Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Jordan Guillot has a total of 32 co-authors. Here are the top co-authors by number of shared publications:

  • Arvind Suresh

    Affiliation Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America

    5 shared publications
  • Atul J. Butte

    Affiliation Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America

    5 shared publications
  • Brenda Miao

    Affiliations Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America, Quality Health, Inc., San Francisco, California, United States of America

    5 shared publications
  • Madhumita Sushil

    Affiliation Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America

    5 shared publications
  • Marina Sirota

    Affiliation Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America

    5 shared publications
  • Rohit Vashisht

    Affiliation Bakar Computational Health Sciences Institute, University of California, San Francisco, San Francisco, California, United States of America

    5 shared publications

Latest Publications

  • Open Access

    Extracting adverse event nature, severity, timelines and resulting interventions from clinical notes of patients receiving CAR-T cell therapy using large language models

    by Jordan Guillot, Brenda Miao, Arvind Suresh, Madhumita Sushil, Christopher Y. K. Williams, Rohit Vashisht, Tomiko T. Oskotsky, Marina Sirota, Atul J. Butte

    Chimeric Antigen Receptor T-cell (CAR‑T) therapy, genetically engineered patient T cells targeting tumor antigens, has transformed care for hematologic malignancies but requires careful tracking of adverse events (AEs) often documented only in unstructured electronic health record (EHR) notes. We evaluated a Large Language Model (LLM)–based approach in UCSF’s secure environment to extract AEs, dates, grades, and interventions within 30 days post‑infusion for six commercial CAR‑T products (2012–2023), benchmarking against two evaluators. Using GPT‑4‑0314 in a zero‑shot setting with four prompts (prespecified AEs, non‑prespecified AEs, CRS, ICANS), we compared outputs against dual annotations on a random sample of 50 notes using accuracy, precision, recall, F1, and Cohen’s kappa. From 4,762 progress notes for 293 patients (median age 65.6), CRS occurred in 80.2% (median onset 4 days); neutropenia 70.0% (16 days); neutropenic fever 64.8% (4 days); ICANS in 34.8%. Interventions included tocilizumab and corticosteroids. Grades were frequently undocumented (CRS 62.3%, ICANS 56.1%); documented cases were mainly CRS grade 1 (59.4%) and ICANS grade 2 (28.0%). Performance was high on CRS and ICANS grading (accuracy of 0.97 and 0.91, respectively). Moderate performances were assessed for prespecified AE extraction (accuracies 0.62–0.76), and non‑prespecified AEs (accuracies 0.76–0.84). Inter‑rater reliability (IRR) was strong for CRS/ICANS presence and grade (kappa 0.86–0.96), moderate for dates and interventions, and weaker for broader AE attributes. LLM‑derived insights can augment AE monitoring and real‑world evidence generation by unlocking unstructured clinical detail and characteristic timelines after CAR T. However, performance varied for broader AE attributes, warranting cautious use. Performance was highest for detecting and grading CRS and ICANS, with strong to near-perfect IRR. While cautious use of LLMs is warranted due to the variable performance observed in this study, these results support further evaluation of supervised CRS/ICANS extraction in controlled EHR-based research or pilot settings, with monitoring and prospective validation.

  • Open Access

    Extracting adverse event nature, severity, timelines and resulting interventions from clinical notes of patients receiving CAR-T therapy using large language models

    by Jordan Guillot, Brenda Miao, Arvind Suresh, Madhumita Sushil, Christopher Yk Williams, Rohit Vashisht, Tomiko T. Oskotsky, Marina Sirota, Atul J. Butte

    Chimeric Antigen Receptor T-cell (CAR-T) therapy, where genetically engineered patient T cells target tumor antigens, has transformed care for hematologic malignancies but requires careful tracking of adverse events (AEs) often documented only in unstructured EHR notes. We evaluated a Large Language Model (LLM)–based approach in UCSF’s secure environment to extract AEs, dates, grades, and interventions within 30 days post-infusion for six commercial CAR-T products (2012–2023), benchmarking against two evaluators. Using GPT-4-0314 in a zero-shot setting with four prompts (prespecified AEs, non-prespecified AEs, CRS, ICANS), we compared outputs against dual annotations on a random sample of 50 notes using accuracy, precision, recall, F1, and Cohen’s kappa. From 4,762 progress notes for 293 patients (median age 65.6), CRS occurred in 80.2% (median onset 4 days); neutropenia 70.0% (16 days); neutropenic fever 64.8% (4 days); ICANS in 34.8%. Interventions included tocilizumab and corticosteroids. Grades were frequently undocumented (CRS 62.3%, ICANS 56.1%); documented cases were mainly CRS grade 1 (59.4%) and ICANS grade 2 (28.0%). Performance was high on CRS and ICANS grading (accuracy of 0.97 and 0.91, respectively). Moderate performances were assessed for prespecified AE extraction (accuracies 0.62–0.76), and non-prespecified AEs (accuracies 0.76–0.84). Inter-rater reliability was strong to near-perfect for CRS/ICANS presence and grade (kappa 0.86–0.96), moderate for dates and interventions, and weaker for broader AE attributes. LLM-derived insights can augment AE monitoring and real-world evidence generation by unlocking unstructured clinical detail and characteristic timelines after CAR T. However, performance varied for broader AE attributes, warranting cautious use. Performance was highest for detecting the presence and grade of CRS and ICANS, with strong to near-perfect inter-rater reliability. While cautious use of LLMs for broad AE extraction is warranted due to the variable performance observed in this study, these results support integrating high-performing CRS/ICANS extraction into EHR workflows. Author summary: Chimeric Antigen Receptor T-cell (CAR-T) therapy has transformed care for blood cancer but requires careful tracking of adverse events (AEs). We asked whether a large language model could read routine clinical notes and extract AEs after CAR T-cell therapy. We analyzed de-identified notes from the first month after infusion. The model identified when two key side effects occurred—cytokine release syndrome (a whole-body inflammatory reaction) and neurotoxicity (brain and nerve symptoms)—and how severe they were, with accuracy similar to human reviewers. It also captured when side effects started and what treatments were given, though performance was more variable for the wider range of side effects beyond these two. In our data, these reactions often arose within the first week; blood count problems and infections were also common. Because many notes did not state severity explicitly, the model sometimes could not assign a grade. Our findings suggest that language models can help unlock important details hidden in clinical notes and could be incorporated into electronic records to support faster, more reliable side-effect monitoring and research. We recommend careful, supervised use and continued validation, especially for broader side-effect categories.

  • Polymédication chronique : description et risque

    by Jordan Guillot

    La polymédication est définie comme l’administration de nombreux médicaments de façon simultanée ou comme l’administration d’un nombre excessif de médicaments. L’objectif général de la thèse était de décrire la polymédication chronique et le risque associé à partir des bases de données médico-administratives. Un indicateur de description de la polymédication chronique, défini par l’usage concomitant et chronique de 5 médicaments ou plus, a été développé pour décrire la prévalence et les médicaments impliqués. À partir des données de l’Assurance Maladie Française, nous avons estimé que la polymédication chronique concernait plus de 5 % des français de tout âge, et impliquait majoritairement les médicaments cardiovasculaires. Afin d’approfondir la caractérisation de la polymédication chronique, une description de la part potentiellement inappropriée a été conduite en utilisant les données de l’Assurance Maladie Française et de la Veterans Health Administration aux États-Unis. À partir de ces données nous avons estimé que la part inappropriée de la polymédication chronique était majeure puisque concernant, en France, 65 % des sujets âgés (≥ 65 ans) et 46 % des sujets d’âge moyen (45-64 ans), et aux États-Unis, 67 % des vétérans. Les inhibiteurs de la pompe à proton, les médicaments psychiatriques et les médicaments du diabète étaient les médicaments potentiellement inappropriés les plus fréquents. La polymédication, qu’elle soit appropriée ou non est un facteur de risque de mortalité. Chez les vétérans américains nous avons montré que la polymédication chronique et les médicaments potentiellement inappropriés augmentaient le risque de décès.

  • Open Access

    Risk Prediction in Patients With Metabolic Dysfunction–Associated Steatohepatitis Using Natural Language Processing

    by Jordan Guillot, Christopher Y.K. Williams, Shadera Azzam, Balu Bhasuran, Gail Fernandes, Boshu Ru, Joe Yang, Xiao Zhang, R. Ravi Shankar, Jin Ge, et al.


  • Contribution of Potentially Inappropriate Medications to Polypharmacy-Associated Risk of Mortality in Middle-Aged Patients: A National Cohort Study

    by Jordan Guillot, Amy C. Justice, Kirsha S. Gordon, Melissa Skanderson, Antoine Pariente, Julien Bezin, Christopher T. Rentsch

    Background The role of potentially inappropriate medications (PIMs) in mortality has been studied among those 65 years or older. While middle-aged individuals are believed to be less susceptible to the harms of polypharmacy, PIMs have not been as carefully studied in this group. Objective To estimate PIM-associated risk of mortality and evaluate the extent PIMs explain associations between polypharmacy and mortality in middle-aged patients, overall and by sex and race/ethnicity. Design Observational cohort study. Setting Department of Veterans Affairs (VA), the largest integrated healthcare system in the US. Participants Patients aged 41 to 64 who received a chronic medication (continuous use of ≥ 90 days) between October 1, 2008, and September 30, 2017. Measurement Patients were followed for 5 years until death or end of study period (September 30, 2019). Time-updated polypharmacy and hyperpolypharmacy were defined as 5–9 and ≥ 10 chronic medications, respectively. PIMs were identified using the Beers criteria (2015) and were time-updated. Cox models were adjusted for demographic, behavioral, and clinical characteristics. Results Of 733,728 patients, 676,935 (92.3%) were men, 479,377 (65.3%) were White, and 156,092 (21.3%) were Black. By the end of follow-up, 104,361 (14.2%) patients had polypharmacy, 15,485 (2.1%) had hyperpolypharmacy, and 129,992 (17.7%) were dispensed ≥ 1 PIM. PIMs were independently associated with mortality (HR 1.11, 95% CI 1.04–1.18). PIMs also modestly attenuated risk of mortality associated with polypharmacy (HR 1.07, 95% CI 1.03–1.11 before versus HR 1.05, 95% CI 1.01–1.09 after) and hyperpolypharmacy (HR 1.18, 95% CI 1.09–1.28 before versus HR 1.12, 95% CI 1.03–1.22 after). Patterns varied when stratified by sex and race/ethnicity. Limitations The predominantly male VA patient population may not represent the general population. Conclusion PIMs were independently associated with increased mortality, and partially explained polypharmacy-associated mortality in middle-aged people. Other mechanisms of injury from polypharmacy should also be studied.


  • Potentially inappropriate medication use by level of polypharmacy among US Veterans 49–64 and 65–70 years old

    by Jordan Guillot, Christopher T. Rentsch, Kirsha S. Gordon, Amy C. Justice, Julien Bezin

    Background: Potentially inappropriate medications (PIMs) are medications contra‐indicated in particular circumstances. We sought to characterize PIMs by level of polypharmacy by age, sex, and race/ethnicity.Methods: We performed a cross‐sectional drug dispensing study using electronic health records available through the US Department of Veterans Affairs. We extracted pharmacy fill and refill records during fiscal year 2016 (i.e., October 1, 2015–September 30, 2016) for all patients aged 49–70 who accessed care in the preceding fiscal year. PIMs were defined by the combined Beers and Laroche (henceforth Beers Laroche) criteria used for older patients and the PROMPT criteria used for middle‐aged.Results: In the 1 499 586 patients aged 49–64, PIMs prevalence by PROMPT in patients with 0–4, 5–9, and ≥10 medications was 14.0%, 62.2%, and 86.1%, respectively, and by Beers Laroche was 14.3%, 63.4%, and 85.7%, respectively. In the 1 249 119 patients aged 65–70, PIMs prevalence by Beers Laroche was 14.8%, 59.9%, and 83.3%, and by PROMPT was 13.9%, 57.4%, and 82.0%, respectively. Meaningful differences in prevalence were shown by sex and race/ethnicity according to both set of criteria (e.g. PROMPT in patients with 5–9 medications: 66.1% women vs. 59.3% men; standardized‐mean‐differences [SMD] = 0.14; 61.7% of White vs. 54.5% of non‐White; SMD = 0.15). The most common PIMs were digestive, analgesic, antidiabetic, and psychotropic medications.Conclusion: Prevalence of PIMs was high and increased with polypharmacy. Beers Laroche and PROMPT provided similar estimations inside and outside their target age, suggesting that PIMs are common among those with polypharmacy regardless of age.


  • Chronic polypharmacy at all age: A population‐based drug utilization study

    by Jordan Guillot, Sandy Maumus‐Robert, Antoine Pariente, Julien Bezin

    Polypharmacy is a growing concern often described only in older people by cumulating all drugs taken. We aimed to describe chronic polypharmacy in France, regardless of age. A cross sectional descriptive study was performed using the 1/97th representative sample of the French health insurance nationwide database (EGB). All subjects alive on January 1st, 2015 and covered by the French healthcare insurance were included, and their information collected until December 31st, 2015 or date of death. Drug exposures were estimated from drug dispensing dates and treatment durations. Chronic uses of drug were defined as drugs used daily for more than 6 months. Chronic polypharmacy corresponded to the exposure to five chronic uses of drug or more. In 2015, information of 584,862 subjects was collected (mean age: 42.2 years). Prevalence of chronic polypharmacy was 5.6% and incidence 1.1%. Prevalence of chronic polypharmacy increased noticeably from 0.2% for subjects aged 18 to 40 years to a maximum of 29.2% for subjects aged 80 to 90 years, and then decreased to 23.6% for subjects aged 90 years and more. Lipid modifying agents were the most frequent drugs involved in chronic polypharmacy (10% of exposure). According to age, the most important differences between the younger and older people were found for cardiovascular drugs (43.5% vs. 45.7% of exposure, respectively) and nervous system drugs (13.7% vs. 11.5% of exposure, respectively). This population-based study showed increasing of chronic polypharmacy and evolution of chronic drug patterns with age.


  • Open Access

    The Burden of Potentially Inappropriate Medications in Chronic Polypharmacy

    by Jordan Guillot, Sandy Maumus-Robert, Alexandre Marceron, Pernelle Noize, Antoine Pariente, Julien Bezin

    We aimed to describe the burden represented by potentially inappropriate medications (PIMs) in chronic polypharmacy in France. We conducted a nationwide cross-sectional study using data from the French National Insurance databases. The study period was from 1 January 2016 to 31 December 2016. Chronic drug use was defined as uninterrupted daily use lasting ≥6 months. Chronic polypharmacy was defined as the chronic use of ≥5 medications, and chronic hyperpolypharmacy as the chronic use of ≥10 medications. For individuals aged ≥65 (older adults), PIMs were defined according to the Beers and Laroche lists, and for individuals aged 45–64 years (middle-aged) PIMs were defined according to the PROMPT (Prescribing Optimally in Middle-aged People’s Treatments) list. Among individuals with chronic polypharmacy, 4009 (46.2%) middle-aged and 18,036 (64.8%) older adults had at least one chronic PIM. Among individuals with chronic hyperpolypharmacy, these figures were, respectively, 570 (75.0%) and 2544 (88.7%). The most frequent chronic PIM were proton pump inhibitors (43.4% of older adults with chronic polypharmacy), short-acting benzodiazepines (older adults: 13.7%; middle-aged: 16.1%), hypnotics (6.1%; 7.4%), and long-acting sulfonylureas (3.9%; 12.3%). The burden of chronic PIM appeared to be very high in our study, concerning almost half of middle-aged adults and two-thirds of older adults with chronic polypharmacy. Deprescribing interventions in polypharmacy should primarily target proton pump inhibitors and hypnotics.


  • Polypharmacy: A general review of definitions, descriptions and determinants

    by Jordan Guillot, Sandy Maumus-Robert, Julien Bezin


  • Open Access

    Extracting adverse event nature, severity, timelines and resulting interventions from clinical notes of patients receiving CAR-T cell therapy using large language models

    by Jordan Guillot, Brenda Miao, Arvind Suresh, Madhumita Sushil, Christopher Y. K. Williams, Rohit Vashisht, Tomiko T. Oskotsky, Marina Sirota, Atul J. Butte