Infection control measures

Several measures have been used to reduce the prevalence of HAIs and antibiotic-resistant bacteria (ARBs) in hospitals:

• Standard and contact precautions: Standard precautions are applied in hospitals in order to prevent basic infection. This includes hand hygiene policies and the use of personal protective equipment. Contact precautions are used in addition to standard precautions, comprising measures aimed at the discontinuation of pathogen-specific transmission pathways. These measures can include gowning, gloving, wearing a mask, and using patient-dedicated non-critical care equipment (e.g. stethoscopes) (Tacconelli et al., Reference Tacconelli, Cataldo and Dancer2014).

• Isolation or single-room care: If a patient is infected or colonized with the targeted pathogen, the patient can be transferred to a single room or into a cohort isolation together with patients colonized by the same pathogen. An alert code for patients previously colonized with ARB following single-room isolation has proven to be an effective strategy in preventing further spread of ARBs. This pre-emptive isolation remains active until the current colonization status of the patient has been verified. However, in the context of increasing colonization rates, there is an ongoing discussion on whether or not contact precautions are both effective and cost-effective, especially concerning Gram-negative bacteria (Tschudin-Sutter et al., Reference Tschudin-Sutter, Lucet and Mutters2017).

• Active screening cultures: Since many ARBs spread within the community, the number of patients with undetected colonization status can be highly dependent on the pathogen, setting, and country (Harris et al., Reference Harris, Nemoy and Johnson2004). In order to prevent hospital-wide spread of ARBs, active screening for colonization followed by strict contact precautions are recommended (Weintrob et al., Reference Weintrob, Roediger and Barber2010). For Gram-negative pathogens, the combination of screening the perirectal and groin areas results in the detection of 95% of carriers. For MRSA, the combination of screening throat and groin areas detects approximately 90% of carriers (Marshall & Spelman, Reference Marshall and Spelman2007). This evidence indicates the need for active screening procedures as a prevention strategy since the majority of patients that enter the hospital test positive for ARBs.

• Environmental cleaning (EC): Cleaning of particular surfaces near infected or colonized patients has been shown to be fundamental in HAI prevention and control (Barker, Alagoz & Safdar, Reference Barker, Alagoz and Safdar2017; Dancer, Reference Dancer2011). However, the pathogens on dry hospital surfaces vary in their resilience to EC. Strong evidence for EC effectiveness has been demonstrated by the control of outbreaks of Acinetobacter baumannii

• (Tankovic et al., Reference Tankovic, Legrand and De Gatines1994). Other examples of EC strategies have been used for Pseudomonas aeruginosa, which is spread via various pathways but most typically originates in biofilms in sinks (Salm et al., Reference Salm, Deja and Gastmeier2016). However, there is minimal evidence proving the effectiveness of EC in preventing HAIs in an endemic setting. EC is primarily used in a bundled approach and it is therefore difficult to assess its effectiveness as a single measure (Tacconelli et al., Reference Tacconelli, Cataldo and Dancer2014).

• Universal decolonization: This strategy is used for reducing the rates of HAIs. It has shown to be successful in preventing bloodstream infections, such as extended spectrum beta-lactamase-positive Enterobacteriaceae (Huang et al., Reference Huang, Septimus and Kleinman2013). In this approach, all patients, regardless of their colonization status, receive a daily chlorhexidine bath and mupirocin nasal ointment. Apart from this, there are no other promising regimens for long-term eradication of Gram-negative gut pathogens (Tacconelli et al., Reference Tacconelli, Cataldo and Dancer2014).

Cost–effectiveness of infection control measures to

prevent HAIs

A 2007 study evaluated the complex relationship between the rate of HAIs and the cost–effectiveness of preventive measures (Halton & Graves, Reference Halton and Graves2007). The study model presupposed the effectiveness of the prevention measure under evaluation. The findings demonstrated that infections due to ARBs led to a prolonged length of hospital stay, associated with increased costs for additional diagnostics, therapeutic interventions, and the additional number of hospital bed days. These effects derive from both patient complications and blocking of beds to prevent further patient contact and infection. These costs have a greater impact on hospitals that operate on a diagnosis-related group (DRG)-based system since prospective remuneration on admission is not typically covered by insurance companies. In order to assess the cost–effectiveness of a prevention measure, the excess costs of the HAI under consideration and the necessary investment to prevent the infection need to be known (Figure 4.1).

Figure 4.1 Relationship between the number of hospital-acquired infections and investments in infection control

Notes: HAI: hospital-acquired infection.

Line A (dotted) represents the costs of hospital infections, which is also the savings that would result from prevention. Line B (solid line) summarizes the relationship between the cost and the effectiveness of infection control strategies. Line C (dashes) is the sum of lines A and B for every incidence rate of hospital infections, representing the total cost of infection control strategies for HAIs. The point “X” represents the incidence of infection that minimizes total cost, which indicates a rational objective for decision-makers.

Source: Graves, Plowman & Roberts, 2001.

Robust data on the excess costs associated with HAIs are scarce. Currently, there is one meta-analysis that has estimated the excess costs of HAIs in the USA between 1986 and 2013 (Zimlichman et al., Reference Zimlichman, Henderson and Tamir2013). This study found that the additional costs range from approximately $900 (catheter-related UTI) to $46 000 (central line-associated bloodstream infection due to MRSA) (Table 4.1). These data should be interpreted cautiously as it represents the costs in only one national health care system. However, since most OECD countries use DRG-based payment systems, and in the absence of more reliable data, these figures can serve as orientation (OECD, 2014).

The most expensive aspects of a DRG-based hospital payment system are blocked beds and prolonged stays due to complications from HAIs. Beds are blocked in a multi-bed room in order to isolate an infected patient, leading to the non-availability of the remaining beds and reduced occupancy rates. Together, both infection control strategies account for approximately 80% of excess costs related to AMR (Conterno et al., Reference Conterno, Shymanski and Ramotar2007; Hübner et al., Reference Hübner, Hübner and Hopert2014). Although HAIs are commonly associated with additional diagnostics and treatment, the main driver of excess financial costs is the daily cost of hospital beds (Shepard et al., Reference Shepard, Ward and Milstone2013).

Interventions to prevent cross transmission

Similar to other bacteria, ARBs are transmitted within the hospital predominantly via patient contact with the hands of their caregivers (Longtin et al., Reference Longtin, Sax and Allegranzi2011; Pittet et al., Reference Pittet, Hugonnet and Harbarth2000; Pittet et al., Reference Pittet, Allegranzi and Sax2006; Tacconelli et al., Reference Tacconelli, Cataldo and Dancer2014). Patients who are infected or colonized with ARBs carry billions of colony-forming units per millilitre of stool or sputum. Hand hygiene involves cleaning hands with an alcohol-based hand rub to prevent the spread of bacteria. Health care workers who do not rigorously do this can carry hundreds of thousands of colony-forming units on their hands, which can then be transmitted to other patients. In the case of ARBs, this leaves the affected patients prone to diminished treatment options if exogenous infection occurs (Sax et al., Reference Sax, Allegranzi and Uçkay2007). Although this mode of pathogenesis is well-known and accepted in the medical world, compliance with hand hygiene policies by health care workers in hospitals is often as low as 40% (Longtin et al., Reference Longtin, Sax and Allegranzi2011). In order to facilitate the promotion of good hand hygiene in hospitals, the WHO developed an educational tool consisting of five indications of when hands should be disinfected: before patient contact, before an aseptic task, after exposure to bodily fluids, after patient contact, and after contact with patient surroundings (Sax et al., Reference Sax, Allegranzi and Uçkay2007). However, to improve compliance, multimodal strategies or intervention bundles should be used as they are found to be more successful (Damschroder et al., Reference Damschroder, Banaszak-Holl and Kowalski2009). Most importantly, a positive organizational culture is connected to low HAI rates and stabilizing high levels of hand hygiene compliance. Although difficult to assess, the success of this type of culture seems to be associated with the existence of role models who engage in hand hygiene and infection prevention.

From the late 1980s to early 2000s, health care systems in many industrialized countries have been restructured with the goal of decreasing hospital costs and increasing productivity (Clements et al., Reference Clements, Halton and Graves2008). This most often leads to shorter hospital stays per patient, enhanced patient throughput and hospital capacity. At the same time, AMR has been on the rise for MRSA, extended spectrum beta-lactamases (ESBL), vancomycin-resistant enterococci, and carbapenem-resistant Enterobacteriaceae (ECDC, 2010; Gagliotti et al., Reference Gagliotti, Balode and Baquero2011; Gastmeier et al., Reference Gastmeier, Schröder and Behnke2014; de Kraker & van de Sande-Bruinsma, Reference de Kraker and van de Sande-Bruinsma2007). This has created a vicious cycle characterized by overcrowding and understaffing that works against high levels of hand hygiene compliance. This eventually led to increased HAI rates, hospital costs, and more cost pressure (Clements et al., Reference Clements, Halton and Graves2008). Although many countries have acknowledged this as a problematic situation, few have taken action or initiated measures to relieve the pressure (Kaier, Mutters & Frank, 2012).

Surveillance systems of HAIs

In many industrialized countries, HAIs are benchmarked to allow for comparison across different hospitals (Haustein et al., Reference Haustein, Gastmeier and Holmes2011; Tacconelli et al., Reference Tacconelli, Sifakis and Harbarth2017). This has the potential to identify best practices, improve standards of care, and stabilize the performance of the health services offered. A useful comparison of outcome indicators requires consistent definitions, surveillance methods, and standardized rates. These rates must also be risk-adjusted for differences across the patient population and types of medical procedures. In Europe, there are several national surveillance systems for HAIs and a centralized surveillance system for ARBs – EARS-Net. In 2011–2012 and in 2016–2017, the ECDC also carried out two Europe-wide point prevalence surveys (PPS) of HAIs and antimicrobial use in acute hospitals (ECDC, Reference Davey, Brown and Charani2013; 2016).

Surveillance definitions for HAIs are somewhat complex and can lead to disagreement among the clinicians responsible; for example, with ventilator-associated pneumonia and surgical site infection. Surveillance systems should therefore be the responsibility of professionals trained in HAI surveillance (Gastmeier et al., Reference Gastmeier, Geffers and Brandt2006). Continuous surveillance systems that assess primarily for infection incidence are time-consuming and often cost-effective only for larger institutions. By contrast, measuring infection prevalence with cross-sectional surveys (e.g. PPS) is less resource-intensive (Haustein et al., Reference Haustein, Gastmeier and Holmes2011; Tacconelli et al., Reference Tacconelli, Sifakis and Harbarth2017).

However, this system is more applicable for assessing the overall burden of HAIs than for benchmarking between hospitals. The use of HAI indicators for benchmarking in the different national surveillance systems is well established, yet there are substantial differences with respect to the indicators, methods, and reporting techniques. An example of a successfully functioning national surveillance system is the German KISS system (Krankenhaus-Infektions-Surveillance-System). Based on the US national nosocomial infections surveillance system model, this voluntary and confidential system has been in operation since 1997 (Gastmeier et al., Reference Gastmeier, Sohr and Schwab2008). Currently, approximately two thirds of all German hospitals participate in KISS (Schröder et al., Reference Schröder, Schwab and Behnke2015; Leistner et al., Reference Leistner, Piening and Gastmeier2013). This system provides several small surveillance modules for various risk groups (e.g. ICU patients, surgical patients), ARBs, and C. difficile, and the use of an alcohol-based hand rub.

ECDC collects laboratory-based data for AMR using the European Antimicrobial Resistance Surveillance Network (EARS-Net). Although all existing national and international surveillance systems can provide comprehensive information on AMR and HAIs, the results are usually published years after the data are collected (Tacconelli et al., Reference Tacconelli, Sifakis and Harbarth2017). This diminishes their utility for clinical, institutional, and regulatory decision-making. Further, the delay may result in misalignment of targeted resources and research priorities since information will not be up to date at the time of policy development.

Measures for outbreak control

Outbreaks of HAIs pose major challenges for hospital management and the infection control department of an affected institution. Although outbreaks typically affect only one department or ward, they are often publicly perceived as a malfunction of the entire hospital. In order to control current and future outbreaks, such events should be rigorously investigated. Infection and microbiological diagnoses should be recorded and analysed continuously. This can act as an early warning system. Such systems can be based on microbiology data from the laboratory, surveillance data, or clinical data. The delay between a possible outbreak and its detection depends on the type of system used. In order to ensure cost–effectiveness of the system, it should be adapted to the individual hospital based on a risk assessment by an infection control or hospital epidemiology expert.

The results of the analysis should then be communicated to all affected players since this information presents an excellent training opportunity. In the case of an outbreak, an alert signal is provided by the surveillance system. The outbreak alert can result from clusters of:

• the same pathogen (e.g. ARBs in different microbiological specimens) (Salm et al., Reference Salm, Deja and Gastmeier2016);

• the same types of infection (e.g. central venous catheter-associated bloodstream infections due to different pathogens) (Price et al., Reference Price, Hacek and Noskin2002);

• a combination of both (e.g. surgical site infections with Candida albicans) (Pertowski et al., Reference Pertowski, Baron and Lasker1995).

An alert signal initiates outbreak investigation. This should be focused on epidemiological and microbiological data in order to identify the likelihood of an outbreak. The first evidence that can indicate a potential outbreak is the comparison of baseline and epidemic rates (Barker, Alagoz & Safdar, Reference Barker, Alagoz and Safdar2017). At this early stage, it is necessary to create a line list that aggregates relevant information on all potentially affected patients with epidemiological data. Microbiological sampling of pathogens from infection, environment, and patient colonization can then be used for further investigation (Lippmann et al., Reference Lippmann, Lübbert and Kaiser2014). This should be conducted in a timely fashion since most microbiological laboratories dispose of their samples after 7 to 14 days. Some infection control departments store pathogens of interest in order to allow for retrospective analysis (Salm et al., Reference Salm, Deja and Gastmeier2016). These samples can then be investigated to verify their genetic relatedness using pulsed-field gel electrophoresis or whole genome sequencing (Kampmeier et al., Reference Kampmeier, Knaack and Kossow2017; Sax et al., Reference Sax, Bloemberg and Hasse2015; Snitkin et al., Reference Snitkin, Zelazny and Thomas2012; Welinder-Olsson et al., Reference Welinder-Olsson, Stenqvist and Badenfors2004). The most cost-effective epidemiological approach is a case–control study, which has the potential to yield information that either reinforces suspected risk factors or leads to undiscovered connections between cases (Moolenaar et al., Reference Moolenaar, Crutcher and San Joaquin2000; Salm et al., Reference Salm, Deja and Gastmeier2016).

Outbreak control measures should be directed at the pathogen, the suspected routes of transmission, and its epidemiology within the institution. An orientation on likely transmission pathways and possible control activities can be found in the Centers for Disease Control and Prevention’s (CDC) Guidelines for isolation precautions: preventing transmission of infectious agents in healthcare settings (Siegel et al., Reference Siegel, Rhinehart and Jackson2007). At the outset of an outbreak, the pathogenesis is often unknown.

Given this, control measures are often conducted using a broad, horizontal approach. Therefore, it is imperative for contact to be made with the key clinical players (e.g. physicians, nurses, and department chiefs) as well as to raise awareness in all affected departments (Moolenaar et al., Reference Moolenaar, Crutcher and San Joaquin2000). All players should be informed of the outbreak’s course to improve compliance with control measures and ensure that the outbreak serves as a learning opportunity for prevention of future outbreaks under similar conditions.

Taxonomy of ABS interventions

There have been several systematic reviews on interventions to change the prescribing behaviour of professionals in hospitals (Davey et al., Reference Davey, Marwick and Scott2005; 2013; 2017). Davey et al. performed a critical appraisal using the Cochrane Effective Practice and Organisation of Care Group (EPOC) taxonomy (Davey et al., Reference Davey, Brown and Fenelon2017). They appraised all interventions relevant to improving antibiotic prescribing categorized as persuasive, restrictive, or structural. The EPOC definitions of the interventions and examples of the intervention components are given in Table 4.2.

Evidence on effectiveness of ABS interventions

In their most recent Cochrane review, Davey et al. (Reference Davey, Brown and Fenelon2017) concluded that there is high-certainty evidence that interventions are effective in increasing compliance with antibiotic policy and reducing duration of antibiotic treatment. Lower use of antibiotics probably does not increase mortality but does reduce length of stay. Enablement (i.e. increasing means/reducing barriers to increase capability or opportunity) consistently increased the effect of interventions, including those with a restrictive component. Although feedback to health care professionals further increased the effect of an intervention, it was used in only a minority of enabling interventions. Interventions were successful in safely reducing unnecessary antibiotic use in hospitals despite the fact that the majority did not use the most effective behaviour change techniques (Davey et al., Reference Davey, Brown and Fenelon2017).

Schuts et al. (Reference Schuts, Hulscher and Mouton2016a) performed a systematic review and meta-analysis of the evidence on selected ABS objectives, evaluating for the effect of 14 ABS objectives on four predefined patient outcomes: clinical outcome, adverse events, costs, and bacterial resistance rates. The ABS objectives consisted of a mixture of 11 consensus-derived quality indicators of antibiotic use and three additional objectives from the 2007 Infectious Diseases Society of America guidelines on ABS (van den Bosch et al., Reference van den Bosch, Geerlings and Natsch2015; Dellit et al., Reference Dellit, Owens and McGowan2007). They identified 145 unique studies with data on nine out of the 14 stewardship objectives. Objective characteristics are summarized in Table 4.3. Overall, the quality of evidence was generally low and heterogeneity between studies was mostly moderate to high. For the objectives empirical therapy according to guidelines, de-escalation of therapy, switch from intravenous to oral treatment, therapeutic drug monitoring, using a list of restricted antibiotics, and bedside consultation, the overall evidence showed significant benefits for one or more of the four outcomes. Guideline-adherent empirical therapy was associated with a 35% relative risk reduction of mortality (RRR) and de-escalation with a RRR of 56% (Schuts et al., Reference Schuts, Hulscher and Mouton2016a). Evidence of effects was less clear for adjusting therapy according to renal function, discontinuing therapy based on lack of clinical or microbiological evidence of infection, and having a local antibiotic guide. Schuts et al. concluded that for several ABS objectives there is abundant, but low quality, evidence on clinical outcomes, adverse events, costs, and resistance rates in hospitals (Schuts et al., Reference Schuts, Hulscher and Mouton2016a).

Methodology of ABS interventional studies

This section provides an overview of the methodology used when testing for the effectiveness of antibiotic stewardship programmes. Studies have found overall positive effects but the methodology has not been clearly assessed for external validity or generalization to other populations.

In a narrative review, de Kraker et al. (Reference de Kraker, Abbas and Huttner2017) evaluated for the differences between various study designs in their ability to provide a framework for assessing the quality of evidence for ABS interventions. Relevant literature was identified using a database search of Cochrane and PubMed. The authors found that random time effects and bias can jeopardize the validity of causal inference in ABS research. The most important risks include simultaneously implemented strategies and regression to the mean. Inclusion of homogeneous intervention and control arms, through randomization of the intervention, can limit these risks. However, contamination, that is spill over from the intervention to the control arm, can play an important role for ABS. Therefore, it is recommended that randomization is conducted at the cluster rather than the individual-level. However, it can be challenging to identify enough representative clusters and implementation of a cluster-randomized control trial can be costly. Controlled interrupted time series design has a high validity as well, and is easier to implement, although time-varying confounding should be considered. To detect any unintended consequences, it is crucial to include multiple process, clinical outcome, microbiological and financial measures (de Kraker et al., Reference de Kraker, Abbas and Huttner2017).

A recent study reviewed published systematic reviews retrieved from Medline to study the evidence base of antibiotic use recommendations and behavioural change interventions (Hulscher & Prins, Reference Hulscher and Prins2017). It found that most current studies used designs prone to confounding by indication, where participants with less complex or less severe illness may be more likely to have received appropriate antibiotic treatment, which will confound the association between appropriate use and the outcomes tested. Much could be learnt from behavioural sciences. The literature demonstrated that the quality of evidence is low for the positive effects of appropriate antibiotic use in hospital patients. In addition, it found that all types of behavioural change intervention might work. Although effects were positive overall, there were large differences in improvement between studies that tested similar change interventions. Confirming findings elsewhere, the research showed a clear need for studies that use an appropriate study design, i.e. both randomized and controlled, to test for the effectiveness of appropriate antibiotic use in achieving meaningful outcomes (Davey et al., Reference Davey, Brown and Fenelon2017; de Kraker et al., Reference de Kraker, Abbas and Huttner2017).

ABS guidelines

With growing evidence on the benefit of particular prescription practices, national and international health agencies have issued and regularly updated guidance to address AMR by encouraging appropriate use of antibiotics. In 2014, the CDC recommended that all acute care hospitals in the United States implement ASPs (Fridkin & Srinivasan, Reference Fridkin and Srinivasan2014). Importantly, the CDC recommends a commitment to stronger leadership to enable dedication of the necessary human, financial, and information technology resources to ASPs. In 2015, the WHO published its Global action plan on antibiotic resistance which urges all countries to optimize the use of antibiotic agents (World Health Organization, Reference Organization2015). In 2016, the Infectious Diseases Society of America issued new antibiotic stewardship guidelines which focused on practical advice for implementing ASPs (Barlam et al., Reference Barlam, Cosgrove and Abbo2016). These replace former, outdated guidelines and focus on specific strategies that are thought to be more beneficial to ensure that the ASP will be effective and sustainable. They recommend that ASPs should tailor interventions based on local issues, resources, and expertise. To ensure this, the guidelines recommend that the ASP is led by physicians and pharmacists and rely on the expertise of infectious diseases specialists. Most recently, ECDC has issued Proposals for EU guidelines on the prudent use of antibiotics in humans in 2017 (European Commission, Reference Commission2017). Many other EU Member States have issued national guidelines for antibiotic use and ASPs in response to EU-wide calls for action.

Some best ABS practices

National surveillance data on AMR show higher antibiotic use and higher resistance levels in the south and east of Europe compared to the north and west. The Netherlands is an example from the latter, with low consumption and AMR, and a long tradition of antibiotic policies in hospitals. By contrast, national data from Italy show a high antibiotic consumption (ESAC-Net). However, individual hospitals in Italy have started with ASPs focusing on local issues. Box 4.1 shows a selection of exemplary local good antibiotic stewardship practices.

Cost–effectiveness of interventions

Governments have limited financial resources. Interventions that are both effective and cost-saving are necessary to reduce the high cost burden of AMR on public health and health system functioning.

A study reviewed the literature on cost–effectiveness of ABS programmes in hospital settings of OECD countries up to June 2014 (Coulter et al., Reference Coulter, Merollini and Roberts2015). The type of ABS strategy and the clinical and cost outcomes were evaluated in 36 studies on adult patients. The main ABS strategy implemented was prospective audit with intervention and feedback, followed by the use of rapid diagnostic technology, e.g. rapid polymerase chain reaction (PCR)-based methods or matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF), for the treatment of bloodstream infections. All but one of the 36 studies reported that ABS resulted in a reduction in pharmacy expenditure. Among 27 studies measuring changes to health outcomes specifically, either no change was reported after the ASP or the additional benefits achieved from these outcomes were not quantified. Only two studies performed a full cost–effectiveness analysis (CEA) (Brown & Paladino, Reference Brown and Paladino2010; Scheetz et al., Reference Scheetz, Bolon and Postelnick2009). Both CEAs used a decision-tree model from the hospital perspective and did not evaluate societal costs. Both studies found the interventions to be cost-effective. The earlier study used a model comparing costs and outcomes of bacteraemic patients receiving standard treatment with or without an ASP team consultation (structural intervention). Effectiveness was estimated as quality-adjusted life-years (QALYs) over the lifetime of patients. Incremental cost–effectiveness ratios were calculated to estimate the cost per QALY gained. The later study found that the implementation of rapid testing resulted in improved outcomes for patients. They used data from the literature both from the EU and the USA to inform the model. Rapid PCR testing for MRSA reduced mortality rates while being less costly than empirical therapy in the EU and the USA, even when factoring in a wide range of MRSA prevalence rates and PCR test costs. ABS programmes frequently resulted in a reduction in pharmacy costs. However, there was a lack of consistency in the reported cost outcomes making it difficult to compare the results of the included interventions (Coulter et al., Reference Coulter, Merollini and Roberts2015).

The most recent study on the cost–effectiveness and cost–benefits of ASPs summarized the literature from 2000 to 2007 (Naylor et al., Reference Naylor, Zhu and Hulscher2017). In addition to the CEAs discussed above, it included a CEA that used a Markov model for analysing a bundled ABS strategy (persuasive intervention) conducted in a hospital in Brazil (Okumara et al., Reference Okumara, Riveros and Gomes-da-Silva2016). Overall, it concluded that the cost–effectiveness evidence for ABS is severely limited and remains inadequate for investment decision-making. Robust health economics research is needed to enhance the generalizability and usability of cost–effectiveness results.

Recommendations

The following recommendations can be made:

• Regarding AMR and HAI surveillance, intensified international collaboration is needed in order to overcome existing barriers to high-quality surveillance.

• Robust data from national surveys are needed to provide useful and comprehensive information to decision-makers in local hospitals.

• To increase the success of educational ABS interventions, education of all professionals should start at the undergraduate level and include medical students, pharmacists and nurses.

• Other stakeholders should be involved to promote responsible antibiotic use in hospitals.

Future directions for research

Although there is a large amount of published literature on interventions to curb AMR in hospitals, some of the relevant outcomes relating to patient health such as patient safety or economics have been neglected. In addition, intervention studies should have more robust designs.

Future research should focus on:

• Targeting treatment and assessing other measures of patient safety, assessing different stewardship interventions, and exploring the barriers and facilitators to implementation. More research is required on unintended consequences of restrictive interventions.

• Robust study designs such as cluster-randomized controlled trials, or interrupted time series including a control arm. A detailed process evaluation should be provided to adequately inform implementation of successful ABS strategies.

• High-quality health economics research on ABS with an appropriate health-economic methodological choice to enhance the generalizability and applicability of cost–effectiveness results.