Siebeneck, Eric

Abstract Background The BioFire® FilmArray® Gastrointestinal Pathogen Panel (GIPP) tests for 22 potential pathogens in stool, including Clostridioides difficile. C. difficile nucleic acid is detected via nested polymerase chain reaction (PCR) targeting toxin A (TcdA) and/or toxin B (TcdB) genes. While PCR-based diagnostic methods are highly sensitive, they lack specificity to differentiate colonization from C. difficile infection (CDI). Mayo Clinic Arizona (MCA) uses enzyme immunoassay (EIA) testing for toxin A/B and glutamate dehydrogenase antigens with reflex to PCR for discordance as the primary algorithm for diagnosing CDI; however, GIPP PCR testing circumvents this algorithm. When used in patients with low pretest probability for CDI, clinical false positives can lead to inappropriate diagnosis and treatment of colonization. GIPA Results in the Electronic Medical Record Pre-Intervention MCA Standalone C. difficile Testing Algorithm Abbreviations – EIA, enzyme immunoassay; GDH, glutamate dehydrogenase; NAAT, nucleic acid amplification test Methods We convened a multidisciplinary quality improvement team to reduce the detection and treatment of C. difficile colonization secondary to GIPP test results. Surveys were sent to providers and nurses as part of a formal gap analysis and included questions around ordering patterns, knowledge of appropriate tests, and associated costs. Baseline data was compiled from GIPP orders in MCA from 1/1/2022-9/30/2022. Results There were 1978 GIPP orders for 1694 patients during the study period (∼7 GIPPs/day). Most orders (65%) originated from ambulatory settings. The most common positive target was C. difficile (355/723, 49%), which resulted in treatment for suspected CDI in 94.9% (337/355) of cases. Of the 49 providers who responded to the survey, 86% ordered the GIPP at least monthly. Testing for C. difficile was the 2nd most common reason for ordering the GIPP. Most providers underestimated the cost of the panel and were unfamiliar with the differences between the GIPP and EIA testing algorithm. Nurses sought clarification on stool samples suitable for GIPP testing. Following education and modifications to ordering in the electronic health record, the decision was made to remove the C. difficile target from the GIPP. Provider Understanding of C. difficile Testing Methods GIPA Results in the Electronic Medical Record Post-Intervention Conclusion Multiplex PCR panels are comprehensive but require careful implementation to facilitate diagnostic stewardship. Facilities with robust standalone C. difficile testing algorithms may consider removing the C. difficile target from the GIPP. Disclosures All Authors: No reported disclosures

Abstract Background Inappropriate use of carbapenems is a major driver of antimicrobial resistance. The intensive care unit (ICU) at Mayo Clinic Arizona had a higher meropenem (MER) days of therapy (DOT) per 1000 patient days (PD) compared to other Mayo Clinic Enterprise sites, despite low rates of Pseudomonal resistance to narrower spectrum anti-pseudomonal beta-lactams. Methods A multidisciplinary team of infectious diseases (ID) and critical care medicine (CCM) physicians and pharmacists was created to reduce inappropriate use of MER in the ICU. The baseline rate of appropriateness was determined by evaluating all MER days in the ICU during Q4 2019 and Q4 2020. Each MER day was designated as appropriate or inappropriate based on pre-determined criteria after review by at least one ID and one CCM member. Discordant days were evaluated by the whole team to achieve consensus. The team provided education to the CCM department on the initiative, baseline data, and strategies to reduce inappropriate MER use. The active intervention included prospective audit and feedback of all patients on MER performed by the ID clinical pharmacists Mon-Fri 0700-1530 in Q4 2022. DOT/1000 PD was assessed for the pre-intervention and intervention periods. Results The baseline rate of inappropriate MER use in the ICU was 52% (442/846 days). The two most common reasons for inappropriate MER use in the baseline period were: susceptible organism for which a narrower agent could have been used (93/442, 21%), and infectious workup/cultures negative (89/442, 20%). During the intervention period, the inappropriate rate of MER use was reduced to 23%, below the pre-determined goal of 30%. The MER DOT/1000 PD in the baseline period were 103.9 (Q4 2019) and 151.5 (Q4 2020) versus 66.1 (Q4 2022) in the intervention period. There was no difference in ICU sepsis-related mortality in the pre-intervention vs intervention period (13.4 vs 0%, p=0.04). Conclusion A multidisciplinary team approach, including education and prospective audit and feedback, reduced inappropriate MER use in the ICU of an academic medical center from 52 to 23%. An associated decrease in MER DOT/1000 patient days was also observed. Disclosures All Authors: No reported disclosures

Abstract Objective Hospital acquired infections (HAIs) are one of the top 10 leading causes of death within the United States. While current standard of HAI risk prediction utilizes only a narrow set of predefined clinical variables, we propose a graph convolutional neural network (GNN)-based model which incorporates a wide variety of clinical features. Materials and Methods Our GNN-based model defines patients’ similarity based on comprehensive clinical history and demographics and predicts all types of HAI rather than focusing on a single subtype. An HAI model was trained on 38 327 unique hospitalizations while a distinct model for surgical site infection (SSI) prediction was trained on 18 609 hospitalization. Both models were tested internally and externally on a geographically disparate site with varying infection rates. Results The proposed approach outperformed all baselines (single-modality models and length-of-stay [LoS]) with achieved area under the receiver operating characteristics of 0.86 [0.84–0.88] and 0.79 [0.75–0.83] (HAI), and 0.79 [0.75–0.83] and 0.76 [0.71–0.76] (SSI) for internal and external testing. Cost-effective analysis shows that the GNN modeling dominated the standard LoS model strategy on the basis of lower mean costs ($1651 vs $1915). Discussion The proposed HAI risk prediction model can estimate individualized risk of infection for patient by taking into account not only the patient’s clinical features, but also clinical features of similar patients as indicated by edges of the patients’ graph. Conclusions The proposed model could allow prevention or earlier detection of HAI, which in turn could decrease hospital LoS and associated mortality, and ultimately reduce the healthcare cost.