About

Niaz Banaei is a member of the Advisory Board at the Center for African Studies at Stanford University. The page does not provide specific details about his research focus, background, or key contributions.

Research topics

• Medicine
• Biology
• Internal medicine
• Immunology
• Genetics
• Pathology
• Nanotechnology
• Computer Science
• Materials science
• Microbiology
• Optics
• Bioinformatics
• Dermatology
• Chemistry
• Gastroenterology
• Optoelectronics
• Virology
• Physics
• Statistics
• Computational biology
• Physical therapy
• Emergency medicine

Selected publications

• Transposable elements are driving rapid adaptation of Enterococcus faecium

  Nature · 2026-04-22 · 3 citations

  articleOpen access

  Bacterial pathogens adapt rapidly to clinical and within-host selective pressures1. Insertion sequences (IS) are transposable elements that can contribute to pathogenic adaptation2, but their activity and consequences in contemporary clinical populations are not well characterized. Here, combining large-scale genomic surveys with long-read sequencing of clinical isolates and longitudinal gut metagenomes, we quantify pathogen IS dynamics from global patterns to within-host evolution. Across 19,485 publicly available high-contiguity ESKAPEE pathogen genomes, Enterococcus faecium genomes are the most IS dense, dominated by replicative ISL3 family elements, which have proliferated in clinical lineages over the past 30 years. We find extensive chromosomal structural variation, largely involving ISL3, within a new single-hospital collection of bloodstream isolates. Long-read metagenomic sequencing of 28 longitudinal stool samples from 12 haematopoietic cell transplantation (HCT) recipients demonstrates within-host IS dynamics and their regulatory consequences. In one patient, an ISL3 insertion upstream of a folate transporter formed a strong promoter, increasing transcription and improving relative fitness under folate limitation. Enhanced folate scavenging may enable E. faecium to thrive in the setting of microbiome collapse, which is common in HCT and other critically ill patients3. Together, these results show that a recent ISL3 expansion is driving rapid evolution in healthcare-associated E. faecium, with consequences for its metabolic fitness that may help explain its increasing clinical burden. Several other pathogens also show elevated IS loads in our survey, which suggests that IS expansion-mediated evolution might be more broadly relevant. Over three decades, rapid expansion of the transposable element ISL3 has reshaped Enterococcus faecium, which helps to explain this pathogen’s growing clinical threat.

  PublisherDOI

• P-780. Machine Learning Models for Early Urinary Tract Infection (UTI) Prediction from Electronic Health Records

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background Antibiotic resistance is a growing global threat, and urinary tract infections (UTIs) are a leading driver of inappropriate antibiotic use. Diagnosing true UTIs at the time of culture order is challenging due to variable symptoms and delayed test results, often leading to over-treatment, resistance, drug-related complications, and increased costs. Missed diagnoses risk progression to severe infection. Predictive models using routinely collected electronic health record (EHR) data offer a promising, real-time solution to support stewardship and early decision-making.Table 1:EHR-integrated models feature set description.Figure 1:EHR-integrated models’ performance in terms of AUC-ROC and precision-recall curve. Methods We developed machine learning models to predict clinical UTIs at the time of urine culture order using structured EHR data. Our dataset included 300,381 urine cultures from 164,327 adult patients at two academic and community hospitals (2015- 2024). A previously validated electronic phenotype (Ma et al., 2024) served as the proxy label, combining microbiologic and treatment criteria. Features included demographics, vital signs, and labs within 24 hours prior to and 2 hours after culture order. Two enhanced models incorporated recent diagnoses and antibiotic use (14 days prior). Models were trained with XGBoost. Performance was assessed on a held-out test set (2023- 2024) using ROC-AUC, NPV, and PPV at sensitivity ≥ 80%.Table 2:EHR-integrated models’ AUC-ROC, specificity, negative predictive value (NPV), and positive predictive value (PPV) at threshold where sensitivity is greater than or equal to 80%. Results The baseline model, using only vital signs and labs, achieved an ROC-AUC of 0.73. At 80% sensitivity, it demonstrated high NPV (95%) but limited specificity (51%) and PPV (18%), making it more useful for ruling out UTIs than confirming them. Adding recent ICD-coded diagnoses (Model I) improved ROC-AUC to 0.81, with better specificity (67%) and PPV (24%), and excellent NPV (99%). Model II, which also included prior antibiotic prescriptions, achieved the highest performance: ROC-AUC 0.89, specificity 80%, NPV 97%, and PPV 35%. Conclusion Structured EHR data available at the time of urine culture order can be leveraged to accurately predict clinical UTIs. Incorporating recent diagnoses and antibiotic use significantly enhances performance, enabling scalable, real-time decision support to improve diagnostic precision, guide empiric therapy, and advance antimicrobial stewardship. Disclosures Jonathan H. Chen, MD, PhD, Reaction Explorer: Ownership Interest

  PublisherDOI

• P-417. Machine Learning Prediction of Pediatric Bacteremia: Development of EHR-Based Models for Diagnostic and Clinical Decision Support

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background Pediatric blood cultures are frequently ordered but have low positivity rates (< 4%) in emergency departments (EDs), highlighting the need for better-targeted testing. Accurate prediction can reduce unnecessary cultures, conserve resources, and support stewardship—particularly during the global blood culture bottle shortage. Models developed for adults perform poorly in children due to physiological and clinical differences; in prior work, applying an adult model to pediatric data yielded an AUC of 0.61. We excluded infants < 90 days, who have distinct risk factors (e.g., perinatal history), and developed machine learning models to predict bacteremia in children aged > 90 days to ≤ 18 years using electronic health record (EHR) data.Table 1:PedsBactoScore Point-Based Scoring System Derived from Logistic Regression CoefficientsEach feature contributes a fixed number of points based on clinically meaningful thresholds. The total score is used to stratify risk of bacteremia at the point-of-care.Table 2:Performance Metrics of Pediatric Bacteremia Prediction ModelsComparison of PedsBactoRisk and PedsBactoScore models on the pediatric test set. Metrics include AUC with 95% confidence intervals, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) at pre-specified thresholds. Methods We analyzed 26,829 blood culture orders from 9,362 pediatric emergency department (ED) encounters at Stanford Medicine | Children's Health. To preserve temporal validity, data were split chronologically, with the most recent encounters used as the test set. We developed two models: PedsBactoRisk, a logistic regression model, and PedsBactoScore, a simplified point-based tool derived from the most influential PedsBactoRisk predictors. The PedsBactoScore rubric is shown in Table 1. Results Table 2 summarizes model performance. We evaluated sensitivity, specificity, PPV, and NPV, focusing on thresholds achieving 90% sensitivity. PedsBactoRisk achieved an AUC-ROC of 0.75; PedsBactoScore, 0.64. While PedsBactoRisk showed superior performance, PedsBactoScore allows easier implementation via its interpretable scoring system. PedsBactoScore performance is shown across thresholds to illustrate sensitivity–specificity trade-offs. Conclusion PedsBactoRisk demonstrated the highest overall performance (AUC: 0.75), but PedsBactoScore offers a pragmatic, interpretable bedside tool with strong sensitivity. Both models support more judicious blood culture use by identifying low-risk patients with high sensitivity. Future work will focus on integrating provider notes using large language models to enhance predictive accuracy and extending this approach to infants < 90 days by incorporating maternal and delivery data. Disclosures Jonathan H. Chen, MD, PhD, Reaction Explorer: Ownership Interest

  PublisherDOI

• Impact of stopping contact precautions on <i>vanA</i> plasmid transmission, Northern California, 2021–2023

  Antimicrobial Stewardship & Healthcare Epidemiology · 2026-01-01

  articleOpen access

  Abstract Objective: To evaluate the impact of discontinuing contact precautions for vancomycin-resistant enterococci (VRE) on strain and plasmid transmission using long-read whole-genome sequencing (WGS). Study design: Before-after trial of adults with Enterococcus bloodstream infections pre-(Jan–Oct 2021) and post-(Oct–Dec 2021 and Jan–Oct 2023) discontinuation of contact precautions for VRE infections. Setting: Quaternary referral and transplant academic medical center. Patients: Hospitalized adults (≥18 yr) with E. faecalis or E. faecium bacteremia. Methods: Classical epidemiology identified potential transmissions via shared unit exposure within a 14-day window. Blood culture isolates underwent long-read WGS to assess strain and vanA plasmid relatedness. Clonal transmission was defined as &lt;20 single-nucleotide polymorphisms. Plasmid similarity was assessed with Mash distance. Findings: Among 288 isolates from 202 patients, there was no significant difference in possible epidemiologic transmissions pre-versus postdiscontinuation (9.5% vs 8.1%; P = .679). Genomic analysis identified four clonal transmission events, two of which occurred postdiscontinuation. Among 70 vanA plasmids from 54 patients, 38 highly related plasmids formed a low-diversity cluster. The proportion of cluster plasmids was not significantly different between periods (47% vs 60%; P = .267). Postdiscontinuation, vanA -positive E. faecium ST117 was more prevalent (22/44 vs 53/75; P = .024). Conclusion: Discontinuation of contact precautions for VRE was not associated with increased transmission of enterococci or vanA plasmids in bloodstream infections. Transmission patterns remained largely stable, though the postdiscontinuation period showed increased prevalence of the dominant E. faecium ST117. These findings suggest limited impact of contact precautions on VRE transmission.

  PublisherOA PDFDOI

• Accuracy of Plasma Cell-Free DNA PCR for Diagnosis of Coccidioidomycosis

  Clinical Infectious Diseases · 2026-05-03

  articleSenior author

  BACKGROUND: Coccidioidomycosis is the most prevalent endemic mycosis in the United States. Diagnosis of coccidioidomycosis can be challenging due to limitations of conventional diagnostics. Detection of fungal plasma cell-free DNA (cfDNA) is a novel non-invasive testing modality for the diagnosis of invasive fungal disease. In this study, we characterized the performance of Coccidioides plasma cfDNA PCR. METHODS: A retrospective study was conducted in patients with suspected coccidioidomycosis who underwent plasma cfDNA PCR testing and conventional fungal testing for C. immitis and C. posadasii. Patients were categorized per the EORTC/MSGERC case definitions as proven or probable for sensitivity analyses and no coccidioidomycosis for specificity analyses. RESULTS: Overall, 362 plasma samples from unique patients were included in this study. The overall sensitivity, specificity, and positive and negative predictive values of Coccidioides plasma cfDNA PCR at 0.04% prevalence were 57.5% (27/47; 95% CI, 42.2-71.7), 100% (315/315; 95% CI, 98.8-100), 100% (95% CI, 87.2-100), and 99.9% (95% CI, 100-100), respectively. The sensitivity was 33.3% (2/6; 95% CI, 4.3-77.7) in patients with pulmonary focal disease, 66.7% (14/21; 95% CI, 43.0-85.4; P≥0.05) in patients with pulmonary multifocal disease, and 52.6% (10/19; 95% CI, 39.6-77.9; P≥0.05) in patients with disseminated coccidioidomycosis. In patients with concurrent antigen results, the sensitivity of cfDNA PCR was 58.3% (14/24; 95% CI, 39.6-77.9) compared with 37.5% (9/24; 95% CI, 18.8-59.4; P≥0.05) for antigen. CONCLUSIONS: Coccidioides plasma cfDNA PCR is modestly sensitive but highly specific which can be useful for rapid and non-invasive diagnosis of coccidioidomycosis.

  PublisherDOI

• 431. Performance of an Expert Recommendation Framework for Blood Culture Stewardship: Comparing Clinician Manual Review and Large Language Model Automation

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background The 2024 blood culture bottle shortage created an urgent need to conserve supplies and prioritize high-yield testing. Institutions turned to expert frameworks like Fabre et al. (2020), which stratify bacteremia risk by clinical presentation, though these frameworks have not been evaluated at scale. In our pilot, unguided LLM queries produced high sensitivity but poor specificity, consistent with prior literature, suggesting a tendency to overestimate infection risk. To address this, we anchored both clinician and LLM classification in the Fabre framework to improve precision and enable scalable clinical decision support.Figure 1:Large Language Model (LLM)-Based Pipeline for Automated Risk Stratification of Bacteremia Schematic diagram illustrating the structured pipeline leveraging a HIPAA-compliant GPT-4 model to automate bacteremia risk assessment. The pipeline integrates the expert recommendation framework (Fabre et al., 2020) within a structured, prompt-based evaluation. Inputs to the LLM include the expert recommendation criteria, structured electronic health record (EHR) patient data, and emergency department (ED) provider notes recorded at the time of blood culture ordering. The figure also details the complete structure of prompts used to generate risk stratification outputs.Table 1:Case Stratification by Clinician Manual vs. Large Language Model (LLM) Automated Classification* BCx = blood cultures. This table shows the distribution of reviewed emergency department (ED) patient cases stratified by bacteremia risk categories using clinician manual classification (n = 108) and automated LLM-based classification (n = 1,013). Risk categories reflect pre-test probability for bacteremia as defined by the expert recommendation framework, for example: - Low Risk: Isolated fever and/or leukocytosis, uncomplicated lower urinary tract infection, non-severe cellulitis. - Intermediate Risk: Acute pyelonephritis, cholangitis, severe community-acquired pneumonia (CAP). - High Risk: Catheter-associated bloodstream infection, meningitis, diskitis/native vertebral osteomyelitis. Methods We conducted a two-phase evaluation. First, four physicians independently reviewed 108 emergency department (ED) cases with blood culture orders (80% positive), stratifying bacteremia risk using the Fabre framework; discrepancies were resolved by an expert reviewer. Second, we automated stratification using a HIPAA-compliant GPT-4 LLM with structured prompts incorporating the framework, electronic health record (EHR) data, and ED provider notes. For both approaches, low-risk classifications were treated as predicted negatives; intermediate/high-risk as predicted positives. LLM evaluation included 1,013 patient-encounter-orders, with intentional oversampling of positives (prevalence 19%).Table 2:Comparison of Performance Metrics for Clinician Manual vs. Large Language Model (LLM) Automated Classification Comparative performance of clinician manual classification and automated LLM-based classification for predicting bacteremia risk in ED patients. Metrics reported include accuracy (i.e., proportion of cases correctly classified), sensitivity (i.e., ability to correctly identify patients with true positive blood cultures), and specificity (i.e., ability to correctly identify patients with negative blood cultures). Results Manual review achieved 86% sensitivity and 57% specificity, missing 14 positive cultures and recommending cultures in 3 negatives. LLM classification yielded 96% sensitivity and 16% specificity, correctly identifying 393 of 409 true positives but over-classifying 315 of 591 negatives. Conclusion The Fabre framework offers structured guidance, but manual application showed limited performance and is impractical at scale. LLM-based automation enabled large-scale stratification but lacked specificity, producing excess false positives. A hybrid model, using the LLM to exclude low-risk cases and refer higher-risk cases for clinician review, may improve accuracy and resource use, highlighting the role of generative AI in scaling framework validation. Disclosures All Authors: No reported disclosures

  PublisherDOI

• Cultryx: Precision Diagnostic Stewardship for Blood Cultures Using Machine Learning

  medRxiv · 2026-03-04

  articleOpen access

  Abstract Background The 2024 blood culture bottle shortage brought diagnostic resource allocation to the forefront, reflecting persistent, foundational challenges with low-value testing and empiric treatment approaches under clinical uncertainty. Objective To determine whether a machine learning approach using electronic medical record data can predict bacteremia more effectively than existing systems and practices to guide diagnostic testing and empiric treatment strategies. Methods In a retrospective cohort of 101,812 adult emergency department encounters (2015-2025), we first established an idealized cognitive baseline by evaluating physician and generative AI (GPT-5) application of the professional society-endorsed Fabre framework on a validation subset. We then trained an XGBoost model (Cultryx) on the full cohort to predict bacteremia, benchmarking its performance against real-world clinical heuristics (SIRS, Shapiro Rule). Results For the idealized baseline, physicians applying the Fabre framework achieved 95.7% sensitivity, but GPT-5 automation failed to replicate this standard (71.6% sensitivity). In real-world benchmarking, Cultryx outperformed all clinical heuristics (AUROC 0.810). SIRS lacked specificity (41.2%), driving diagnostic overuse, while the Shapiro Rule lacked sensitivity (70.2%), missing ~30% of bacteremia cases. In contrast, when calibrated to a strict 95% sensitivity target, Cultryx achieved the highest culture volume deferral rate (26.2%, deferring ~ 15,872 bottles with predicted negative results) while maintaining a 98.9% negative predictive value. Cultryx score , a simplified bedside tool, retained a 20.8% deferral rate. Conclusions Machine learning provides a superior, data-driven alternative to mainstream clinical heuristics for predicting bacteremia. By maximizing culture deferment without compromising pathogen detection, Cultryx can conserve diagnostic resources, reduce unnecessary empiric antibiotic exposure, and systematically elevate patient safety. Summary Cultryx, a machine learning model for blood culture stewardship, outperforms standard clinical heuristics in predicting bacteremia. This approach could reduce culture utilization by over 26% while preserving pathogen detection, conserving diagnostic resources, reducing unnecessary antibiotic exposure, and elevating patient safety.

  PublisherOA PDFDOI

• P-416. From Broad to Best: A Structured, Automated, and Scalable EHR Approach to Evaluate Empiric Antibiotic Appropriateness

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background Current antimicrobial stewardship metrics emphasize reducing overall antibiotic use but rarely assess patient-level appropriateness. Tools like the SAAR and EHR alerts benchmark use or trigger rules but do not evaluate whether an agent was clinically appropriate. We developed an automated, scalable metric using SQL, widely adopted and optimized for querying data, to implement the DOOR MAT (Desirability of Outcome Ranking for the Management of Antimicrobial Therapy) framework, which ranks empiric antibiotics by spectrum, favoring narrower agents when susceptible.Figure 1:Cohort Construction Flowchart for Adult Emergency Department (ED) Urinary Tract Infection CasesThis flow diagram illustrates the cohort generation process for adult ED patients with presumed urinary tract infection (UTI), based on urine culture orders and empiric antibiotic prescriptions. Starting from all urine culture orders, sequential filters were applied to isolate cases from the ED with associated empiric antibiotic treatment, while excluding those with recent antibiotic exposure (within 30 days) or non-relevant encounters. This figure provides a visual example of inclusion and exclusion criteria, culminating in the final analytic sample used in the appropriateness analysis. The same logic was applied to generate the pediatric ED and adult outpatient cohorts (not shown). As a proof-of-concept study using a large real-world dataset, limitations include potential misclassification, missing data, and variability in documentation or data completeness across care settings.Figure 2:Antibiotic Spectrum Tiering Used for Appropriateness Classification in the DOOR MAT FrameworkThis figure presents selected examples of how empiric antibiotics were categorized into six hierarchical spectrum tiers, adapted from the World Health Organization’s AWaRE classification. AWaRE groups antibiotics into three categories to guide stewardship prioritization: Access (first-line agents with low resistance potential), Watch (higher resistance potential), and Reserve (last-resort agents for multidrug-resistant infections). These categories were expanded to improve granularity for spectrum-based assessment. Narrow-spectrum agents appear in lower tiers, while broader-spectrum and last-resort agents occupy higher tiers. This new tiering system, developed and validated by an infectious diseases physician, served as the basis for evaluating whether a prescribed agent was optimal, broader than necessary (over-treatment), or lacked adequate activity (under-treatment) relative to culture and antimicrobial susceptibility testing (AST) data. The tiers were used in conjunction with SQL-based cohort construction and join logic to apply the DOOR MAT (Desirability of Outcome Ranking for the Management of Antimicrobial Therapy) framework. This is not a comprehensive list; antibiotics shown here are representative examples from the full tiering system. Methods We used the ARMD EHR dataset to identify presumptive UTI cases based on urine culture orders and empiric antibiotics, excluding patients with antibiotic exposure in the prior 30 days. Cohorts were stratified by care setting (Figure 1). Antibiotics were categorized into six spectrum tiers adapted from WHO AWaRE and validated by an infectious diseases physician (Figure 2). Using the DOOR MAT framework, we applied SQL logic to compare empiric therapy with culture and antimicrobial susceptibility testing (AST) results, classifying each case as optimal, over-treatment, under-treatment, unnecessary, or not assessable. All unique antibiotics and organisms were retained for full assessment.Figure 3:Appropriateness of Empiric Antibiotic Prescribing for Urinary Tract Infections Across Care SettingsThis figure displays a spectrum-based histogram of empiric antibiotic appropriateness for culture-positive urinary tract infection (UTI) cases across adult ED, pediatric ED, and adult outpatient settings. Prescriptions are classified as optimal (green), indicating the narrowest agent with full in vitro activity based on antimicrobial susceptibility testing (AST) or implied susceptibility; over-treatment (yellow to red gradient), where the agent was active but broader than necessary, with color intensity reflecting the number of spectrum tiers above the optimal choice; under-treatment (dark red), where the agent lacked activity against the cultured organism(s); and not assessable (gray), where AST was unavailable and no intrinsic resistance or predictable susceptibility could be inferred. The histogram includes only culture-positive cases to maintain interpretability. The percentage and count of unnecessary prescriptions (defined as empiric antibiotics given for negative cultures) are shown separately to the right, as inclusion in the main histogram would distort the visual scale. This color-coded format enables intuitive assessment of antibiotic use across departments, where green reflects appropriate prescribing and red indicates increasingly inappropriate use.Figure 4:Spectrum Deviation of Commonly Over-Treated Antibiotics in the Adult Outpatient CohortThis bar chart displays the five most frequently prescribed empiric antibiotics in the adult outpatient cohort that were classified as over-treatment based on final urine culture and antimicrobial susceptibility testing (AST) results. Each antibiotic is labeled on the x-axis with the number of spectrum tier deviations above the optimal agent, based on the adapted WHO AWaRE classification system. For example, nitrofurantoin was the most commonly prescribed agent but typically deviated by only one tier from the optimal choice, whereas ciprofloxacin was two tiers broader than necessary in most cases, where an agent such as amoxicillin (lowest tier) would have provided adequate coverage. This figure illustrates how over-treatment varies not only by drug selection but also by degree of unnecessary spectrum, providing insight into stewardship opportunities beyond binary classification alone. Results Of 73,881 adult ED prescriptions, 58.0% were unnecessary. Among culture-positive cases, 3.3% were optimal, 53.3% over-treated, 17.4% under-treated, and 26.0% lacked AST. In 7,213 pediatric ED cases, 64.4% were unnecessary; among positives, 7.4% were optimal, 29.5% over-treated, and 47.2% lacked AST. Among 47,109 adult outpatient prescriptions, 68.1% were unnecessary; among positives, 10.1% were optimal, 56.3% over-treated, and 23.8% lacked AST (Figure 3). Nitrofurantoin and ciprofloxacin were the most overused agents in the outpatient setting (Figure 4). Conclusion This structured, SQL-based framework enables standardized assessment of empiric antibiotic appropriateness using only EHR data. By determining appropriateness along a spectrum relative to culture and susceptibility results, it offers a scalable alternative to manual audit or rule-based alerts. These reproducible measures can support real-time and longitudinal stewardship, particularly in high-volume or outpatient settings. Disclosures Hayden T. Schwenk, MD, MPH, Bristol Myers Squibb: Stocks/Bonds (Public Company)|Karius, Inc.: Consultant, Medical Affairs Jonathan H. Chen, MD, PhD, Reaction Explorer: Ownership Interest

  PublisherDOI

• P-11. Evolution of Gram-Negative Rod Bacteremia Management in Hematology and Oncology Patients

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background Managing uncomplicated Gram-negative rod bacteremia (uGNB) with shorter durations of therapy and oral antibiotic transition has been well-established by randomized trials and large retrospective studies. However, immunocompromised patients, like hematology and oncology patients, are often underrepresented in such trials. Our goal was to describe uGNB management over a contemporary 5-year period in this population. Methods This single-center, retrospective, study included adults with an active hematologic or solid malignancies and monomicrobial Enterobacterales bloodstream infection managed at Stanford Hospital between 2019 and 2024. Duration was categorized as short (≤10 days) or long ( &amp;gt;10 days). Exclusion criteria included bacteremia secondary to endocarditis, meningitis, or osteomyelitis, uncontrolled source of infection, failure to receive at least one active antibiotic, resistance to carbapenems, allogenic bone marrow transplant, and transition to hospice care or death within 14 days of culture. Results 172 patients were included. Primary sources of infection included urinary (52.3%) and intra-abdominal (26.7%) with the most common organism being Escherichia coli (56.9%). Most had a solid organ malignancy (87.2%); 26.7% of these had metastatic disease and 26.2% had localized disease with chemotherapy within 6 months. The prevalence of short antibiotic duration increased from 42% to 63% (Figure 1) and oral antibiotic transition increased from 65% to 78% (Figure 2). When classified by the most common indications, increase in short duration was mainly driven by urinary tract infections and oral antibiotic transition was mainly driven by intra-abdominal infections. Conclusion Despite major comorbidity, contemporary uGNB antibiotic treatment paradigms, namely, shorter duration and oral transition, were increasingly employed in hematology oncology patients. Disclosures All Authors: No reported disclosures

  PublisherDOI

• P-1974. Predicting Bacteremia in Emergency Departments: A Suite of Data-Driven Clinical Decision Tools

  Open Forum Infectious Diseases · 2026-01-01

  articleOpen access

  Abstract Background The global blood culture bottle shortage in 2024 highlighted the critical need to optimize test utilization. While blood cultures are essential for diagnosing bacteremia, fewer than 10% yield true-positive findings. Overuse increases the risk of contamination unnecessary antibiotic exposure and hospitalizations, and strains resources. To address this, we developed a suite of predictive models leveraging structured and unstructured electronic health record (EHR) data to better stratify bacteremia risk and support targeted blood culture ordering.Table 1:BactoScore Scoring System This table summarizes conditions assigned to each feature in the BactoScore system, derived from the coefficients of the BactoRisk model. This design ensures straightforward application in clinical settings. A cumulative score of 4 or higher indicates a high likelihood of a positive blood culture, achieving a sensitivity of 0.95.Table 2:Comparison of the performance of BactoPro, BactoPlus, BactoRisk, and BactoScore against common SIRS-based criteria (Systemic Inflammatory Response Syndrome) and the numerical components of the Shapiro method While the full Shapiro method involves a broad range of features, this comparison focuses solely on the numerical components of Shapiro. BactoPlus incorporates additional features such as prior antibiotic usage and diagnoses, which enhance the model’s predictive power. Our results demonstrate that both BactoRisk and BactoScore outperform the common SIRS-based criteria and the Shapiro method, highlighting their potential for practical use in EDs. Furthermore, the BactoScore model allows flexibility in setting the most appropriate sensitivity threshold for culture ordering based on the clinical scenario. At a sensitivity of 91%, clinicians should order the test when 5 or more criteria are met. Methods We analyzed 135,483 ED blood cultures from patients (≥18 years, no recent bacteremia) at Stanford and Tri-Valley Hospitals (2015–2023). The primary outcome was true bacteremia (excluding contaminants). We developed three models: (1) BactoScore, a simplified, interpretable point-based tool derived from BactoRisk, a logistic regression model using structured clinical and laboratory data; (2) BactoPlus, an XGBoost-based model extending BactoRisk with diagnosis codes and recent antibiotic exposure; and (3) BactoPro, a multimodal model incorporating clinical notes via a large language model (LLM) alongside structured EHR data. Models were compared to SIRS and Shapiro’s score.Table 3:Comparison of SIRS Criteria, Shapiro, BactoRisk, BactoPlus, BactoPro, and BactoScore Models for Blood Culture OptimizationThis table summarizes the key features, pros, and cons of each model, highlighting their applicability, real-time implementation potential, and performance in predicting bacteremia risk and reducing unnecessary blood culture orders in ED settings. Results At 90% sensitivity, all three models outperformed SIRS and Shapiro. BactoPro achieved the highest accuracy by combining clinical notes with structured features. BactoPlus offered comparable performance using enriched structured features. BactoScore, while simpler, retained strong predictive power and interpretability, enabling bedside use with flexible thresholds. At a 4+ point cutoff, BactoScore achieved 95% sensitivity and supported safe culture reduction with minimal missed bacteremia. Conclusion Our tiered suite of predictive models, from the interpretable BactoScore to the advanced, multimodal BactoPro, offers flexible options for improving blood culture stewardship by accurately predicting bacteremia than existing standards. These tools support real-time risk stratification, targeted ordering, and better resource utilization in ED settings. By aligning model complexity with available infrastructure, health systems can select scalable solutions that optimize care while preserving diagnostic safety. Disclosures Jonathan H. Chen, MD, PhD, Reaction Explorer: Ownership Interest

  PublisherDOI

Frequent coauthors

• Indre Budvytiene

  Stanford University

  104 shared

• Benjamin A. Pinsky

  Communities In Schools of Orange County

  87 shared

• Catherine A. Hogan

  45 shared

• Kanagavel Murugesan

  Stanford University

  38 shared

• Fiona Senchyna

  Stanford University

  30 shared

• Rajiv L. Gaur

  29 shared

• Ángel Moreno

  Stanford University

  28 shared

• Cynthia Truong

  Howard Hughes Medical Institute

  26 shared

Labs

• Center for African Studies Advisory Board MembersPI

Education

• M.D.

  Stanford University

• Other, Laboratory Medicine

  University of California, San Francisco

• Other, Infectious Diseases

  New York University

Awards & honors

• Kenneth L. Vosti Infectious Diseases and Stanford University…