Rising rates of antimicrobial resistance have spurred the global advent of antimicrobial stewardship programs (ASPs) throughout both inpatient and outpatient healthcare settings. The implementation of such programs within hospitals has faced many challenges particularly at the administrative level with concerns regarding personnel shortages, finances, and low prioritization (1,2). Despite this, multiple studies have demonstrated the effectiveness of such ASP initiatives with reductions in Clostridium difficile rates, lower rates of broad-spectrum antibiotic use, and potential for cost savings (3–6). A further push has been made in the last decade toward implementation of ASP programs within ICUs given the relatively high rates of broad-spectrum antibiotic use among critically ill patients and associated antibiotic-related complications. Although some hesitancy may be present in implementing these programs within a critically ill population, a recent systematic review of 11 studies identified safe implementation of audit-and-feedback ASP programs within ICUs (7).
Although multiple studies have demonstrated the efficacy and safety of short-term implementation of ASP within ICUs, there is a paucity of long-term data to ensure sustained success of these programs. During initial implementation of or ICU ASP, we reported reduction in broad-spectrum and total antimicrobial use, as well as C. difficile (3). More recently, we detected hospital-wide reductions in healthcare-associated multidrug resistant organism prevalence despite rising rates of local community-associated multidrug resistant organisms over the same time span (8). However, the trend in antimicrobial recommendations over time is unknown, and it is not clear what patient-, infection-, and treatment-related factors predict the need for ASP recommendations and the acceptance of ASP suggestions in ICU. To address this, we conducted a 10-year retrospective analysis of an ASP audit-and-feedback program in our hospital’s ICUs. Our aims for this study were three-fold: 1) to describe long-term trends in suggestion and acceptance rates of ASP interventions, 2) to identify predictors of ASP suggestions being offered, and 3) to identify predictors of acceptance versus rejection of ASP suggestions in the ICU.
MATERIALS AND METHODS
General Study Design
Our study was designed as a retrospective cohort focused on ICUs within Sunnybrook Health Sciences center, a 627-bed academic tertiary care hospital located in Toronto, ON, Canada, with a broad patient population including trauma, cardiovascular surgery, burns, neurosurgery, stroke, oncology, and general medical-surgical ICU patients. The intensive care population was divided among three “level 3” ICUs with mechanical ventilation capability (cardiovascular surgery, burns, and medical/surgical) and two “level 2” medical/surgical ICUs differing from level 3 ICUs by not routinely providing invasive ventilation. All ICUs were considered closed units, which are run by intensivists only. Through the course of the study, there were approximately 20 “level 3” medical/surgical beds, 14 “level 3” cardiovascular beds, 21 “level 2” medical/surgical beds, and 14 “level 3” burn surgery beds. We identified all patients assessed by the ASP team while admitted to an ICU between June 1, 2010, and September 30, 2019, as captured by the Stewardship Program Integrating Resource Information Technology (SPIRIT) database (9). The SPIRIT database was created as a prospective repository, which is autopopulated with all pharmacy and microbiology data for each admitted patient, along with select patient characteristics (9). All antibiotic prescriptions are written orders with no mandatory durations required; however, there are certain restrictions on antibiotic orders (linezolid, daptomycin, tigecycline, fourth and fifth generation cephalosporins, and colistin) which require infectious disease consultation and approval. We did not include patients admitted to the coronary care unit (primarily caring for patients with ischemic cardiac disease and/or receiving percutaneous coronary artery interventions) and assessments made for antifungal and antiviral therapies.
Nature of the ASP Intervention
The ASP team at Sunnybrook Health Sciences Center (SHSC) is a joint effort between pharmacists who specialize in antimicrobial stewardship and infectious disease consultant physicians. This program was initiated at SHSC in 2010 in the ICUs and, ultimately, expanded to the ward setting (3,10). Prior to ASP implementation in SHSC, an internal assessment revealed only a minority of patients (approximately ¼) on targeted antibiotics were being followed by the infectious disease consultation service. Initially, the program was run by one full-time pharmacist and infectious disease staff physician; however, it has expanded to include an additional full-time pharmacist, one part-time pharmacist, and pharmacy resident trainees over the course of the study. Patients who were identified as being on consecutive days of broad-spectrum antibiotics were assessed by the ASP pharmacists via a chart review, brief clinical patient assessment, and discussion with the primary team either in person or via telephone discussion followed by a formal chart order suggestion. Antimicrobial reviews were conducted 5 days a week (Monday to Friday) by the pharmacist team with new patients being automatically rostered based on day of broad-spectrum antibiotic therapy. During 2010–2014, patients were assessed on days 3 and 10 of broad-spectrum antibiotic use, and, from 2015 onwards, patients were assessed on days 3, 7, and 14 of broad-spectrum antibiotic use. We felt this allowed for interventions at standardized time points in therapy, while balancing intensive care physician and pharmacist autonomy in antibiotic initiation, selection, and duration. “Broad-spectrum antibiotics” were defined as third-generation cephalosporins, piperacillin-tazobactam, carbapenems, fluoroquinolones, aminoglycosides, and IV vancomycin. Patients who qualified for an ASP assessment with a particular antibiotic had assessments provided for all other active antibiotics as well. If the ASP pharmacist felt that there was a need to recommend a change in treatment, the case was reviewed with an infectious diseases physician consultant, and suggestions were provided directly to the primary clinical team. The possible suggestion categories were “broaden coverage” (change antibiotic to one with broader microbial spectrum coverage, such as from Ceftriaxone to piperacillin-tazobactum for hospital-acquired pneumonia), “change agent” (change antibiotic to one that aligns with indication of therapy without consideration of broadening or narrowing spectrum of coverage, such as from ceftriaxone to azithromycin for confirmed legionella pneumonia), “change dose”, “change formulation”, “discontinue”, or “narrow coverage” (change antibiotic to one with narrower microbial spectrum coverage, such as ceftriaxone to amoxicillin for low-severity community-acquired pneumonia coverage). The decision of suggestion category assignment was made in real-time by the ASP pharmacists and not altered in our retrospective review of the data unless no category was assigned initially. Alternatively, if the ASP team felt no change was indicated, a recommendation of “continue unchanged” was provided. Patients were excluded from ASP review if they were being actively followed by the infectious diseases consult service or if the antibiotics they were receiving were recently recommended by infectious diseases consult service. Our hospital followed a combination of national and international guidelines which were adjusted for local susceptibilities and antibiograms, if necessary, in our institutional guidelines which are available via an online resource to physicians on the hospital website.
Long-Term Trends in ASP at SHSC.
To evaluate the utilization and need of ASP within the ICUs, the monthly proportion of ICU patients who received an ASP assessment was calculated during the study period. Using the SPIRIT database, each ASP assessment was reviewed to determine whether a suggestion was made, and if so, whether that particular suggestion was accepted or rejected. Acceptance was defined as the ASP suggestion being implemented within 24 hours of the suggestion being made to the primary team. This was adjudicated by author (P.S.) and a second reviewer (J.L.) who independently reviewed 5% (n = 387) of cases to ensure agreement (kappa = 0.95; 95% CI 0.91–0.99; 98% agreement). We identified the ASP acceptance rates on a monthly basis to understand the temporal trend of acceptance over the course of the study period.
Predictors of ASP Suggestions and Acceptance.
We used the Critical Care Information System database in conjunction with the SPIRIT database at SHSC to examine variables that would potentially contribute to a decision by the ASP team to make a suggestion and the primary team to accept or reject an ASP suggestion. The potential predictor variables we extracted were patient demographics, year of intervention, length of stay prior to ASP assessment, ICU level of care, admitting service, mechanical ventilation, central venous access, invasive arterial monitoring, intracranial pressure monitoring, dialysis, vasopressor support, Multiple Organ Dysfunction Score (MODS), laboratory investigations (WBC count, platelet count, and lactate), indication for therapy, antibiotic assessed, antibiotic duration, and ASP suggestion type. All variable values were taken at time of ASP assessment. The analysis was done in two stages: 1) identifying variables that predict a suggestion as opposed to “continue unchanged” and 2) identifying variables that predict acceptance of a suggestion.
For evaluation of long-term trends in the ICU, we measured monthly suggestion rates and quarterly acceptance rates over the course of the study and used autoregressive linear models to assess time trends taking into account autocorrelations. Autocorrelations are inspected graphically and using the Durbin-Watson test.
For the predictive factors described above, continuous variables were expressed as means with sd or medians with interquartile ranges, whereas categorical variables were expressed as frequencies and percentages. Univariate associations between the patients’ factors and outcomes were tested using simple logistic regression.
We conducted multivariable logistic regression analysis to examine patient-, infection-, and treatment-related characteristics associated with the ASP team making a suggestion to alter therapy and adjusting for the year of intervention. The unit of analysis was the individual antibiotic assessment, and to account for clustering of multiple antibiotic assessments within individual patients, the generalized estimating equations method was used. The covariates of interest were prespecified based on clinical grounds of potential to impact assessments. To evaluate predictors of making ASP suggestions and suggestion acceptance, we identified a reference item for each subgroup which we believed would have the most neutral clinical impact. Among the subset of antibiotic assessments in which a suggestion was made to alter therapy, we conducted a similar analysis using multivariable logistic regression analysis to examine patient-, infection-, and treatment-related characteristics associated with suggestion acceptance. This model included all of the same covariates as the first model, with the addition of a variable denoting the type of suggestion that was made. In the primary analysis, we excluded laboratory value covariates given that they were anticipated to be missing in substantial numbers of patients. We conducted a secondary analysis incorporating the laboratory values into the models.
Research Ethics Board
This study was reviewed by the Sunnybrook Health Sciences Center Research Ethics Board and approved.
There were 7,749 antibiotic assessments made by the ASP team during 5,569 ICU patient encounters between June 1, 2010, and September 30, 2019. The mean age of patients was 63 (± 20) years, and 65% were males. The mean MODS was 4.43 (± 2.77); 59% were receiving mechanical ventilation, and 30% vasopressors or inotropes. Assessments were predominantly in level 3 ICUs (68%) with the medical/surgical ICU accounting for 40% of all assessments. An ASP change suggestion was made in 2,826 encounters (36%). The most common suggestions were “discontinue” (1,771; 63%), “narrow” (374; 13%), and “change dose” (264; 9%). The most common indications for therapy were infection of unknown source (1,676, 22%), community-acquired pneumonia (1,197; 15%), and intra-abdominal sepsis (1,152; 15%). Piperacillin-tazobactam was the most assessed antibiotic (1,889; 24%) followed by ceftriaxone (1,597; 21%) and fluoroquinolones (962; 12%).
Long-Term Trends in ASP Suggestions and Acceptance
During the study period, 14% of ICU patients qualified for review by the ASP team (Fig. 1). There was a significant temporal trend toward increasing rates of ASP suggestions (1.23% for every year; 95% CI 0.85–1.62) using an autoregressive model with lag of 1 and 5 months. The overall acceptance rate of ASP suggestions was 67%, which was relatively stable over time (–0.22%; 95% CI –1.58 to 1.15) (Fig. 2) with no significant time trend based on an autoregressive model with lag 1 quarter-year. There was variation in ASP acceptance rates across both ASP suggestion types and ICUs with lower mean acceptance rates noted for suggestions of “narrow” (65%), “change agent” (65%), and “change formulation” (55%) along with lower acceptance rates for burn ICU patients (60%).
Predictors of Making ASP Suggestions
Multiple factors were associated with increased odds of making an ASP suggestion (Supplemental Table 1, Supplemental Digital Content 1, http://links.lww.com/CCM/F925): admission to the cardiovascular surgery service (odds ratio 1.37; 95% CI 1.06–1.76) or burn surgery service (1.88; 1.50–2.36), a urogenital indication for therapy (1.61; 1.19–2.18), and receipt of specific antibiotics (compared with ceftriaxone as referent agent) including aminoglycosides (2.91; 1.85–4.59) and IV vancomycin (2.71; 2.19–3.36). The following factors were associated with decreased odds of making an ASP suggestion: admission to a level 3 ICU (0.79; 0.67–0.93); vasopressor use (0.84; 0.73–0.96); a primary CNS (0.56; 0.34–0.93), intra-abdominal (0.54; 0.42–0.69), or skin/soft-tissue infections (0.65; 0.49–0.86) compared with bacteremia.
Predictors of ASP Suggestion Acceptance
Admission to a level 3 ICU (1.50; 1.14–1.97) was the only variable identified associated with higher acceptance of an ASP suggestion (Supplemental Table 2, Supplemental Digital Content 2, http://links.lww.com/CCM/F926). Factors associated with decreased rates of acceptance included admission to burn surgery service (0.64; 0.45–0.91), community-acquired pneumonia (0.64; 0.42–0.97), ventilator-acquired pneumonia (0.65; 0.44–0.94), and unknown source of infection (0.66; 0.48–0.92). A suggestion of “narrow” (0.65; 0.45–0.94), “change formulation” (0.42; 0.27–0.64), or “change agent” (0.63; 0.40–0.97) were all associated with lower rates of ASP suggestion acceptance compared with a suggestion of “change dose”.
In this retrospective cohort study, we reviewed 10 years of ASP recommendations in a diverse patient population within the ICUs of a tertiary care hospital. Our study showed a stable 67% acceptance rate of ASP suggestions, relatively unchanged after increasing involvement of the ASP service and a proportional increase in recommendations observed. Patients who were receiving very broad-spectrum antibiotics, those associated with potential toxicities (vancomycin and aminoglycosides) and patients admitted to subspecialized care units were more likely to receive an ASP recommendation, whereas patients with the highest acuity of illness and those with an unclear source of infection were less likely to receive recommendations. Patients admitted to the highest acuity ICUs were more likely to have suggestions accepted, whereas those with respiratory or skin infections and recommendations to narrow, change antibiotic agent, or formulation of the antibiotic were less likely to have suggestions accepted.
To determine the sustainability of our ASP program, we trended acceptance rates of suggestions over the course of the study. This method has been used in previous studies tracking the implementation of ASP programs particularly in the short-term and predominantly limited to ward-based acute care settings with acceptance rates varying between 72% and 92% (11–14). One systematic review of ASP in ICUs similarly identified acceptance rate variations between 67% and 96% across a range of ASP models and countries (15). Despite our study showing an increased number of assessments by the ASP team, we were able to show a consistent acceptance rate over the decade. We believe the cause of these increased assessment numbers is multifactorial relating to an evolution of our program approach over time (switching from day 3 and 14 reviews to days 3, 7, and 14 review allowing for earlier intervention), increased expansion of our program from one pharmacist to three, and a change in overall antibiotic use practices based on studies showing safety in less antibiotic days of therapy for certain clinical conditions (16,17).
We identified multiple factors that influenced both the likelihood of an ASP suggestion as well as the likelihood of acceptance of suggestions. Our findings that patients in the specialized units, burns, and cardiovascular surgery were more likely to receive ASP suggestions relate to the uniqueness of these populations both in their patient characteristics and infectious diseases. Burn patients, in particular, are faced with frequent challenging and potentially life-threatening infections after their initial burn, which predispose them to higher rates of antibiotic receipt and rates of multidrug resistant organisms (18). Therefore, these patients remain a priority for ASPs, reflected with increased assessment and suggestions over time compared with the general medical/surgical population. Patients who were mechanically ventilated, on vasopressors, or with higher MODS were less likely to receive ASP suggestions to alter therapy, possibly reflecting a reluctance to narrow or discontinue therapy in the most critically ill (19). Similarly, patients with infectious diagnoses with perhaps greater diagnostic uncertainty—hospital-acquired CNS infections, intra-abdominal sepsis, or skin and soft-tissue infections—were less likely to receive suggestions, whereas urogenital infections, generally with a higher rate of microbiologic confirmation and susceptibility testing, led to more frequent ASP suggestions.
Compared with bacteremia, ASP recommendations made for patients with all pneumonias, skin and soft-tissue infections, and unknown source of infections tended to be less frequently accepted compared with patients with objective bacteremia. There remains significant variance in defining infections within intensive care, particularly ventilator-acquired pneumonias which can have potentially low inter-rater reliability with ranges from 35% to 97% depending on the diagnosis and 70% for pathogen agreement (20). Especially in cases where no pathogen has been speciated, it becomes difficult to alter antimicrobial therapy particularly if the patient remains unstable.
Last, intensivists were less likely to accept ASP suggestions that were focused toward altering the therapeutic range of antibiotics by either narrowing coverage or changing agent altogether. There was particular reluctance toward changing formulation of antibiotics, which were predominantly suggestions to transition from IV to oral antibiotics. This is likely attributed to the concerns of using oral antibiotics while the patient is still in an ICU environment due to concerns around gut absorption and guaranteed drug delivery (21). Although high-quality studies have shown no adverse effect of early oral antibiotic use for severe urinary tract infections or community-acquired pneumonias, dedicated high-quality studies in ICU patients are currently lacking which may make clinicians wary of switching to oral antibiotics particularly while the patient may still be on oxygen therapy or low-dose vasopressors (22,23).
Given this is a retrospective cohort study, there may be unmeasured confounding influencing the likelihood of ASP suggestions and/or acceptance. Our particular ASP implementation format may be unique to our center or the model of care of critically ill patients, and therefore, the findings may not apply to a broad intensive care population employing different ASP formats. Specifically, hospitals, regions, and countries with higher or lower antimicrobial resistance profiles may have different approaches to ICU stewardship to achieve more conservative or liberal antibiotic use targets.
Our study describes the successful implementation of an ASP program within an intensive care setting with a broad and diverse population. We show stable acceptance rates of suggestions over time despite an increasing need for stewardship reviews over a decade-long study period. Our findings support the need for a persistent presence of audit-and-feedback over time with a variety of patient-, infection-, and treatment-related characteristics that may impact the decision of the intensivist to accept or reject ASP suggestions in critically ill patients. Further work is needed to foster acceptance of suggestions to narrow antibiotic treatment and to improve uptake of suggestions for infection sources such as pneumonia that are typically treated without pathogen confirmation and identification.
1. Pope SD, Dellit TH, Owens RC, et al. Infectious Diseases Society of America; Society for Healthcare Epidemiology of America: Results of survey on implementation of Infectious Diseases Society of America and Society for Healthcare Epidemiology of America guidelines for developing an institutional program to enhance antimicrobial stewardship. Infect Control Hosp Epidemiol. 2009; 30:97–98
2. Broom J, Broom A, Plage S, et al. Barriers to uptake of antimicrobial advice in a UK hospital: A qualitative study. J Hosp Infect. 2016; 93:418–422
3. Elligsen M, Walker SA, Pinto R, et al. Audit and feedback to reduce broad-spectrum antibiotic use among intensive care unit patients: A controlled interrupted time series analysis. Infect Control Hosp Epidemiol. 2012; 33:354–361
4. Diaz Granados CA. Prospective audit for antimicrobial stewardship in intensive care: Impact on resistance and clinical outcomes. Am J Infect Control. 2012; 40:526–529
5. Katsios CM, Burry L, Nelson S, et al. An antimicrobial stewardship program improves antimicrobial treatment by culture site and the quality of antimicrobial prescribing in critically ill patients. Crit Care. 2012; 16:R216
6. Malani AN, Richards PG, Kapila S, et al. Clinical and economic outcomes from a community hospital’s antimicrobial stewardship program. Am J Infect Control. 2013; 41:145–148
7. Lindsay PJ, Rohailla S, Taggart LR, et al. Antimicrobial stewardship and intensive care unit mortality: A systematic review. Clin Infect Dis. 2019; 68:748–756
8. Peragine C, Walker S, Simor A, et al. Impact of a comprehensive antimicrobial stewardship program on institutional burden of antimicrobial resistance: A 14-year controlled interrupted time-series study. Clin Infect Dis. 2019ciz1183
9. Elligsen M, Walker SA, Simor A, et al. Prospective audit and feedback of antimicrobial stewardship in critical care: Program implementation, experience, and challenges. Can J Hosp Pharm. 2012; 65:31–36
10. Palmay L, Elligsen M, Walker SA, et al. Hospital-wide rollout of antimicrobial stewardship: A stepped-wedge randomized trial. Clin Infect Dis. 2014; 59:867–874
11. Bartlett JM, Siola PL. Implementation and first-year results of an antimicrobial stewardship program at a community hospital. Am J Health Syst Pharm. 2014; 71:943–949
12. Toth NR, Chambers RM, Davis SL. Implementation of a care bundle for antimicrobial stewardship. Am J Health Syst Pharm. 2010; 67:746–749
13. Li Z, Cheng B, Zhang K, et al. Pharmacist-driven antimicrobial stewardship in intensive care units in East China: A multicenter prospective cohort study. Am J Infect Control. 2017; 45:983–989
14. Waters CD. Pharmacist-driven antimicrobial stewardship program in an institution without infectious diseases physician support. Am J Health Syst Pharm. 2015; 72:466–468
15. Mertz D, Brooks A, Irfan N, et al. Antimicrobial stewardship in the intensive care setting–A review and critical appraisal of the literature. Swiss Med Wkly. 2015; 145:w14220
16. Sawyer RG, Claridge JA, Nathens AB, et al. STOP-IT Trial Investigators: Trial of short-course antimicrobial therapy for intraabdominal infection. N Engl J Med. 2015; 372:1996–2005
17. Pugh R, Grant C, Cooke RPG, et al. Short-course versus prolonged-course antibiotic therapy for hospital-acquired pneumonia in critically ill adults. Cochrane Database Syst Rev. 2015; 2015:CD007577
18. Lachiewicz AM, Hauck CG, Weber DJ, et al. Bacterial infections after burn injuries: Impact of multidrug resistance. Clin Infect Dis. 2017; 65:2130–2136
19. Lawrence K, Kollef M. Antimicrobial stewardship in the intensive care unit advances and obstacles. AJRCCM. 2009; 176:434–438
20. Klein Klouwenberg PM, Ong DS, Bos LD, et al. Interobserver agreement of Centers for Disease Control and Prevention criteria for classifying infections in critically ill patients. Crit Care Med. 2013; 41:2373–2378
21. Béïque L, Zvonar R. Addressing concerns about changing the route of antimicrobial administration from intravenous to oral in adult inpatients. Can J Hosp Pharm. 2015; 68:318–326
22. Oosterheert JJ, Bonten MJ, Schneider MM, et al. Effectiveness of early switch from intravenous to oral antibiotics in severe community acquired pneumonia: Multicentre randomised trial. BMJ. 2006; 333:1193
23. Pohl A. Modes of administration of antibiotics for symptomatic severe urinary tract infections. Cochrane Database Syst Rev. 2007; 2007:CD003237