Nicholas P. Marshall

All Publications

• Cultryx: Precision Diagnostic Stewardship for Blood Cultures Using Machine Learning. medRxiv : the preprint server for health sciences Marshall, N. P., Chen, W., Amrollahi, F., Haredasht, F. N., Maddali, M. V., Ma, S. P., Zahedivash, A., Black, K. C., Chang, A., Deresinski, S. C., Goldstein, M. K., Asch, S. M., Banaei, N., Chen, J. H. 2026

  Abstract

  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.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.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).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. Cultryxscore, a simplified bedside tool, retained a 20.8% deferral rate.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.

  View details for DOI 10.64898/2026.02.27.26347214

  View details for PubMedID 41867170

  View details for PubMedCentralID PMC13004107

• Context-Aware, Bundled Clinical Decision Support Drives Pediatric Outpatient Antibiotic Guideline Condordance from 59% to 94% IDWeek Marshall, N. P., Puckett, L. M., Cabrera-Haro, L., Krishna, S., Bio, L. L., O'Brien, L. E., Kling, S., Holdsworth, L., Lee, G. M., Schwenk, H. T. 2026
• Perception, Practice, and Capacity: An International Survey of Informatics and AI in Infectious Diseases IDWeek Waxse, B., Pradhan, A., Marshall, N. P., Diaz, G., Yamshchikov, A., Chang, A. 2026
• Mycoplasma pneumoniae Testing, Diagnosis, and Treatment in Hospitalized Children Across the US: A 10-Year Retrospective Analysis Pediatric Hospital Medicine (PHM) Conference Bassett, H. K., Chilukuri, N., Cunningham, C., Allan, J., Marshall, N. P., Pageler, N., Schwenk, H. T., Ritter, V., Schroeder, A. 2026
• Can a General-Purpose Coding Agent Analyze a Production Hospital Data Warehouse? Marshall, N. P., Haberkorn, W., Faulkenberry, J., Palakkattil, G., Nateghi Haredasht, F., Schwenk, H. T., Chen, J. H., Morse, K. medRxiv. 2026

  Abstract

  Background: Health systems answer most questions by having expert analysts hand-write queries against a complex electronic health record data warehouse, a slow, resource-intensive process. Whether an autonomous coding agent can do this accurately is unknown. Methods: In a single-center quality-improvement evaluation, we posed ten questions about a common pediatric infection, acute otitis media. The questions went to analysts, whose adjudicated answers were the reference, and to an autonomous coding agent (OpenAI Codex), which wrote and ran read-only queries on a full copy of the production data warehouse (Epic Caboodle). We ran the agent under four conditions: an autonomous baseline (each question answered three times at two reasoning-effort settings), a variant in which it listed its assumptions, interactive analyst feedback, and reuse of a corrected definition across related questions. Outcomes were accuracy, patient-level agreement (F1), reproducibility, and efficiency (time and tokens). Results: Working autonomously, the agent wrote valid queries and never fabricated data, but rarely produced the exact answer. At medium effort, it came within 5% of the reference on 27 of 30 runs but exact on only 10. Higher effort produced no improvement. Reproducibility was the greater weakness, with the three runs returning an identical answer on only 3 of 10 questions. When a run matched the reference, it had found the same patients (F1 = 1.00); the exception was an over-counted procedure (F1 = 0.72). With analyst feedback on four questions, it answered two exactly, and reusing the corrected definition on related questions restored reproducibility and accuracy. Conclusions: A general-purpose coding agent matched an adjudicated analyst reference on most routine questions but not reproducibly, because key definitions depended on warehouse knowledge that the data dictionary omits. Supplying that knowledge, by prompt or feedback, restored reproducibility. Under expert analyst supervision, the agent is already a capable drafting aid and a promising step toward broader hospital analytic support.

  Preprint
• External Validation of Machine Learning for Antimicrobial Susceptibility Prediction Across Three Health Systems and Culture Types IDWeek Chen, J. S., Nateghi Haredasht, F., Marshall, N. P., Chen, W., Konsker, H. B., Cooper, L. N., Medford, R., Ma, S. P., Chang, A., Deresinski, S. C., Goldstein, M., Asch, S. M., Banaei, N., Chen, J. H. 2026
• Cultryx Mini: An Age-Banded Score for Pediatric Emergency Department Blood Culture Stewardship IDWeek Marshall, N. P., Chen, W., Nateghi Haredasht, F., Konsker, H. B., Archer, E. J., Kuklinski, C., Khan, A., Seddik, T., Ma, S. P., Deresinski, S. C., Goldstein, M., Asch, S. M., Banaei, N., Chang, A., Morse, K., Chen, J. H. 2026
• Off-Label Performance of Three Published Blood Culture Decision Rules in a Pediatric Emergency Department IDWeek Marshall, N. P., Chen, W., Nateghi Haredasht, F., Konsker, H. B., Archer, E. J., Kuklinski, C., Khan, A., Seddik, T., Ma, S. P., Deresinski, S. C., Goldstein, M., Asch, S. M., Banaei, N., Chang, A., Morse, K., Chen, J. H. 2026
• Antibiotic Resistance Microbiology Dataset (ARMD): A Resource for Antimicrobial Resistance from EHRs. Scientific data Nateghi Haredasht, F., Amrollahi, F., Maddali, M. V., Marshall, N., Ma, S. P., Cooper, L. N., Johnson, A. O., Wei, Z., Medford, R. J., Kanjilal, S., Banaei, N., Deresinski, S., Goldstein, M. K., Asch, S. M., Chang, A., Chen, J. H. 2025; 12 (1): 1299

  Abstract

  The Antibiotic Resistance Microbiology Dataset (ARMD) is a de-identified resource derived from electronic health records (EHR) that facilitates research in antimicrobial resistance (AMR). ARMD encompasses big data from adult patients collected from over 15 years at two academic-affiliated hospitals, focusing on microbiological cultures, antibiotic susceptibilities, and associated clinical and demographic features. Key attributes include organism identification, susceptibility patterns for 55 antibiotics, implied susceptibility rules, and de-identified patient information. This dataset supports studies on antimicrobial stewardship, causal inference, and clinical decision-making. ARMD is designed to be reusable and interoperable, promoting collaboration and innovation in combating AMR. This paper describes the dataset's acquisition, structure, and utility while detailing its de-identification process.

  View details for DOI 10.1038/s41597-025-05649-7

  View details for PubMedID 40715119

  View details for PubMedCentralID PMC12297523

• Performance of an Expert Recommendation Framework for Blood Culture Stewardship: Comparing Clinician Manual Review and Large Language Model Automation IDWeek Marshall, N. P., Amrollahi, F., Maddali, M., Black, K. C., Zahedivash, A., Nateghi Haredasht, F., Ma, S. P., Chang, A., Deresinski, S., Banaei, N., Goldstein, M. K., Asch, S. M., Chen, J. H. 2025

  View details for DOI 10.1093/ofid/ofaf695.144

• First, do NOHARM: towards clinically safe large language models Wu, D., et al ArXiv. 2025

  Abstract

  Large language models (LLMs) are routinely used by physicians and patients for medical advice, yet their clinical safety profiles remain poorly characterized. We present NOHARM (Numerous Options Harm Assessment for Risk in Medicine), a benchmark using 100 real primary care-to-specialist consultation cases to measure frequency and severity of harm from LLM-generated medical recommendations. NOHARM covers 10 specialties, with 12,747 expert annotations for 4,249 clinical management options. Across 31 LLMs, potential for severe harm from LLM recommendations occurs in up to 22.2% (95% CI 21.6-22.8%) of cases, with harm of omission accounting for 76.6% (95% CI 76.4-76.8%) of errors. Safety performance is only moderately correlated (r = 0.61-0.64) with existing AI and medical knowledge benchmarks. The best models outperform generalist physicians on safety (mean difference 9.7%, 95% CI 7.0-12.5%), and a diverse multi-agent approach improves safety compared to solo models (mean difference 8.0%, 95% CI 4.0-12.1%). Therefore, despite strong performance on existing evaluations, widely used AI models can produce severely harmful medical advice at nontrivial rates, underscoring clinical safety as a distinct performance dimension necessitating explicit measurement.

• Machine Learning Models for Early Urinary Tract Infection (UTI) Prediction from Electronic Health Records IDWeek Marshall, N. P., Amrollahi, F., Nateghi Haredasht, F., Ma, S. P., Maddali, M. V., Chang, A., Deresinski, S., Banaei, N., Goldstein, M. K., Asch, S., Chen, J. H. 2025

  View details for DOI 10.1093/ofid/ofaf695.991

• From Broad to Best: A Structured, Automated, and Scalable EHR Approach to Evaluate Empiric Antibiotic Appropriateness IDWeek Chen, W., Marshall, N. P., Amrollahi, F., Nateghi Haredasht, F., Maddali, M. V., Ma, S. P., Chang, A., Deresinski, S., Goldstein, M. K., Asch, S., Banaei, N., Schwenk, H. T., Chen, J. H. 2025

  View details for DOI 10.1093/ofid/ofaf695.633

• Predicting Bacteremia in Emergency Departments: A Suite of Data-Driven Clinical Decision Tools IDWeek Marshall, N. P., Amrollahi, F., Nateghi Haredasht, F., Ma, S. P., Maddali, M. V., Chang, A., Deresinski, S., Banaei, N., Goldstein, M. K., Asch, S., Chen, J. H. 2025

  View details for DOI 10.1093/ofid/ofaf695.2141

• Machine Learning Prediction of Pediatric Bacteremia: Development of EHR-Based Models for Diagnostic and Clinical Decision Support IDWeek Marshall, N. P., Amrollahi, F., Nateghi Haredasht, F., Black, K. C., Zahedivash, A., Maddali, M. V., Ma, S. P., Chang, A., Deresinski, S., Goldstein, M. K., Asch, S., Banaei, N., Chen, J. H. 2025

  View details for DOI 10.1093/ofid/ofaf695.634

• Pre-Transplant Initiation of Direct-Acting Antiviral Therapy for Donor-Derived Hepatitis C in a Pediatric Patient International Society for Heart and Lung Transplantation (ISHLT) Kushner, L. E., Marshall, N. P., Schwenk, H. T., Almond, C., Ma, M., Puckett, L., Lee, J., Chen, S. F., Profita, E. ISHLT 2025 Scientific Program. 2025
• Machine Learning-Based Prediction of Antimicrobial Susceptibility: A Step Towards Precision Antimicrobial Stewardship. AMIA ... Annual Symposium proceedings. AMIA Symposium Amrollahi, F., Haredasht, F. N., Vansomphone, A., Marshall, N., Maddali, M. V., Ma, S. P., Chang, A., Deresinski, S. C., Goldstein, M. K., Kanjilal, S., Medford, R. J., Cooper, L. N., Asch, S. M., Banaei, N., Chen, J. H. 2024; 2024: 138-146

  Abstract

  Antimicrobial resistance (AMR) represents an urgent global health crisis exacerbated by the frequent empirical use of broad-spectrum antibiotics. AMR is exacerbated by inherent delays in obtaining culture results and antimicrobial susceptibility data after sample collection. In this study, we developed and validated Machine Learning (ML) models using routinely collected EHR data from inpatient and outpatient encounters to predict antibiotic resistance at the time of blood, urine or respiratory bacterial culture collection. The models demonstrated robust predictive accuracy, particularly in inpatient settings where clinical data was more consistently available. Notably, the model independently identified patterns that predict resistance, similar to how a clinician would attempt to predict resistance using prior culture and susceptibility data combined with their clinical training and knowledge of microbiological resistance patterns. Integrating these predictive tools into clinical workflows could significantly enhance empirical antibiotic selection, reduce unnecessary broad-spectrum antibiotic use, and meaningfully advance antimicrobial stewardship efforts.

  View details for DOI 10.1023/A:1012487302797

  View details for PubMedID 41726396

  View details for PubMedCentralID PMC12919605

• A Multi-Phase Analysis of Blood Culture Stewardship: Machine Learning Prediction, Expert Recommendation Assessment, and LLM Automation. AMIA ... Annual Symposium proceedings. AMIA Symposium Amrollahi, F., Marshall, N., Haredasht, F. N., Black, K. C., Zahedivash, A., Maddali, M. V., Ma, S. P., Chang, A., Deresinski, S. C., Goldstein, M. K., Asch, S. M., Banaei, N., Chen, J. H. 2024; 2024: 147-156

  Abstract

  Blood cultures are often overordered without clear justification, straining healthcare resources and contributing to inappropriate antibiotic use-pressures worsened by the global shortage. In the study of 135,483 emergency department (ED) blood culture orders, we developed machine learning (ML) models to predict the risk of bacteremia using structured electronic health record (EHR) data and provider notes via a large language model (LLM). The structured model's AUC improved from 0.76 to 0.79 with note embeddings and reached 0.81 with added diagnosis codes. Compared to an expert recommendation framework applied by human reviewers and an LLM-based pipeline, our ML approach offers higher specificity without compromising sensitivity. The recommendation framework achieved sensitivity 86%, specificity 57%, while the LLM maintained high sensitivity (96%) but overclassified negatives, reducing specificity (16%). These findings demonstrate that ML models integrating structured and unstructured data can outperform consensus recommendations, enhancing diagnostic stewardship beyond existing standards of care.

  View details for PubMedID 41726397

  View details for PubMedCentralID PMC12919529

• Surgical Prophylaxis in Pediatric Liver Transplant Recipients: Variations in Protocol and Impact on 30-Day Surgical Site Infections IDWeek Marshall, N. P., Foster, C. B., Aragon, V., Hashimoto, K., Gonzalez, B. E. 2022

  View details for DOI 10.1093/ofid/ofac492.1770