018; Nissui Pharmaceutical Co., Tokyo, Japan).
The acceptance rate was de ned as the number of feedback accepted by prescribers divided by the total number of feedback made by an ID specialist. Acceptance or nonacceptance was decided by chart review within 24 hours after the recommendations were made. A carbapenem-related annual cost reduction and estimated savings were calculated based on the annual cost of purchasing carbapenem. The cost of purchasing carbapenem was adjusted to the cost of branded meropenem in 2018 considering uctuations in drug prices and cost differences between brands and generics. The exchange rate of 1 USD to 110 JPY in 2019 was used.

Statistical analysis
We used segmented regression analysis of an interrupted time series to evaluate trends in monthly antimicrobial consumption calculated by DOTs per 100 patient-days and trends in the incidence of detected microorganisms, which was represented as a change in level due to the initiation of the intervention and change in trend between two periods [12]. The monthly antimicrobial consumption using DDDs per 100 patient-day, number of detected hospital-acquired microorganisms/infections, all-cause inhospital mortality, and length of hospital stay were compared between the two periods using the chisquared (χ 2 ) test for nominal variables and Mann-Whitney U test for continuous variables. A two-tailed pvalue of < 0.05 was regarded as statistically signi cant, and the 95% con dence interval (95% CI) was used. All analyses were performed using Stata/SE v. 14.2 (StataCorp, College Station, TX, USA).

Impact of the intervention on carbapenem consumption
During the study period, 90,076 patients were admitted to the AIH, including 59,195 in the baseline period and 30,881 in the intervention period. Out of these, 3,935 patients received carbapenem (meropenem and doripenem) injection: 2,694 in the baseline period and 1,241 in the intervention period. Among the 1,241 courses of carbapenem in the intervention period, an ID specialist provided a total of 96 instances of feedback against carbapenem use for ≥ 14 days. The predominant content of the feedback was carbapenem discontinuation (n = 45), followed by further investigation (n = 25), and change to other antimicrobials (n = 22) ( Table 1). The overall acceptance rate of the ID specialist's feedback was 76%, and the ID specialist devoted an average FTE of 0.16 to the intervention. After the initiation of the intervention, the trend in monthly carbapenem consumption (DOTs per 100 patient-days) changed in level (coe cient:  The interrupted time series analysis showed that the trend in monthly piperacillin-tazobactam consumption temporarily increased after the initiation of the intervention, but the level change was not signi cant (coe cient: 0.25; 95% CI: -0.30 to 0.79, p = 0.36, Fig. 3). Thereafter, the trend in piperacillin-

Discussion
In this before-after study, we evaluated the impact of an ID specialist-led daily postprescription carbapenem intervention with weekly feedback for long-term carbapenem users. Our intervention was effective in reducing carbapenem consumption without causing a change in in-hospital mortality or an increase in the consumption of alternate broad-spectrum antimicrobials. We found that the number of courses of carbapenem, but not the duration of carbapenem use, decreased, indicating a possible bene cial effect of our intervention on improving prescribing behavior.
The present study has three important ndings. First, it is noteworthy that our study demonstrates the positive effect of our ASP even in a resource-limited setting. There is limited information on in-hospital ASP in Asia, and resources and settings are diverse among each country and hospital [9,13,14]. In Japan, the resources were generally insu cient to meet the demands for implementing an ideal ASP [15], and therefore there is a necessity in developing ID-trained pharmacists and ID specialists [9]. Although PPRF is a well-established intervention, there are few resource-rich Japanese hospitals implementing early PPRF against overall antimicrobial use for 48-72 hours. Therefore, we must make a practical plan to implement the stewardship according to available institutional resources: e.g., focusing on broadspectrum antimicrobials and/or longer duration of use, and providing once-weekly intervention [16][17][18][19]. In our hospital, the number of ID physicians was particularly insu cient; there were no ID fellows, and one (part time) and two (full time and part time) ID specialists worked in the baseline and intervention periods, respectively. As a result, only an FTE of 0.16 was devoted to the additional stewardship activity by an ID specialist; however, it is noteworthy that our strategy had a positive effect on reducing carbapenem consumption. Hence, we highlight the positive impact of our ASP strategy, which may be a reasonable and acceptable alternative, particularly in resource-limited settings.
Second, our study indicates that a predominant factor in the decrease of carbapenem consumption might be in behavior change among prescribers. In our study, ESBL-producing Enterobacteriaceae strains were more frequently identi ed in the intervention period than in the baseline period, a trend similar to that seen in national and local surveillance data in Japan [11]. Therefore, it is meaningful that carbapenem consumption decreased in the intervention period, despite being one of the rst-line agents for ESBLproducing Enterobacteriaceae. Interestingly, although we did not provide alternate empirical antimicrobial choices or restrict carbapenem prescription, empirical carbapenem use decreased, which suggests that our intervention was indirectly or comprehensively relevant to changing prescribing behavior. Prescription trends of all available intravenous antimicrobials at our facility showed that the monthly trend in piperacillin-tazobactam consumption increased only temporarily before decreasing, but the change in level was not signi cant after the initiation of the intervention. There was no compensatory increase in the other antipseudomonal agents including fourth-generation cephalosporins, uoroquinolones, and aminoglycosides (Table S1). Further, carbapenem-sparing agents such as ceftazidime-avibactam and ceftolozane-tazobactam were not available in our hospital during the study period. As a result, the comprehensive evaluation enables us to know that our intervention might promote optimal use of antimicrobials-that is, reduce unnecessary empirical carbapenem use and using narrow-spectrum antimicrobials such as rst-and third-generation cephalosporins-without a change in in-hospital mortality. Thus, we emphasize that not only a reduction in carbapenem consumption but also the positive impact of ASPs on optimizing the use of antimicrobials; the concepts supported by several previous reports [19,20].
Third, our study suggests that, in addition to intervention made by the ID specialist, several factors comprehensively affected the reduction in carbapenem consumption and optimal use of antimicrobials.
We only provided direct intervention for 96 (7.7%) of 1,241 courses of carbapenem, and the duration of carbapenem use was not in uenced by this. Similar to our weekly feedback, weekly e-mail-based noti cation alone failed to show a positive impact on reducing carbapenem consumption in a previous study [19]. However, by adding the weekly feedback to the daily intervention on carbapenem use for ≥ 14 days, synergistic positive effects and a competitive mindset among departments may occur. Further, a hospital president approved our intervention and announced the importance of the intervention and information on antimicrobial resistance to all medical staff at the facility before the initiation of our intervention. The credit and education might yield a further positive in uence on growing awareness problems on antimicrobials and changing prescribing behavior. Moreover, leadership of ID consultation services changed from hospitalists to the ID specialist since April 2017 (four month before the initiation of our intervention). We conducted several in-hospital noti cations to increase the use of ID consultation services. The campaign and daily ID consultation service without ASP activities might achieve a positive impact on optimal use of antimicrobials. We believe that, in addition to daily intervention, the weekly feedback, president's credit, education of medical staff, and campaigns of using ID consultation services all comprehensively and synergistically contributed to a reduction in carbapenem consumption and optimal use of antimicrobials. Hence, we believe that taking action is important without giving up ASP activities in resource-limited settings.
This study has several limitations. First, this is a single-center study in Japan, and thus, it is uncertain whether our hospital-based ASP strategy can be applicable worldwide. Schweitzer et al described that only 48% and 23% of studies evaluating antimicrobial stewardship interventions reported clinical and microbiological outcomes, respectively [21]. However, we evaluated these outcomes, and it is noteworthy that our ASP strategy led to a decrease in carbapenem consumption with prescribing behavior change in a resource-limited setting. Second, we could not investigate the length of hospital stay among patients with infectious diseases and infectious disease-related total medical cost reduction. However, we would like to emphasize the impact on cost reduction that our ASP accomplished on overall intravenous antimicrobial-related saving of $255,352 in addition to annual estimated carbapenem-related savings ($83,745) on the basis of the annual purchasing cost. Third, there is evidence that the implementation of hospital-based ASP had a positive in uence on microbiological outcomes [13,22,23]; however, our study did not show the positive microbiological impact on the hospital-acquired multidrug-resistant pathogens and infections. This may be because of the insu cient number of events and short observation period. Further, several studies showed a synergistic effect of hand-hygiene improvement combined with ASP on microbiological outcomes in the hospital [14,24]. Thus, we should promote the implementation of hospital-based ASP with hand-hygiene intervention as well as a local-regional ASP to improve microbiological outcomes, and continue further outcome evaluation.
This before-after study showed the positive impact of an ID specialist-led ASP on reductions in carbapenem consumption and antimicrobial-related cost without adversely affecting the increase of broad-spectrum antimicrobial consumption and in-hospital mortality. We believe that an important path to success in hospital-based ASP is investigating institutional/local issues of antimicrobial prescription and making a practical plan according to available institutional resources. In particular, the favorable and reliable relationship between prescribers and the AST is vital in bringing about high compliance and prescribing behavior change among prescribers, which can nally lead to optimal use of antimicrobials. Even in resource-limited settings, we advocate that taking any action is warranted even if it is not PPA and early PPRF.   Supplementary Files