Challener, Douglas W.

Artificial intelligence (AI) has emerged as a promising adjunct in the diagnosis and management of infective endocarditis (IE), a disease characterized by diagnostic complexity and significant morbidity. Machine learning (ML) models such as SABIER and SYSUPMIE have demonstrated strong predictive accuracy for early IE diagnosis, embolic risk stratification, and postoperative mortality, surpassing traditional clinical scoring systems. In imaging, AI-enhanced echocardiography and advanced modalities like FDG-PET/CT offer improved sensitivity, specificity, and reduced inter-observer variability, potentially transforming clinical decision making. Additionally, AI-powered microbiological techniques, including MALDI-TOF mass spectrometry combined with ML and neural network-based metagenomic classifiers, show promise in rapidly identifying pathogens and predicting antimicrobial resistance. Despite encouraging early results, widespread adoption faces barriers, including data limitations, interpretability issues, ethical concerns, and the need for robust validation. Future directions include leveraging generative AI as clinical consultative tools, provided their capabilities and limitations are carefully managed. Ultimately, collaborative efforts addressing these challenges could transform IE care, enhancing diagnostic accuracy, clinical outcomes, and patient safety.

Abstract Purpose Data regarding the incidence, onset, and management of complex outpatient antimicrobial therapy–induced neutropenia (COIN) are limited, and pharmacist involvement in COIN management has not been reported. This report describes the incidence, onset, and management of COIN at a large academic medical center. Methods This multisite retrospective cohort study examined adult patients undergoing serum laboratory monitoring for antimicrobials over a 4-year period. Patients receiving chemotherapy or immunosuppressants were excluded. Data collected included the antimicrobial regimen, complete blood count, duration of antimicrobial therapy until outcome occurrence, infectious syndrome, type of intervention, outcome, and pharmacist involvement. Logistic regression was used to assess risk factors for COIN incidence. The primary outcome was the incidence of COIN by antimicrobial drug. Results From 4,261 treatment episodes, 161 cases of COIN were identified (3.8% COIN incidence). The most common antimicrobials associated with COIN were intravenous piperacillin/tazobactam, cefepime, and meropenem. The majority of COIN events were attributed to a combination of at least 2 antimicrobials. Piperacillin/tazobactam was significantly associated with higher odds of developing COIN (odds ratio [OR], 1.86; 95% confidence interval [CI], 1.04-3.34; P = 0.038), whereas ertapenem was associated with significantly lower odds of COIN (OR, 0.43; 95% CI, 0.24-0.77; P = 0.005). The median time to neutropenia diagnosis was 22 days (interquartile range, 14-29 days) from the time of inpatient antimicrobial start. Intervention for COIN occurred in 66.5% of cases, with neutrophil count recovering in 96.9% of these patients. Clinical pharmacists initiated 74.8% of interventions. Conclusion COIN can be effectively identified and managed by a multidisciplinary team reviewing routine laboratory monitoring, enabling most patients to safely complete antimicrobial treatment.

Abstract Background There is a lack of comprehensive, large cohort studies investigating the predictors for severity and hospitalization in human granulocytic anaplasmosis (HGA). Methods We conducted a retrospective cohort study including all cases of HGA identified by positive blood Anaplasma polymerase chain reaction (PCR) at the Mayo Clinic Laboratory, Rochester between 2011 and 2021. Multivariable logistic regression was performed to evaluate risk factors associated with hospitalization. Results A total of 465 cases were identified. Of those, 67% (n = 312) were managed in the outpatient setting. Hospitalized patients (n = 153, 33%) were more likely to be older (median age of 71 vs 61; P ≤ .001) and immunocompromised (17% vs 7%; P ≤ .001). The triad of leukopenia, thrombocytopenia, and transaminitis was observed in 24% of the cases, but was associated with risk of hospitalization (odds ratio [OR] 1.78, 95% confidence interval [CI]: 1.05–3.03; P ≤ .033). The presence of a coinfection did not impact mortality or hospitalization. Conclusions The risk factors for hospitalization in patients with HGA include altered mental status, higher absolute neutrophil count (ANC), advanced age, comorbidities, and immunosuppression. The classic hematological and biochemical profile may be absent in the majority of cases. Co-testing may be of higher benefit in select cases.

Background Patients enrolling in hospice care are asked their preferences for antibiotic treatment. There is little information regarding which patients are more likely to receive antibiotics. To better inform discussions, we sought to characterize the use of antibiotics in the outpatient hospice setting. Methods We performed a retrospective review of patients enrolled in outpatient hospice within the Mayo Clinic Health System from 1/1/2017 through 1/1/2023. We calculated what percent of patients received antibiotics based on Hospice Qualifying Condition (HQC), age at enrollment, sex, survival time, and Charlson Comorbidity Index and calculated adjusted odds ratios (aOR). We documented which antibiotics were prescribed based on HQC. Results Of 6452 patients identified, 1259 (19.5%) received antibiotic prescriptions. Cephalosporins were the most common class of antibiotics prescribed (22.8% of antibiotics prescribed), followed by fluoroquinolones (20.3%) and penicillin derivatives (14.9%). Patients with Pulmonary HQCs were most likely to receive antibiotics (28.6% aOR 1.85 [1.51-2.25]), those with neurologic HQCs were least likely (14.9% aOR 0.66 [0.53-0.83]). There was no difference of age for those receiving antibiotics (80.7 yr) vs those not receiving (80.3 yr [ P = 0.25]), or for sex (male vs female aOR 1.10 [0.96-1.26). 4.6% of those surviving <7d received antibiotics (aOR .22 [0.17-0.29]), compared to 47.7% of those surviving >6mo (aOR 4.46 [3.50-5.69]). Discussion 19.5% of patients will receive antibiotics during their hospice course, with more frequent prescriptions in those surviving longer periods and those enrolled for pulmonary conditions. It is important to clarify patient preferences regarding antimicrobial utilization at the time of hospice enrollment.

Abstract Background At the time of hospice enrollment, patients are asked whether they would potentially want antibiotics. This is little information regarding which patients are more likely to receive these antibiotics or the impact on comfort or longevity. To help guide discussions with patients and families and target future stewardship interventions, we sought to characterize the use of antibiotics in the outpatient hospice setting. Methods We performed a retrospective review of patients enrolled in hospice with the Mayo Clinic Health System -Midwest from 1/1/2017 through 1/1/2023. For those with documented Hospice Qualifying Condition (HQC), we assessed age at enrollment, gender, survival time (ST), and Charleston Comorbidity Index (CCI). We documented which antibiotics were prescribed and compared the rate of prescription based on the above characteristics. Results We identified 6452 patients with identifiable HQC. Of these, 1281 patients (19.8%) received 4173 antibiotic prescriptions. After removing refilled prescriptions, there were 2037 new prescriptions. Cephalosporins were the most common class of antibiotics prescribed (24.4% of antibiotics prescribed), followed by fluoroquinolones (19.9%), penicillin derivatives (14.4%), and trimethoprim (10.4%) (fig 1). Pulmonary (28.6%) and GI (31.9%) HQCs were most likely to receive antibiotics, neurologic was least likely (fig 2). There was no difference of age for those receiving antibiotics (80.6 yr) vs those not receiving (80.3 yr [p=0.53]), or for gender (male vs female OR 0.97 [0.86-1.10). CCI was higher in those patients who received antibiotics (4.3) than those not receiving (3.93, p&lt; 0.001). Patients were more likely to receive antibiotics the longer they survived on hospice (4.7% surviving &lt; 7d, 47.7% surviving &gt;6mo – fig 3). Conclusion Antibiotics are an important pharmaceutical tool in the hospice setting. Nearly 20% of patients in hospice will receive antibiotics during their hospice course, with higher chance in those surviving longer periods and those enrolled for GI or pulmonary reasons. Discussions of antibiotic are crucial for these groups, and antimicrobial steward efforts should consider focusing on these patients. Disclosures All Authors: No reported disclosures

Abstract Background The COVID-19 pandemic has underscored challenges with patient adherence to medical directives, particularly vaccine acceptance. While sociodemographic factors have been linked to non-COVID-19 vaccine hesitancy, the relationships between patient experience, the COVID-19 pandemic, and adherence to routine vaccinations are uncertain.Figure 1:Yearly influenza vaccine refusal, influenza vaccine administrations, and overall patient satisfaction since establishing care with a primary care provider from 2017-2023.Panel A (Influenza Refusals): the number of vaccine refusals among individuals who are Intermittently Vaccinated and Never Vaccinated for seasonal influenza since establishing care with a primary care provider over time.Panel B (Influenza Vaccinations): the number of vaccines administered over time since establishing care with a primary care provider among those who are Always Vaccinated and Intermittently Vaccinated for seasonal influenza.Panel C (Overall Satisfaction): the number of patients who indicated that they were maximally satisfied with their overall care following any interaction with the healthcare system over time. Methods We retrospectively analyzed influenza vaccination records of adult primary care patients in Olmsted County, Minnesota. Vaccination history was examined during two phases: Pre-Pandemic (1/1/2017—2/28/2020) and Pandemic-Plus (3/1/2020—12/31/2023). Vaccination status was defined as “always vaccinated” (AV), “never vaccinated” (NV), or “intermittently missed years of vaccinations” (IV) for seasonal influenza. Patient dissatisfaction was determined if patients negatively ranked their overall quality of care on standard patient experience surveys. We explored the relationship between influenza vaccination rates, healthcare engagement, and patient satisfaction.Figure 2:Multinomial logistic regression of patient satisfaction on vaccination status across the phases of the COVID-19 Pandemic, adjusted for engagement in the health care system, participation, and area of deprivation index (ADI) for patients that were (A) intermittently vaccinated compared to always vaccinated and (B) never vaccinated comparted to always vaccinated. The “Pre-pandemic phase” refers to influenza immunization data and patient satisfaction data prior to March 2020. The “Pandemic-plus phase” refers to immunization data and patient satisfaction data from March 2020 to January 2024. Low ADI ranking refers to areas with a low level of deprivation, whereas high ADI ranking refers to areas with a high level of deprivation. Individuals were considered low if they came from an area with a state ADI ranking from 1 to 5 and were considered high if they came from an area with a state ADI ranking from 6 to 10. Results From 1/1/2017 to 12/31/2023, 21% of individuals were AV (n=36,632), 52% were IV (n=90,776), and 27% were NV (n=46,418) for seasonal influenza. Post-pandemic onset, 7% (n=3,556) of the Pre-Pandemic AV and 32% (n=16,710) of the Pre-Pandemic IV never received another influenza vaccine. In the Pandemic-Plus phase, the IV and NV had a higher likelihood of disengagement with the healthcare system [odds ratio (OR) (95% CI): 1.59 (1.55, 1.64); 4.21 (4.09, 4.33)], residing in areas of greater deprivation [1.58 (1.53, 1.62); 1.99 (1.94, 2.05)], and self-identifying their race as Black [2.29 (2.09, 2.51); 3.29 (3.01, 3.60)] compared to the AV. After adjusting for healthcare engagement and area of deprivation, dissatisfaction OR increased from 1.03 (0.99,1.07) Pre-Pandemic to 1.07 (1.03,1.12) in the Pandemic-Plus phase among IV compared to AV. Similarly, the OR of dissatisfaction rose from 1.11 (1.05, 1.18) Pre-Pandemic to 1.16 (1.09, 1.24) in the Pandemic-Plus phase among those who were NV compared to those AV. Conclusion Higher healthcare dissatisfaction was observed among those inconsistently adherent to vaccination recommendations. Targeted efforts to enhance patient satisfaction may boost vaccination rates. Disclosures All Authors: No reported disclosures

Abstract Background Minocycline is a second-generation tetracycline with a broad antimicrobial spectrum. It is often used in chronic suppression of musculoskeletal infections, given activity against gram positive organisms. Efficacy of Minocycline against gram negative skin and soft tissue infection has not previously been defined. In this study, we review in vitro minocycline efficacy against isolates from skin, muscle, and bone sources.Fig 1.Organisms isolated Methods We conducted a retrospective review of all minocycline susceptibility testing performed on musculoskeletal GNB isolates performed at the Mayo Clinic reference lab between 2013-2022. We describe the isolates broken down by site of culture and organism isolated. Then, we calculated the overall susceptibility of these isolates to Minocycline and susceptibility of the most common 6 organisms isolated.Fig 2.Sites of gram negative isolation Results During the study timeframe, we identified 13,567 gram negative isolates from musculoskeletal sites. Of 330 species identified, 19 were identified in at least 100 isolates (fig 1). Of 36 unique sites, 19 sites were each identified in at least 100 isolates (fig 2). 1920 isolates were tested for susceptibility to minocycline. Of these, 1584 (82.5%) were susceptible to minocycline. Of the most common organisms identified and tested for minocycline susceptibility, 99% of Stenotrophomonas maltophilia, 83% of Enterobacter cloacae, 80% of Escherichia coli, 68% of Acinetobacter baumanii, 1% of Proteus mirabilis, 53% of Klebsiella pneumoniae, 65% of Serratia marcescens isolates were susceptible to minocycline (fig 3). Pseudomonas aeruginosa isolates were not tested, due to innate resistance.Fig 3.Minocycline susceptibility compared to key organisms Conclusion A majority of skin and soft tissue cultures positive for Gram negative isolates were below the waist line. Of these, we reported organisms most commonly found. Minocycline showed in vitro activity against some commonly isolated gram negative organisms, especially E.coli, Acinetobacter, and Stenotrophomonas. Further research is needed into the efficacy of minocycline in gram negative skin and soft tissue infections, but it may represent a viable oral option, especially in the end of life. Disclosures All Authors: No reported disclosures

Abstract Background The integration of Generative AI (GAI) into healthcare systems is increasingly recognized for its potential to transform patient management. The primary aim of this research was to evaluate and quantify the performance of large language models (LLMs) in generating actionable travel medicine advice. Architectural design of the Travel Clinic LLM project. Four phases of Discovery, Design, Evaluation, and Implementation/Deployment. Methods This study utilized two iterative phases of evaluation. In the initial phase, LLMs were prompted with detailed clinical scenarios including demographic data, medical and immunization histories, and specific travel plans. These prompts were designed to mimic typical inquiries encountered in travel consultations. The LLMs' initial responses were generated using the CDC’s Yellow Book as a foundational knowledge base. In the subsequent phase, the prompts were refined for greater specificity and clarity, and the knowledge base was enhanced by transitioning to Travax’s Travelers’ Health database. Additional structured data inputs included an exhaustive list of vaccines from our pharmacy formulary and a detailed table of vaccine contraindications. The responses were evaluated and scored by ID clinicians from the Mayo Clinic. Results Initial findings after first iteration revealed limited efficacy with recall at 23.9%, an F1 score of 38.6%, accuracy also at 23.9%, and precision maintained at 100%, utilizing the CDC's Yellow Book. With the implementation of Travax and refined prompting techniques, preliminary results suggest a notable improvement in the quality of responses, though detailed scoring is presently underway. Improvements in the LLM’s performance can be attributed to several key adjustments: the adoption of a more comprehensive knowledge base, refined prompt engineering, and the incorporation of structured data to support more accurate and detailed recommendations. The collaborative engagement of Mayo Clinic with Google and Travax facilitated a synergistic approach to optimizing the AI model's utility and integration. Future plans include embedding the LLM into our EMR system. Conclusion The findings from this study highlight the significance of strategic collaborations between large healthcare centers, IT industry, and specialized knowledge database firms in effectively harnessing GAI for clinical use. Disclosures All Authors: No reported disclosures

Abstract Objective This study aimed to conduct a scoping review of machine learning (ML) techniques in outpatient parenteral antimicrobial therapy (OPAT) for predicting adverse outcomes and to evaluate their validation, implementation and potential barriers to adoption. Materials and methods This scoping review included studies applying ML algorithms to adult OPAT patients, covering techniques from logistic regression to neural networks. Outcomes considered were medication intolerance, toxicity, catheter complications, hospital readmission and patient deterioration. A comprehensive search was conducted across databases including Cochrane Central, Cochrane Reviews, Embase, Ovid MEDLINE and Scopus, from 1 January 2000 to 1 January 2024. Results Thirty-two studies met the inclusion criteria, with the majority being single-centre experiences primarily from North America. Most studies focused on developing new ML models to predict outcomes such as hospital readmissions and medication-related complications. However, there was very little reporting on the performance characteristics of these models, such as specificity, sensitivity and C-statistics. There was a lack of multi-centre or cross-centre validation, limiting generalizability. Few studies advanced beyond traditional logistic regression models, and integration into clinical practice remains limited. Discussion ML shows promise for enhancing OPAT outcomes by predicting adverse events and enabling pre-emptive interventions. Despite this potential, significant gaps exist in development, validation and practical implementation. Barriers include the need for representative data sets and broadly applicable, validated models. Conclusion Future research should address these barriers to fully leverage ML’s potential in optimizing OPAT care and patient safety. Models must deliver timely, accurate and actionable insights to improve adverse event prediction and prevention in OPAT settings.

Background Antibiotics are frequently prescribed at the end of life, though the benefits and harms are not well understood. Methods We abstracted relevant findings from articles published in English in the past 25 years to answer questions generated by discussion among the authors and with stakeholders in Palliative Care and Infectious Diseases. Findings Prescribing practices vary based on individual situation and geographic location. Patients with cancer and those hospitalized receive more antibiotics than those enrolled in outpatient hospice. Urinary tract infections and pulmonary infections are the most common conditions treated with antibiotics at the end of life -most often with penicillin derivatives and vancomycin in the hospital, fluoroquinolones in outpatient, and cephalosporins in both settings. When asked, patients most often prefer limiting antibiotics to symptom management at the end of life. Physicians’ over-estimation of patient preference for antibiotics and the increased probability of misdiagnosis increases antibiotic prescription rates. Antibiotics can improve symptoms when used for specific diseases at the cost of drug reactions, resistant organisms, and delayed discharge. Antibiotic use has variable results on survival duration. Antimicrobial stewardship exists in hospital and long-term care facilities, but not outpatient hospice groups. Stewardship interventions could increase proper use of antibiotics, but more information is needed to apply these interventions to hospice groups. Conclusions Antibiotics at the end of life are impactful and efforts to educate patients and providers will be invaluable in optimizing care.