Stewardship processes were compared across 123 hospitals that differed on a risk-adjusted post-discharge antibiotic use metric. Low-performing hospitals were less likely than high-performing hospitals to report routine interactions between their stewardship physician and pharmacist(s) (OR 0.12, 95% CI 0.03–0.55) and to have local antibiotic-prescribing guidelines (OR 0.21, 95% CI 0.05–0.93)

A surveillance system for measuring patient-level antimicrobial adverse drug events (ADE) may support stewardship activities, however, design and implementation questions remain. In this national survey, stewardship experts favored simple, laboratory-based ADE definitions although there were tensions between feasibility, ability to identify attribution without chart review, and importance of specific ADE.

Background: CAP is often inappropriately treated with agents active against multidrug-resistant organisms (MDRO; methicillin-resistant S. aureus [MRSA] and P. aeruginosa [PSA]) and for prolonged duration. We assessed the relationship between antibiotic use with ATS/IDSA guideline-unjustified empiric and definitive MDRO therapy and prolonged duration in non-ICU inpatients with CAP at 105 VA Medical Centers. Methods: From VA Corporate Data Warehouse data, we identified patients with discharge ICD-10-CM codes consistent with CAP from 1/2022-3/2023, excluding cases with 14 days of antibiotic therapy, ICU admission, concurrent infections, or severe immunocompromise. We considered as jultified empiric (≤third day of hospitalization) therapy: anti-MRSA therapy for patients with prior positive MRSA cultures, anti-PSA therapy for patients with prior positive PSA cultures, and both anti-MRSA & anti-PSA therapy in patients with severe pneumonia and intravenous antibiotics in the prior 3 months. Definitive (>third day of hospitalization) anti-MDRO therapy was considered unjustified in patients who had achieved clinical stability and whose cultures did not grow MRSA or PSA. Prolonged duration (>6 days of therapy) was unjustified if patients were clinically stable or discharged by day 5. Results: The median age of the 29,260 patients was 75 (IQR 69,81); 4.6% were women. While 33% and 22% of patients received empiric or definitive MDRO therapy, such therapy was jultified in 12% and 0.5%, respectively. Median facility use of empiric and definitive MDRO therapy was 31% (IQR 25%,38%) and 20% (15%,23%), respectively (Figure 1); this use was unjustified in 89% (85%,93%) and 100% (100%,100%), respectively. Pearson’s correlation coefficient between MDRO therapy and rates of unjustified empiric and definitive MDRO therapy for CAP was 0.54 and 0.61, respectively (Figure 2). Although 99% of patients were discharged or stable by day 5, 42% received prolonged therapy. The median frequency of prolonged therapy was 39% (33%,48%); facility rates of prolonged therapy had a correlation of 0.56 with total antibiotic use and 0.46 with MDRO therapy (Figure 3). Discussion: Based on electronic documentation, we identified 1) substantial opportunities to reduce unjustified anti-MDRO therapy and the duration of therapy in hospitalized non-ICU patients with CAP; 2) a moderate correlation of unjustified anti-MDRO therapy with increased MDRO antibiotic use and of prolonged duration of therapy with increased total and MDRO antibiotic use. The correlation of lower quality prescribing with increased antibiotic use provides further impetus for tools such as dashboards (Figure 4) to assist antibiotic stewards in designing and monitoring interventions to reduce unjustified therapy.

Background: Approximately 40% of all antibiotics related to an acute-care hospital stay are prescribed at the time of hospital discharge. However, there is no metric to compare hospitals on their antibiotic-prescribing at this transition of care. In this study, we sought to build a risk-adjusted metric for comparing hospitals on their overall post-discharge antibiotic use. Methods: This was a retrospective study across all acute-care admissions within the Veterans Health Administration during 2018-2021. For patients discharged to home or self-care, data on antibiotics administered while inpatient and those prescribed at discharge were collected. To predict post-discharge antibiotic use (days of therapy, DOT), we built a zero-inflated negative binomial model with a random intercept for each VA medical center. Data were split into training and testing sets to measure model performance and absolute error. Covariates included patient demographics, medical specialty at discharge, comorbidities, discharge diagnoses of infection, and the length of inpatient antibiotic therapy. Outliers, defined as DOT ≥ 30, were excluded, and the predicted random intercept was used to determine hospital performance. To compare hospitals with a positive versus negative random intercept in our model (i.e. higher vs. lower than expected overall post-discharge use, respectively), we calculated mean total antibiotic duration (inpatient + post-discharge) for two uncomplicated infection types: community-acquired pneumonia (CAP) and skin and soft tissue infections (SSTI). Results: 1,804,400 patients were discharged to home or self-care across 130 hospitals. The mean age was 67.8 (SD 12.9), and 93.7% were male. Antibiotics were prescribed to 41.5% while hospitalized and 19.5% at discharge. The median number of post-discharge DOT among those prescribed post-discharge antibiotics was 7 (IQR 4-12). The predictive model detected post-discharge antibiotic use with fidelity, including accurate identification of any post-discharge antibiotic exposure (area under the precision-recall curve=0.97 ) and reliable prediction of the number of post-discharge DOT in those who were exposed (mean absolute error = 1.65; Figure 1). At negative versus positive random intercept hospitals (Figure 2), antibiotic duration for CAP and SSTIs was 7.3 versus 8.1 days (p < 0 .001) and 9.4 vs. 10.2 days (p < 0 .001), respectively. Conclusion: A model using electronically available data was able to accurately predict antibiotic use prescribed at hospital discharge. Hospitals with lower than expected overall post-discharge antibiotic use also prescribed shorter courses of antibiotic therapy for uncomplicated cases of CAP and SSTI, which may reflect more robust processes at these sites to reduce antibiotic overuse at discharge.

Disclosure: Michi Goto: Contracted Research – Merck

Background: Diagnosis and management of suspected urinary tract infection (UTI) in outpatient settings has been shown to be suboptimal. We previously developed a set of stewardship metrics for UTIs based on electronic health record (EHR) data (Antimicrobial Stewardship & Healthcare Epidemiology 2022;2 suppl 1:S5–S6. doi:10.1017/ash.2022). A tier-based approach was used to more fully capture antibiotic use associated with genitourinary (GU) symptoms and diagnoses. Herein we report a preliminary analysis of validity and reliability of these metrics based on chart abstraction. Methods: The study cohort consisted of patients who visited Veterans Affairs emergency departments or primary care clinics between 2015 and 2022 and who had a GU diagnosis based on International Classification of Disease, Tenth Revision (ICD-10) codes, divided into 3 categories: tier 1 (antibiotics always indicated), tier 2 (antibiotics sometimes indicated), and tier 3 (antibiotics not indicated). Visits related to urological procedures, nontarget settings, or concomitant non-GU infections were excluded. Cases were randomly sampled for manual review from within 8 strata based on tier, use of antibiotics, and visit type. An infectious disease physician and pharmacist abstracted charts using a standardized data-collection instrument. Clinical judgments regarding diagnosis and treatment were recorded on a Likert scale without knowledge of how the patient was managed. The intraclass correlation coefficient (ICC) was used to estimate interrater reliability. Results: To date, 148 cases have been reviewed (50 by both reviewers). Mean (SD) age was 67.5 (15.3) years and 12.2% were female. In a majority of tier 1 and 2 visits in which antibiotics were given, the reviewers found evidence for GU infection (69.7%) and favored prescribing of antibiotics (60.6%) (Table). In contrast, most patients in the tier 3 category who received antibiotics were judged to have noninfectious conditions (eg, benign prostatic hypertrophy) and to not require antibiotics. In the subset of records examined by both reviewers, the interrater reliability of judgments of whether antibiotics were warranted was good (ICC = .704). Conclusions: This preliminary validation provides support for a tier-based approach for stewardship metrics for GU conditions that relies upon electronic data to identify patients for whom antibiotics are generally not indicated.

Disclosures: None

We explored experiences and perceptions surrounding the Self-Stewardship Time-Out Program (SSTOP) intervention across implementation sites to improve antimicrobial use. Semistructured qualitative interviews were conducted with Antibiotic Stewardship physicians and pharmacists, from which 5 key themes emerged. SSTOP may serve to achieve sustainable promotion of antibiotic use improvements.

To investigate factors that influence antibiotic prescribing decisions, we interviewed 49 antibiotic stewardship champions and stakeholders across 15 hospitals. We conducted thematic analysis and subcoding of decisional factors. We identified 31 factors that influence antibiotic prescribing decisions. These factors may help stewardship programs identify educational targets and design more effective interventions.

Background: Antibiotic resistance is pervasive in the Veterans’ Affairs (VA) healthcare system, with rates of fluoroquinolone and trimethoprim–sulfamethoxazole (TMP/SMX) resistance approaching 30% in E. coli urinary isolates. The efficacy of antimicrobial treatment is critically dependent on the susceptibility of the infecting pathogen; however, prescription decisions are often made empirically in practice. We analyzed susceptibility profiles of enteric gram-negative rods (Enterobacterales) from clinical urine cultures collected from ambulatory patients receiving care in VA clinics and emergency departments. Our goals were (1) to develop a predictive model to support choice of empiric antibiotics pending results of susceptibility testing and (2) to examine the relationship between past antibiotic exposures and susceptibility profiles to enhance understanding of antibiotic selective pressure. Methods: We obtained 265,076 positive cultures from 157,422 unique patients from 2015 to 2020. We trained random forest multinomial classifiers to estimate the risk of a positive urine culture isolate being resistant to the multinomial outcome: fluoroquinolone, TMP–SMX, cephalosporin, or any combination of these 3 agents. Data sources evaluated for model generation included demographics, comorbidities, trend and seasonal terms, treatment history for multiple antimicrobial treatments summarized using number of prescriptions in weekly intervals, and sample history summarized by number of resistant and susceptible cultures in weekly intervals. Using 5-fold cross validation, we assess the performance of the clinical prediction using the area under the receiver operating characteristic curve (AUC) for each multinomial outcome. In addition to prediction, we modeled the direct effect of treatment on resistance using multinomial group lasso (MGL). This method allows variable selection in variable groupings, such as all variables related to the fluoroquinolone treatment history, which allowed us to assess the effect of a patient’s complete course of treatment on resistance. Results: In cross-validation analysis, our random forest model was best at predicting outcomes with fluoroquinolone resistant phenotypes compared to non–fluoroquinolone-resistant phenotypes (Table 1). From MGL, we found that having a prescription for fluoroquinolone treatment 4–8 weeks prior to a urinalysis was positively associated with fluoroquinolone resistance and negatively associated with fluoroquinolone susceptible phenotypes (Fig. 1). Conclusions: Our results show that a patient’s sample and treatment history are highly predictive of a future resistance. Fluoroquinolone treatment is especially associated with increased risk of fluoroquinolone single- and multidrug resistances. A history of either fluoroquinolone or trimethoprim-sulfamethoxazole (TMP-SMX) treatment is a stronger indicator of a future resistant phenotype than cephalosporin or penicillin.

Funding: None

Disclosures: None

Background: Tracking antibiotic use is a core element of antimicrobial stewardship. We developed a set of metrics based on electronic health record data to support an outpatient stewardship initiative to improve management of urinary tract infections (UTIs) in Veterans’ Affairs (VA) emergency departments (EDs) and primary care clinics. Because UTI diagnostic codes only capture a portion of genitourinary (GU)-related antibiotic use, a tier-based approach was used to evaluate practices. Methods: Metrics were developed to target practices related to antibiotic prescribing and diagnostic testing (Table 1). GU conditions were divided into 3 categories: tier 1, conditions for which antibiotics are usually or always indicated; tier 2, conditions for which antibiotics are sometimes indicated; and tier 3, conditions for which antibiotics are rarely or never indicated (eg, benign prostatic hypertrophy with symptoms). Patients with visits related to urological procedures, nontarget providers, and concomitant non-GU infections were excluded. Descriptive analyses included calculation of the correlation matrix for the 7 metrics and the construction of box plots to display interfacility variability. Results: Metrics were calculated quarterly for 18 VA medical centers, including affiliated clinics, in a western VA network, from July 2018 to June 2020 (Table 1). Tier 3 GU conditions accounted for 1,276 of 11,840 (11%) of GU-related antibiotic use. Metrics 1 and 6b were strongly correlated with each other and were also positively correlated with metrics 2 and 5 (coefficients > 0.5) (Fig. 1). Substantial interfacility variation was observed (Fig. 2). Conclusions: Stewardship metrics for suspected or documented UTIs can identify opportunities for practice improvement. Broadly capturing GU conditions in addition to UTIs may enhance utility for performance feedback. Antibiotic prescribing for tier 3 GU conditions is analogous to unnecessary antibiotic use for acute, uncomplicated bronchitis and upper respiratory tract infections.

Funding: None

Disclosures: None

Effective stewardship strategies such as an “antibiotic timeout” to encourage prescriber reflection on the use of broad-spectrum antibiotics are critical to reduce the threat of multidrug-resistant organisms. We sought to understand the facilitators and barriers of the implementation of the Antibiotic Self-Stewardship Timeout Program (SSTOP), which used a template note integrated into the electronic health record system to guide decision making regarding anti- methicillin-resistant S. aureus (MRSA) therapy after 3 days of hospitalization. We conducted interviews at 10 Veterans’ Affairs medical centers (VAMCs) during the preimplementation period (N = 16 antibiotic stewards) and postimplementation (N = 13 antibiotic stewards) ~12 months after program initiation. Preimplementation interviews focused on current stewardship programs, whereas postimplementation interviews addressed the implementation process and corresponding challenges. We also directly asked about the impact of COVID-19 on stewardship activities at each facility. Interviews were transcribed and analyzed using consensus-based inductive and deductive coding. Codes were iteratively combined into barrier and facilitator groupings. Barriers identified in the preimplementation interviews included challenges with staffing, the difficulties of changing prescribing culture, and academic affiliates (eg, rotating physician trainees). Facilitators included intellectual support (eg, providers who understand the concept of stewardship), facility support, individual strengths of antibiotic stewards (eg, diplomacy, strong relationships with surgeons), and resources such as VA policies mandating stewardship. By the postimplementation phase, all sites reported a high volume of COVID-19 cases. Additional demands were placed on infectious disease providers who comprise the antibiotic stewardship teams, which complicated the implementation of SSTOP. Many barriers and facilitators mentioned were similar to those identified during preimplementation interviews. Staffing problems and specific providers not “getting it [stewardship activities]” continued, whereas facilitators centered around strong institutional support. Specific pandemic-related barriers included slow down or stoppage of stewardship activities including curbing of regular MRSA screening practices, halting weekly stewardship rounds, and delaying stewardship committee planning. Pandemic-specific staffing problems occurred due to the need for “all hands on deck” and challenges with staff working from home, as well as being pulled in multiple directions, (eg, writing COVID-19 policies). Furthermore, an increase in antibiotic use was also reported at sites during COVID-19 surges. Our findings indicate that SSTOP implementation met with barriers at most times; however, pandemic-specific barriers were particularly powerful. Sites with strong staffing resources were better equipped to deal with these challenges. Understanding how the program evolves with subsequent COVID-19 surges will be important to support the broad implementation of SSTOP.

Funding: No

Disclosures: None

Background: Hospitals are required to have antibiotic stewardship programs (ASPs), but there are few models for implementing ASPs without the support of an infectious disease (ID) specialist, defined as an ID physician and/or ID pharmacist. In this study, we sought to understand ASP implementation at hospitals within the Veterans’ Health Administration (VHA) that lack on-site ID support. Methods: Using a mandatory 2016 VHA survey, we identified acute-care hospitals that lacked an on-site ID specialist. For each hospital, antibiotic use (2018–2019) was quantified as days of therapy (DOT) per 1,000 days present, based on NHSN methodology for tracking all antibacterial agents. From July 2019 through April 2020, we conducted semistructured interviews with personnel involved in or affected by ASP activities at 7 qualifying hospitals. All interview transcripts were analyzed using thematic content analysis. Results: Of the 7 acute-care hospitals, 6 (86%) had a long-term care unit; 3 (43%) had an intensive care unit; and 2 (29%) had full-time employment equivalents dedicated to stewardship. Sites averaged 1,075 (SD, ±654) and 148 (SD, ±96) admissions per year in acute-care and long-term care, respectively. At the site-level, mean antibiotic use was 486 DOT (SD, ±98) per 1,000 days-present in acute-care and 207 DOT (SD, ±74) per 1,000 days present in long-term care. We interviewed 42 personnel across the 7 sites. Although sites reported using similar interventions to promote antibiotic stewardship, the shape of these interventions varied. The following 4 common themes were identified: (1) The primary responsibility for ASPs fell on the pharmacist champions, who were typically assigned multiple other non-ASP responsibilities. (2) The pharmacist champions were more successful at gaining buy-in for stewardship initiatives when they had established rapport with clinicians, but at some sites, the use of contract physicians and frequent staff turnover were potential barriers. (3) Some sites felt that having access to an off-site ID specialist was important for overcoming institutional barriers to stewardship and improving the acceptance of their stewardship interventions. (4) In general, stewardship champions struggled to mobilize institutional resources, which made it difficult to advance their programmatic goals. Conclusions: In this study of 7 hospitals without local ID support, we found that ASPs are largely a pharmacy-driven process. Remote ID support, if available, was seen as helpful for implementing stewardship interventions. These findings may inform the future implementation of ASPs in settings lacking local ID expertise.

Funding: No

Disclosures: None

Antibiotic prescribing practices across the Veterans’ Health Administration (VA) experienced significant shifts during the coronavirus disease 2019 (COVID-19) pandemic. From 2015 to 2019, antibiotic use between January and May decreased from 638 to 602 days of therapy (DOT) per 1,000 days present (DP), while the corresponding months in 2020 saw antibiotic utilization rise to 628 DOT per 1,000 DP.