Amid global efforts to address the growing issue of antimicrobial resistance, antimicrobial stewardship programs (ASPs) have been promoted to drive appropriate antibiotic use (AU). 1-4 AU is a measure of antibiotic consumption used by ASPs to analyze and report how antibiotics are prescribed at their institution, which can be used to demonstrate the progress and value provided by ASPs.
Historically, the most commonly used AU metrics have been defined daily dose (DDD) or days of therapy (DOT) per patient admissions, per patient days (PD), or per days present. Reference Polk, Fox, Mahoney, Letcavage and MacDougall5,Reference Morris6 Although these metrics may be valuable for intrafacility comparisons, meaningful interfacility comparison may be limited when comparing facilities with differences in hospital epidemiology and patient populations. To facilitate AU standardization, reporting to the National Healthcare Safety Network, and benchmarking between similar facilities, the Centers for Disease Control and Prevention developed a new metric called the standardized antimicrobial administration ratio (SAAR). Reference van Santen, Edwards and Webb7 The SAAR is a ratio of observed to predicted AU. Predictive models are used to estimate the number of predicted DOT for given locations and antimicrobial categories. Various hospital and location-level factors are incorporated into these predictive models including hospital bed size, number of ICU beds, medical school affiliation, and location type. For example, a hospital with a transplant center designation would be expected to have a higher prevalence of opportunistic and nosocomial pathogens (eg, Pseudomonas aeruginosa), and more frequent use of broad-spectrum agents (eg, antipseudomonal β-lactams or APBLs) would be expected. Currently, bacterial burden is not directly adjusted for in calculations. Because broad-spectrum antibiotics are used to treat infections due to certain bacterial isolates, and the prevalence of such isolates can vary widely, it would be reasonable to adjust AU calculations for the proportion of such isolates at a particular hospital. Reference Al-Hasan, Winders, Bookstaver and Justo8
Methods
The AU in DOT per 1,000 PD and microbiologic data from 2015 and 2016 were requested from 32 hospitals in the Southeastern Research Group Endeavor-45 (SERGE-45) research network located in the southeastern United States. Hospitals in the SERGE-45 research network range from small community hospitals to large academic medical centers. Data were collected on characteristics of the hospitals and their ASPs, including formulary agents, protection criteria, and prospective audit and feedback. Protection criteria are defined as prior authorization or preapproved indications.
Antimicrobial use of APBLs included total AU of piperacillin-tazobactam, ceftazidime, cefepime, meropenem, doripenem, and imipenem-cilastatin at each hospital. The AU of carbapenems included total AU of meropenem, doripenem, imipenem-cilastatin, and ertapenem. The AU of anti-methicillin resistant Staphylococcus aureus (MRSA) agents included total AU of vancomycin, daptomycin, and linezolid. The AU of anti–vancomycin–resistant enterococci (VRE) agents included the total AU of daptomycin and linezolid. The prevalences of bacterial isolates at each hospital were calculated utilizing antibiogram data as follows: The prevalence of P. aeruginosa was the P. aeruginosa isolate count divided by the total gram-negative isolate count. The prevalence of extended spectrum β-lactamase (ESBL)–producing bacteria was the ESBL isolate count divided by the total gram-negative isolate count. The prevalence of MRSA was the MRSA isolate count divided by the total gram-positive isolate count. And the prevalence of VRE was the VRE isolate count divided by the total gram-positive isolate count. The denominators were selected to include all possible organisms of concern when choosing to treat with agents that target gram-positive or gram-negative bacteria. Adjusted AU (a-AU) by microbiological burden was calculated as previously proposed by Al-Hasan et al. Reference Al-Hasan, Winders, Bookstaver and Justo8 For example, a-AU APBL is the AU APBL divided by (the prevalence of P. aeruginosa at that hospital divided by the average prevalence of P. aeruginosa across all hospitals in network). Similar formulas were used to calculate the a-AU of carbapenems, anti-MRSA agents, and anti-VRE agents based on the prevalences of ESBLs, MRSA, and VRE, respectively. Reference Al-Hasan, Winders, Bookstaver and Justo8
Only hospitals submitting all necessary data were included in each independent analysis. Hospitals were ranked by AU and a-AU from lowest to highest in each antimicrobial category in 2015 and 2016. The rankings of each hospital were compared using AU and a-AU for various antimicrobial categories in both years. To quantify the magnitude of change in rankings between AU and a-AU, the proportion of hospitals that had ≥2 positions change in ranking was calculated. The proportion of hospitals that underwent a change in quartile of ranking based on AU and a-AU (ie, from the first to second quartiles or vice versa) has also been reported.
To examine the impact of hospital size on change in rankings, the χ2 test was used to compare differences in rankings between hospitals with ≤200 beds, 201–500 beds, and >500 beds. In this analysis, the rankings of all reported antimicrobial categories in both years of study were evaluated in each hospital (up to 8 categories per hospital). The level of significance for statistical testing was defined as P < .05 (2-sided). REDCap version 7.3.4 software was used for data collection and management. Excel 2016 software (Microsoft, Redmond WA) and JMP Pro version 13.0 software (SAS Institute, Cary, NC) were used for statistical analyses.
Results
The AU in DOT per 1,000 PD and microbiologic data were available for analysis for at least 1 year from 26 hospitals. However, 6 hospitals with comparable characteristics were only able to submit DDD data and were excluded. Participating hospitals were assigned to numbers from 1 to 26 to maintain anonymity (Table 1). Hospitals varied in bed capacity: 7 (27%) had ≤200 beds, 10 (38%) had 201–500 beds, and 9 (35%) had >500 beds. Moreover, 21 hospitals (81%) had formal ASPs during the study period. The median full-time equivalent for stewardship pharmacists was 1 and for physician champions was 0.25. All hospitals utilized a certain degree of formulary restrictions, protection criteria, or prospective audit and feedback for antimicrobial agents, most commonly carbapenems.
Table 1. Hospital Characteristics
Note. CLSI, Clinical and Laboratory Standards Institute; ATM, aztreonam; CAZ, ceftazidime; DAP, daptomycin; DOR, doripenem; ETP, ertapenem; FEP, cefepime; IPM, imipenem-cilastatin; LZD, linezolid; MEM, meropenem; TZP, piperacillin-tazobactam.
a For most of 2015–2016.
Hospitals had a median AU of 143 DOT per 1,000 PD for APBL, 32 for carbapenems, 120 for anti-MRSA agents, and 10 for anti-VRE agents. The average prevalences of P. aeruginosa, ESBLs, MRSA, and VRE across participating hospitals are shown in Table 2. After adjustment for microbiological burden, the median a-AU was 144 DOT per 1,000 PD for APBL, 25 for carbapenems, 112 for anti-MRSA agents, and 12 for anti-VRE agents.
Table 2. Prevalence of Pertinent Bacteria Across Participating Hospitals in 2015 and 2016
Note. ESBL, extended-spectrum β-lactamase; MRSA, methicillin-resistant Staphylococcus aureus; VRE, vancomycin-resistant enterococci.
Most hospitals in 2015 and 2016 moved ≥2 positions in the ranking in either direction using the a-AU of all antibiotic classes studied (Table 3, Figs. 1 and 2, and Supplementary Figs. 1–6 online). The use of a-AU resulted in a shift in quartile of hospital ranking for many hospitals as well (Table 3). For example, 4 hospitals (17%) moved from first (lowest use) to second quartiles or vice versa, 4 (17%) moved between second and third, and 4 (17%) between the third and fourth quartiles for AU of APBL in 2015 after an the adjustment for microbiological burden.
Table 3. Comparison of Antibiotic Use and Adjusted Antibiotic Use by Microbiological Burden
Note. APBL, antipseudomonal β-lactams; AU, antibiotic use; DOT, days of therapy; MRSA, methicillin resistant Staphylococcus aureus; PD, patient days; VRE, vancomycin-resistant enterococci.
a Relative change in AU = (a-AU – AU)/AU × 100.
b Absolute change in AU = a-AU – AU.
Fig. 1. Actual antibiotic use versus adjusted antibiotic use of antipseudomonal β-lactams in 2016 ranked by actual antibiotic use. Note. a-AU, adjusted antibiotic use; APBL, antipseudomonal β-lactams; AU, antibiotic use; DOT, days of therapy; PD, patient days.
Fig. 2. Actual antibiotic use versus adjusted antibiotic use of carbapenems in 2016 ranked by actual antibiotic use. Note. a-AU, adjusted antibiotic use; AU, antibiotic use; DOT, days of therapy; PD, patient days.
When ranked from lowest to highest AU and a-AU for all antibiotic categories in both years, smaller hospitals were more likely to have an increase in hospital ranking: 31 of 52 (60%) for ≤200 beds versus 26 of 77 (34%) for 201–500 beds versus 24 of 68 (35%) for >500 beds (P = .007). Smaller hospitals were also less likely to have a decrease in hospital ranking: 14 of 52 (27%) for ≤200 beds versus 44 of 77 (57%) for 201–500 beds versus 35 of 68 (51%) for >500 beds (P = .002). This trend was most prominent for APBLs. Smaller hospitals were more likely to have an increase in hospital ranking based on APBL use compared to others (11 of 13 [85%] for ≤200 beds vs. 4 of 20 [20%] for 201–500 beds vs. 6 of 17 [35%] for >500 beds; P < 0.001) and less likely to have a decrease in ranking in this antimicrobial category (2 of 13 [15%] for ≤200 beds vs. 14 of 20 [70%] for 201–500 beds vs. 11 of 17 [65%] for >500 beds; P = 0.005).
Discussion
Adjusting AU by microbiological burden greatly changed how hospitals compared to each other with respect to use of broad-spectrum antimicrobial agents. Most hospitals in this study moved ≥2 positions in the ranking of AU for broad-spectrum antimicrobial agents and nearly one-half shifted in the quartile of hospital ranking. The greatest relative change in a-AU was observed for anti-VRE agents, likely due to relatively lower use of these agents. We propose that adjusting for the microbiological burden of certain bacterial isolates allows for a more balanced comparison of AU among hospitals at the national level or within a regional network. This comparison may be a more fair and may show individual hospitals where they have undiscovered problems.
Smaller hospitals (≤200 beds) were more likely to see an increase in position in the ranking than larger hospitals. This finding was more profound when APBL use was evaluated. This was conceivable given the relatively lower prevalence of resistant bacteria at these hospitals. Previous studies have demonstrated lower prevalence of P. aeruginosa among bloodstream and other clinical isolates in small community-hospitals than larger referral tertiary care medical centers. Reference Al-Hasan, Eckel-Passow and Baddour9,Reference Baddour, Hicks and Tayidi10 After adjustment for microbiological burden, there is even less justification for high use of APBLs in small hospitals compared to larger ones. Similar differences were not seen among hospitals with prospective audit and feedback or protection criteria for at least 1 agent of each class of APBL (ie, cephalosporin, carbapenem, penicillin) and other hospitals without such restrictions (results not shown).
Adjustment for institutional characteristics such as bed size, ICUs, and complexity of patient population has been implemented for several stewardship metrics, including incidence of hospital-onset Clostridioides difficile infection. Although it is convenient to use the same formula for AU as suggested in the SAAR, an association between the need for broad-spectrum antibiotics and hospital characteristics remains to be determined. In fact, a few studies looking at incorporating additional patient-specific factors into predictive models for AU have been suggested. Reference Yu, Moisan and Tartof11,Reference Goodman, Pineles and Magder12 From an antimicrobial stewardship standpoint, antibiotics are used to treat infections caused by specific bacteria. Adjustment of AU by microbiological burden emphasizes this concept and encourages targeted antimicrobial therapy based on actual or predicted microbiologic etiology of infections rather than broad-spectrum empiricism solely based on clinical indications. To encourage this method of prescribing, patient-specific risk factors for resistant pathogens can be added into treatment guidelines, and education on AU and a-AU could be provided to front-line prescribers. Calculation of the prevalence of bacterial isolates is readily available using institutional antibiograms. This makes adjustment of AU based on bacterial burden relatively simple and convenient. In addition, it may be more up to date than adjustments based solely on hospital characteristics because antibiograms are updated annually. Although a recent study has also shown that MRSA prevalence has an effect on hospital-level anti-MRSA agent use, validation of this novel method of adjusting AU and comparison with other formulas for adjustment would be valuable in future studies. Reference Livorsi, Nair and Lund13
Contrary to most traditional stewardship metrics, this novel metric may encourage healthcare providers to obtain appropriate cultures. Obtaining blood cultures with subsequent growth of P. aeruginosa, for example, justifies the use of APBLs in patients with sepsis. Empirical antimicrobial therapy for “culture-negative” infections due to lack of effort to obtain cultures or obtaining low-yield cultures increases use of broad-spectrum agents without documentation of microbiological burden. On the other hand, the incidence of central-line–associated bloodstream infections and hospital-onset C. difficile infections would increase with more blood cultures and C. difficile tests obtained. This may discourage clinicians from obtaining appropriate cultures to avoid a heavy burden of hospital-acquired infections or the financial repercussions of publicly reported metrics.
This study implemented a novel method to adjust AU based on microbiological burden. The inclusion of 26 hospitals in 8 states adds strength to this work. However, this study has several limitations. All hospitals were from the southeastern United States, and these results may not be generalizable to hospitals in other areas with very high rates of multidrug-resistant organisms. In addition, microbiologic data were collected from antibiograms, which may be affected by culture frequency, susceptibility testing, and selective reporting of microorganisms. However, if a hospital regularly sends more cultures than others, the percentage of the organism would likely remain the same since both numerator and denominator are increased. The ability of microbiology labs to designate isolates as ESBL producing may also differ between hospitals. In addition, antibiograms do not take into account cultures taken at outside hospitals. Finally, not all bacteria in antibiograms are clinically relevant. Many urinary and respiratory isolates may represent colonization. It would be useful to compare microbiological burden based on overall and sterile antibiograms in future investigations. A limitation of the definitions is the overlap between carbapenems and APBLs and between anti-MRSA and anti-VRE agents. Hospitals would want to look at their rankings in all categories to better understand their use. In addition, similar to other metrics, a-AU does not measure appropriateness of therapy, although it may be a step in the right direction.
In conclusion, adjusting AU by microbiological burden allows for a more balanced comparison among hospitals that have different rates of organisms and antimicrobial resistance patterns. As shown, the a-AU considerably changes how hospitals compare among each other.
Acknowledgments
We thank all participating hospitals that submitted data: Blount Memorial Hospital, Candler Hospital, Duke University Hospital, East Alabama Medical Center, Grady Health System, Lexington Medical Center, McLeod Regional Medical Center and affiliated hospitals, Mt. Pleasant Hospital, Medical University of South Carolina, Nash UNC Health Care, Orlando Health and affiliated hospitals, Prisma Health Midlands hospitals, Providence Health, Roper Hospital, St Francis Hospital, St Joseph’s Hospital, St Mary’s Hospital, University of Arkansas for Medical Sciences Medical Center, University of Mississippi Medical Center, University of Tennessee Medical Center, Vanderbilt University Medical Center, Vidant Medical Center, Wake Forest Baptist Medical Center, Wellington Regional Medical Center, and Wilson Medical Center.
Financial support
No financial support was provided relevant to this article.
Conflicts of interest
P.B.B. reports that he is on the speakers bureau for bioMerieux, that he is a content developer and speaker for T.R.C. Healthcare, and that he is a content developer and speaker for FreeCE.com. B.M.J. reports that he is on the speakers bureaus for Allergan, Tetraphase, and Paratek. E.B.C. reports being on the speakers bureau for Merck and Paratek. C.M.B. reports receiving honoraria and grant funding from Merck, grant funding from ALK Abello, and honoraria from Paratek Pharmaceuticals, bioMerieux, and La Jolla Pharmaceutical Company. G.M.G. reports having consulted for ASHP Consulting. MMS reports being on the speakers bureau for Accelerate Diagnostics. H.R.W. reports being on the speakers bureau for bioMerieux. S.H.M. is currently an employee of Accelerate Diagnostics. All other authors report no conflicts of interest relevant to this article.
Supplementary material
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2020.1285
