Impact of a comprehensive antimicrobial stewardship program on antimicrobial utilization and clinical isolates distribution in ICUs

Objective This study aimed to evaluate the effect of a comprehensive antimicrobial stewardship program (ASP) and provide clinical evidence for the scientific stewardship of antimicrobials in intensive care units (ICUs) of a teaching hospital. Methods Between January 2013 and December 2018, we conducted a prospective study, based on an antimicrobial computerized clinical decision support system (aCDSS) deployed in 2015 in ICUs of a tertiary and teaching hospital. The primary outcomes included initial and overall use prevalence of antimicrobials. The second outcomes were the detection rate of common clinical isolates before and after therapeutic antimicrobial use, and the change in patterns of resistance of 5 common clinical isolates in the ICU.


Introduction
Antimicrobial resistance (AMR) is a global public health threat, especially in hospitalized patients with substantial health and economic consequences. The choice of antimicrobials for the treatment of multidrug-resistant organism (MDRO) infections have become increasingly limited and expensive [1,2]. Inappropriate or sub-optimal utilization of antimicrobials over the past few decades has driven increases in rates of resistance, with the consequent effects of worse patient outcomes [3], undermining various aspects of health care and creating new disease burden [4,5]. Thus, optimizing antimicrobial therapy is of great importance since the link between inappropriate use of antimicrobials and the emergence of AMR has been well established, particularly in intensive care units (ICUs) where broad-spectrum antimicrobials are frequently prescribed [6,7] and it turns out that 30%-60% of them are unnecessary, inappropriate, or suboptimal [8][9][10][11].
Antimicrobial stewardship (AMS) involves a multifaceted approach to optimize antimicrobial use, minimize unintended consequences and combat the emergence of AMR [12][13][14][15][16][17]. The guidelines for the implementation of antimicrobial stewardship program (ASP) were published in 2007 and updated in 2016 by the American Society of Infectious Diseases (IDSA) and the Society for Healthcare Epidemiology of America (SHEA) [18,19].
Although ASP in China strived to improve rational use of antimicrobials and indeed 4 achieved some success, gaps still existed [20], especially in the ICU where patients often needed urgent effective treatment while antimicrobial resistance had been identified as a serious problem [21], the obstacles to AMS were significant in this vulnerable and complex patient population [22]. Point prevalence surveys (PPS) in 53 countries around the world showed that only 20% of adults hospitalized patients receiving therapeutic antimicrobials were infected with identified pathogens [23], targeted antimicrobial treatment is facing difficulties all over the world. Considering that the use of antimicrobial agents can affect the distribution and drug resistance of clinical isolates in medical institutions significantly [24][25][26], timely acquisitions of appropriate microbiological specimens before antimicrobial use is critical [27]. Clinical decision support system (CDSS), as one of the components of ASPs, serving to integrate multiple variables and allow physicians to make more appropriate treatment decisions. Computerized systems appear to be logical choices to maintain ASP competence and have become increasingly appealing because of the growing migration of other clinical systems to computerized platforms.
We developed and deployed an interactive, web-based computerized antimicrobial clinical decision support system (aCDSS), allowing and assisting stewards to obtain more reliable microbial outcomes on distribution and resistance patterns of clinical isolates and assess antimicrobial use, providing facility-level epidemiological data and illustrated trends in antimicrobial use. Insufficient studies describing the long-lasting impact of ASP in ICUs in a real-life setting have been documented so far [28][29][30][31][32], thus, our primary objective of this investigation was to describe the implementation and the effects of a comprehensive ASP focusing on CDSS in ICUs of our hospital during 6 consecutive years.

Materials And Methods
Design and Setting 5 A 6-year prospective study was conducted at the Second Affiliated Hospital of Zhejiang University School of Medicine (SAHZU), a tertiary care and Joint Commission International   (JCI) accredited hospital, with 133-bed combined ICUs. During the study period from   January 1, 2013 to December 31, 2018, all patients admitted to ICU for more than 24 hours and prescribed antimicrobials were eligible for inclusion.

Interventions
Hospital-wide AMS strategies had been implemented under the guidance of a crosssectoral committee consisting of infectious-disease-trained physicians, infection control practitioners, pharmacists, and clinical microbiologists. Our ASP was employed since Jan.
2015 which is a bundle of interventions including basic management measures and strengthening interventions aimed to facilitate the appropriate use of antimicrobials and combat AMR in health care settings. Besides, our ICUs established robust infection prevention and control initiatives to curtail onwards transmission of AMR microbes [33], and practices did not change appreciably during the study period.
Basic strategies focused on feedback & monitoring, coupled with weekly review and multidisciplinary team discussion of antimicrobial regimens in randomly selected patients.
Feedback was delivered by quarterly reporting of aggregate microbiological data in ICUs.
In addition, mandate hospital-wide educational modules were made on principles of appropriate antimicrobial use, strategies of related policies and the mechanisms of AMR, for all healthcare professionals involved in antimicrobial ordering, administration and monitoring through online and cased-based learning.
Prior to the study, SAHZU had adopted an electronic medical record system (EMRS) integrated with computerized physician order entry (CPOE) system which allowed submitting medication, laboratory, and radiology orders electronically. In nursing care, a mobile device was used for bedside medication and laboratory orders reconciliation. Data 6 on the timing of medication administration and microbiological specimen sampling were routinely recorded by the device, which was later transferred and stored in the nursing recording System (NRS). We began to develop the aCDSS, a web-based administrative database in late 2012. It was embedded in the EMRS and could access the hospital information system (HIS), EMRS, NRS, laboratory information system (LIS), pharmacy information system (PIS), surgical anesthesia information management system (AIMS) and ii. Pre-audit of indications for broad-spectrum antimicrobials: when prescribed a broadspectrum antimicrobial agent including β-lactam and β-lactamase inhibitor 7 combinations, carbapenems, tigecycline and so on, EMRS would pop up an electronic application form for the use of the corresponding antimicrobial agent after the prescription authority was passed, then the prescriber was asked to select the preset indication according to the actual situation of the patient before submitting the medical order (see Table 1). Stewardship rounds were conducted in ICUs, if there was any difficulty or an antimicrobial prescription modification was recommended by the ASP team, a face-to-face conference was held between team members and attending intensivists to decide the ultimate treatment; iii.
Online clinical decision support: all the diagnostic testing results were communicated to healthcare providers and provided interpretive expertise by the system in real time, assisting prescribers to review antimicrobial prescriptions and help to make preliminary recommendations based on infectious diseases resources; iv.
Intervening in combination therapy: verification and confirmation of indications for combination therapy were provided when prescribed two or more antimicrobials (see Table 2).

v.
Standardizing microbiological specimen sampling practice: the aCDSS ensured timely microbiological specimen sampling for microbial culture before therapeutic antibiotic use in inpatients. When ordered antimicrobials through the CPOE, the system automatically detected patient's surgery information through accessing EMRS. If no surgery information was identified, the aCDSS judged the ordered antibiotic as therapeutic use; otherwise, prescribers should choose the intended indication. In the next step, prescribers were requested to select the infection site on the software interface. The aCDSS detected patient's data on microbiological specimen sampling time through accessing NRS. In cases no data were identified, the prescriber could only quit the aCDSS, submit a laboratory order for bacterial culture through the 8 CPOE, then obtain the specimen and scan the barcode on the specimen to generate the sampling data (see Figure 1).

Electronic application form of the indication of the use of carbapenems
The clinical isolates are only sensitive to the same class of antimicrobials (note the exclusion of contaminated or accompanying strain) The currently used target antimicrobial was ineffective, and exclude other factors that lead to ineffective antimicrobial therapy Initial empiric therapy for severe infections (must be used after qualified microbiological specimens sampling) Receiving high-dose corticosteroids for more than 10 days White blood cell count ≤1000/mm 3 for more than one week The absolute value of neutrophils is ≤500/mm 3 for more than one week.
The number of CD4+ T lymphocytes in HIV-infected patients is ≤400/mm 3

Congenital immune deficiency
After splenectomy

Other choices
Recommended by authoritative guides

Recommended by multidisciplinary discussion
Recommended by a physician with the corresponding antimicrobial prescription authorization Table 1 The schematic  Table 2 The schematic table of indications of combination therapy.

Data Collection and Outcome Assessment
The aCDSS consolidated data on patient-specific demographics, surgery information, antimicrobials prescribed and related administration time, microbiological specimen sampling time, microbial culture results from EMRS, HIS, NRS and LIS (see Figure 2).

Resistant rate of 5 common bacteria including Staphylococcus aureus, Escherichia coli,
Klebsiella pneumoniae, Pseudomonas aeruginosa, and Acinetobacter baumannii was defined as the percentage of isolates resistant to selected antimicrobials. Duplicated isolates cultured from the same specimen from the same inpatient were not included for analysis. Antimicrobial use prevalence was calculated by dividing the number of cases prescribed specific antimicrobial over the total number of cases in the defined period.
Antimicrobial regimens were analyzed within the initial 24h and throughout the whole ICU stay.
The primary outcomes included: (i) initial and overall use prevalence of antimicrobials.

Primary outcomes
Various types of broad-spectrum antimicrobial use prevalence continued to increase from 2013 to 2015, since 2016, where initial use of carbapenems and glycopeptides were counterbalanced by an increase in use of the first/second-generation cephalosporins, βlactam and β-lactamase inhibitor combinations and linezolid, which called "squeezing the balloon" [34] (see Table 3). Further analysis showed that from 2015 to 2018, the proportion of extended-broad spectrum antimicrobials alone, wide-coverage therapy and combination therapy decreased significantly (P < 0.05), the maximum percentage of declines were 63.4%, 13.9% and 42.9%, respectively (see Fig. 3). Similarly, where use of carbapenems, glycopeptides, third/fourth-generation cephalosporins and anti-fungi agents were counterbalanced by an increase in overall use of the first/second-generation cephalosporins and β-lactam and β-lactamase inhibitor combinations (see Table 4). and Moraxella catarrhalis were significantly higher before antimicrobial treatment (P < 0.05), while Acinetobacter baumannii, Burkholderia cepacia, and Candida spp were significantly lower(P < 0.05) (see Fig. 4).

The antimicrobial resistance profiles of 5 major clinical isolates in 4 consecutive years
showed that in addition to Pseudomonas aeruginosa, the resistant rates of strains detected before antimicrobial treatment including Staphylococcus aureus, Escherichia coli, Klebsiella pneumoniae, and Acinetobacter baumannii were significantly lower (P < 0.05), more so in Klebsiella pneumoniae, which even reduced by approximately 2/3 (see Table 5).

Discussion
This study, over a 6-year period and in a real-life setting, analyzed the impact of antimicrobial use on microbial outcomes, along with a comprehensive ASP focusing on aCDSS, feedback & monitoring, weekly review of antimicrobials and multidisciplinary team discussion of antimicrobial regimens and education on antimicrobial use. Measuring the impact of an ASP is crucial to maintain administrative support and we provided real-world 14 experience from established ASPs in critical care settings. The aCDSS consists of several components including intervention in microbiological specimen sampling practice, hierarchical stewardship, education resources, and antimicrobial prescription recommendations. The design of this practice-based study was not to specifically measure the impact of any individual initiative but rather an integrated program in critical care.
We used data from academic hospital ICUs covering 6 years. As one of the wards with highest antimicrobial consumption, ICU is a gathering place for immunocompromised patients, and the nosocomial infection prevalence is much higher due to wide-spread use of broad-spectrum antimicrobials, glucocorticoids, immunosuppressive agents, and various invasive procedures than that of the ordinary wards. An international PPS indicates that 51% of the 1,265 ICUs in 75 countries around the world have infections identified and 71% are prescribed antimicrobials [35]. On one hand, a prompt institution of effective antimicrobial regimens for proper coverage of causative pathogens is vital in critically ill patients and involves physicians with infectious disease expertise considering the complexity of decision-making process; On the other hand, an indiscriminate use of broadspectrum antimicrobials is closely contributed to mortality [36,37], with higher AMR burden [38]. In this regard, timely and accurate retention of microbiological specimens to obtain appropriate cultures before antimicrobial administration is the basis for targeted therapy, obtaining more reliable local antimicrobial susceptibility patterns to increase the likelihood of prescribing appropriate initial antimicrobials, especially for patients with a higher risk of death in patients in the ICU. Our study has demonstrated that the introduction of an antimicrobial decision support tool into the ICU, ensuring the preconditions of microbiological specimens sampling before therapeutic antimicrobials use, was associated with measurable improvements in both initial antimicrobial use and overall use. Given that it has long been recognized that the inability to easily get access on data on local antimicrobial use patterns can be a gap [39][40][41], standardized metrics to individual facilities across a healthcare system has been a major advance. CDSS linked to EMRS and other clinical systems to this computerized platform has been proved to have the potential to facilitate the dissemination of information to intensivists for optimal use in therapeutic decision-making [42][43][44][45][46]. We consider all antimicrobial prescribing important and have shown that it is feasible to make full use of web-based application to help work with intensivists to improve antimicrobial prescribing practice by considering all patients' accompanying microbiology diagnostic testing issues helping the interpretation of some of our results and encouraging evidence-based decisions regarding choice of therapy. Our study achieved significantly reduced use of carbapenems, glycopeptides, third/fourthgeneration cephalosporins and anti-fungi agents in all.
Our study has a number of limitations. Firstly, it is an observational study in a single center without randomization and thus is subject to known biases. Secondly, many confounding factors interfere with the judgment of AMS performance. For example, rapid diagnostic test plays a collaborative role for ASPs [47]. At last, this study occurred in academic ICUs, most of the patients admitted to our ICUs were not "first-hand" patients who had been prescribed antimicrobials in community medical institutions or outpatient clinics before, thus, the microbiological data obtained through aCDSS was already partially biased, Furthermore, academic ICUs differ from community-based ICUs or long-term nursing care institutions in terms of staffing models and admitting more complex patients receiving broader spectrum antimicrobials. This study only represents the efforts of an established ASP with appropriate resourcing for wed-based interventions and may not be generalizable to other centers. In summary, while we were able to show temporal improvements in antimicrobial utilization structure concerning our interventions, larger, appropriately conducted randomized controlled trials or at least a controlled quasi-experimental design with longitudinal time series analysis are needed to further evaluate the effects of ASP including CDSS in critical care settings [48,49].

Conclusions
Computer-assisted decision support was associated with more reliable surveillance data on microbiology and sustained improvements in antimicrobials use in critical care.
Institution of a comprehensive antimicrobial stewardship program combing clinical decision support system has the potential to be an effective means to improve clinical diagnosis and treatment and to optimize antimicrobial use.

Ethics approval
All procedures were performed in accordance with the Declaration of Helsinki of the World Medical Association approved by the Ethics Committee of the Second Affiliated Hospital of Zhejiang University

Consent for publication
Not applicable

Availability of data and materials
The datasets during and/or analysed during the current study available from the corresponding author on reasonable request.

Competing interests
There is no financial and non-financi ompeting interests.  Figure 1 Flow charts of the intervention in microbiological specimen sampling. a. In cases sampling was unnecessary (treatment of syphilis), no appropriate specimen was available (i.e., deep organ infection), or the pathogen had already been identified earlier, the antibiotic prescription was allowed in the absence of sampling data.