Characteristics of included studies

Table 2 shows the characteristics of articles that were included in this review. Most studies were conducted in the United States (51.6%). The vast majority of studies were conducted (80.2%) in inpatient settings. Among studies conducted in inpatient settings, the majority were conducted in adult academic or teaching hospitals (30.0%) and community hospitals (22.3%). Adults also constituted the majority of patient populations where ASPs were conducted (75.3%).

Table 2.

Characteristics of included studies (N = 250)

Table 2 Long description

A table detailing the characteristics of included studies, focusing on regions, settings, age of patient population, specialty, and disease focus. The table has 20 rows and 6 columns. Column headers are Characteristic, Region, Settings, Age of patient population, Specialty, and Disease focus. Row labels include various categories under each characteristic. Values are presented as counts and percentages. Notable trends include the majority of studies conducted in the USA (51.6%), inpatient settings (80.2%), academic or teaching adult hospitals (30.0%), and adult patient populations (75.3%). Specialties are broadly distributed with hospital-wide being the most common (48.6%), and disease focus is varied with a significant portion not disease-specific (40.9%).

^* Articles reported multiple responses (percentages add up to >100%).

^{^} Other diseases and infections included human immunodeficiency virus (HIV), fungal infections, acute exacerbations chronic obstructive pulmonary disease, chorioamnionitis, non-purulent skin and soft tissue infections, and device-associated infections.

ASPs using hospital-wide strategies were reported in nearly half of the articles (48.6%). Emergency departments (10.1%) and intensive care units (8.5%) were the most common specialties where ASPs targeted at the unit level. Additionally, 15% of ASPs were implemented in more than one specialty. Close to half (40.9%) of the studies did not have a specific disease focus. Among strategies that had a specific disease focus, the most common were respiratory tract infections (8.1%), urinary tract infections (6.5%), and bloodstream infections (6.1%).

Study designs and theories

Table 3 provides specifics about study designs used. Pre- and post- assesment was the most commonly used study designs (90.7%). Single-site designs were used in 69.6% of studies, and multiple-site designs in 29.6% (See Table 3). The mean amount of time spent on preintervention data collection was 14.1 months (SD = 12.4), and the mean amount of time spent on postintervention data collection was 18.3 months (SD = 19.2). A total of 70.0% of articles were not overtly informed by any theory, model, or framework (TMF). Among articles that did report using a TMF, 23% of articles reported using a TMF from the CDC, WHO, Infectious Disease Society of America (IDSA), or other large professional organization.

Table 3.

Study design characteristics and ASP intervention strategies

Table 3 Long description

A table detailing study design characteristics and ASP intervention strategies. The table has 12 rows and 2 columns. The columns are labeled ‘Characteristic’ and ‘n (%)’. The rows are labeled with various characteristics and their corresponding values. Row 1: Study designs, Pre-post design, 224 (90.7 percent). Row 2: Study designs, Retrospective design, 180 (72.9 percent). Row 3: Study designs, Single-site design, 172 (69.6 percent). Row 4: Study designs, Multiple sites design, 73 (29.6 percent). Row 5: Study designs, Prospective design, 51 (20.6 percent). Row 6: Study designs, Comparison group, 31 (12.6 percent). Row 7: Study designs, Clustering, 22 (8.9 percent). Row 8: Study designs, Randomization, 13 (5.3 percent). Row 9: Study time, Mean time in preintervention, 14.1 months (SD = 12.4). Row 10: Study time, Mean time in postintervention, 18.3 months (SD = 19.2). Row 11: Theories, Models and Frameworks (TMF), None used, 173 (70.0 percent). Row 12: Theories, Models and Frameworks (TMF), CDC, WHO, IDSA or other large organization TMF, 57 (23.1 percent).

^* Articles reported multiple responses (total percentages add up to >100%).

^{^} Other TMF included System Usability Scale (SUS); WAR score for assessing infection in chronic wounds; evidence-based practice guidelines published in American Journal of Health-System Pharmacy; Mendelson‘s nine essential steps for ASP implementation.

ASP team compositions and involvement

The role and involvement of ASP teams varied across the studies (see Table 3). Fifty one percent of studies were led by an ASP team, although in some studies, ASP teams informed strategies (9.3%) or were not involved at all (13.4%). Pharmacists (45.3%) and medical doctors (44.1%) made up the majority of ASP team members who were involved in the strategy development and delivery.

Interventions and implementation strategies

Table 3 also describes the strategies used as an effort to improve antimicrobial use. Eighty-three percent (n = 208) of studies used more than one implementation strategy, with an average of three strategies used per study. Strategies targeted at clinicians were used in 89.5% of studies. Such interventions included audit/feedback approaches (eg, handshake stewardship), education, and guidelines at the unit level intended to change individual provider behavior. Interventions targeted at the unit level, including order set changes, automatic stops, and protocols that change workflow (prior approval), were used in 56.3% studies. Finally, interventions that focused on patients (eg, shared decision-making) were only used in 8.1% of studies.

Implementation strategies mostly constituted education, which was used in 63.6% of studies, followed closely by new guidelines, clinical pathways, policies, or decision trees (55.5%) and audit and feedback or handshake stewardship (55.1%). Strategies around diagnosis and testing were used in 32.8% of articles. This strategy involved the appropriate use of laboratory testing to guide patient management, including treatment, to optimize clinical outcomes and limit the spread of antimicrobial resistance. Other strategies included electronic medical records (EMR)-based interventions (26.7%), prior approval or restrictions (17.4%), facilitation (used in 13.8%), and quality improvement initiatives (QI), such as Plan-Do-Study-Act (PDSA) cycles (10.5%).

Implementation outcomes and effects

Table 4 describes the outcomes of the studies at the level of the implementation strategy, the antimicrobial resource used, and patient outcomes, if any. Forty-three percent of articles did not track any implementation outcome. Among the implementation outcomes that were tracked were fidelity (measuring the degree to which providers’ interventions adhered to guidelines or protocols; 32.0%) and acceptability (24.7%). Of note, 31.6% of articles reported all beneficial effects and 14.8% reported mixed effects (see Table 5).

Table 4.

ASP outcomes tracked

Table 4 Long description

A table categorizing outcomes tracked in studies related to antimicrobial stewardship programs. The table has three main categories: Implementation outcomes, Resource outcomes, and Patient outcomes. Each category lists specific outcomes with corresponding counts and percentages. Implementation outcomes include categories such as Fidelity, Acceptability, Implementation cost, Appropriateness, Reach, Feasibility, Knowledge, Sustainment, Awareness, and Adaptation. Resource outcomes include Measure of overall antibiotic consumption/amount, Drug choice, Duration, Drug cost, New ASP team development and implementation, Diagnostic switch, Timing of antibiotics, Dose of antibiotic, IV to oral switch, and Other. Patient outcomes include Length of stay, Infections, Mortality, Admissions or readmissions, MDRO emergence, Allergic or adverse reactions, and Other. The table lists the number and percentage of studies tracking each outcome. For example, 105 studies (42.5 percent) did not track any implementation outcome, while 79 studies (32.0 percent) tracked Fidelity. In the Resource outcomes category, 152 studies (61.5 percent) tracked the Measure of overall antibiotic consumption/amount, and 143 studies (57.9 percent) tracked Drug choice. In the Patient outcomes category, 95 studies (38.5 percent) tracked Length of stay, and 88 studies (35.6 percent) tracked both Infections and Mortality.

Table 5.

Reported implementation, patient, and resource effects

Table 5 Long description

A table comparing the outcomes of studies at the level of the implementation strategy, the antimicrobial resource used, and patient outcomes. The table has five rows and four columns. The columns are labeled Reported effects, Implementation outcomes n (%), Resource utilization outcomes n (%), and Primary patient outcomes n (%). The rows are labeled All beneficial, Mixed, Null, All harmful, and None reported. Row 1: All beneficial, 85 (34.4), 125 (50.6), 54 (21.9). Row 2: Mixed, 32 (13.0), 106 (42.9), 59 (23.9). Row 3: Null, 16 (6.5), 15 (6.1), 74 (30.0). Row 4: All harmful, 3 (1.2), 1 (0.4), 3 (1.2). Row 5: None reported, 106 (42.9), — , 57 (23.1).

Resource utilization outcomes and effects

Resource utilization outcomes refer to outcomes related to the use of hospital and material resources, such as duration of therapy, associated costs, and technologies involved in the intervention, excluding the implementation strategy itself. Measures of overall antibiotic consumption (eg, days of therapy) were the most tracked (61.5%), followed by change in drug choice in 57.9% of studies (eg, broad to narrow or replacing an inappropriate antibiotic with an appropriate one). Other outcomes that were tracked were duration (eg, average duration of therapy) in 35.2% of studies and cost (antibiotics/drug costs; cost of prescribing and administering antibiotics) in 23.1% of articles. All beneficial effects on resource utilization outcomes were reported in 50.6% of articles, with mixed effects in 42.9% and null effects in 6.1% (see Table 5).

Resource utilization outcomes by intervention strategy

Quality improvement initiatives, such as PDSA cycles, had the highest proportion of all-beneficial resource outcome effects (76.9%), though this strategy was used in a relatively small number of studies (n = 26). Education (52.9%), guideline or pathway implementation (55.5%), and audit and feedback approaches (51.5%) also showed majority beneficial resource effects. Prior approval and restriction strategies were notable in that most studies using this approach reported mixed effects (69.8%), the highest mixed-effects rate of any strategy, suggesting that while these approaches reliably shift antimicrobial use, the nature and direction of that change is variable across contexts. Facilitation had the highest null effects rate among all strategies (11.8%). Across all strategies, all-harmful resource effects were rare, ranging from 0% to 1.5% (see Table 5).

Primary patient outcomes and effects

For primary patient outcomes, that is outcomes related to the health of patients as a result of the healthcare resources applied to them, length of stay (38.5%), infections such as Clostridioides difficile and surgical site infection (35.6%), and mortality (35.6%) were the most commonly tracked outcomes. Admissions and readmissions were tracked in 23.9% of studies, and MDRO emergence in 17.8%. Fifty-nine studies (23.6%) did not track any primary patient outcome. Fifty-four (21.6%) studies reported all beneficial effects (ie, either improved patient outcomes or no increase in patient harm due to reduced antibiotic use), 23.9% reported mixed effects, and three studies reported all harmful effects (ie, increased length-of-stay or mortality). Null effects were reported in the remaining 30.0% studies indicating that many efforts neither improved nor harmed patient outcomes (see Table 5).

Impact of strategies on implementation, resource and patient outcomes

The Sankey diagrams in Figure 2 further explores the impact of three strategies on implementation, resource, and patient outcomes. Consistently across three strategies, which includes guideline implementation, audit and feedback, and facilitation, beneficial implementation outcomes are most frequently reported. Among studies with beneficial implementation outcomes, the stewardship outcomes tended to be either beneficial or include some beneficial and some null results (ie, mixed). Conversely studies with mixed or null implementation outcomes tended to have mixed or null stewardship and/or patient results. Individual panels also demonstrate outcome variability in the same strategy. Using panel 3 on audit and feedback as an example, while a meaningful proportion of studies with beneficial implementation outcomes also reported beneficial resource and patient-level outcomes, another meaningful proportion reported subsequent mixed outcomes. This demonstrates several possibilities within the literature. First, that audit and feedback works well for some outcomes but not others. Second, that some outcomes may be difficult to measure or not the best indicator of change, like AMR emergence which may require a longer time horizon relative to the length of the study. A third possibility is that different ways of structuring audit and feedback may have different impacts on subsequent outcomes.

Figure 2.

Sankey diagrams of impact of strategies on implementation, resource and patient outcomes.