About

Hamish S Fraser is an Associate Professor of Medical Science and Health Services, Policy and Practice at Brown University. He trained in General Medicine, Cardiology, and Knowledge Based Systems at Edinburgh University in Scotland, and in Clinical Decision Making at MIT and the New England Medical Center. His work focuses on the evaluation of medical information systems, particularly in Low and Middle Income Countries, and understanding the impact of information and communications on healthcare quality worldwide. Fraser has contributed significantly to global health informatics, leading projects that migrate medical informatics tools from high-income to challenging low-income environments, including evaluating electronic health record systems for HIV care in Rwanda and deploying prediction models to improve HIV treatment outcomes in Kenya. He has held grants from organizations such as the CDC, the Rockefeller Foundation, IDRC Canada, and the European Union Horizon 2020 program. Fraser is also involved in developing educational resources, including co-editing a textbook on Global Health Informatics and Digital Health. His recent work includes evaluating symptom checker apps, large language models, and machine learning techniques to enhance diagnostic accuracy and decision support in healthcare settings.

Research topics

• Medicine
• Nursing
• Computer Science
• Business
• Family medicine
• Engineering
• Demography
• World Wide Web
• Knowledge management

Selected publications

• What the AI era doctor should know: a scoping review of proposed artificial intelligence competencies for medical education

  npj Digital Medicine · 2026-05-12

  articleOpen access

  Artificial intelligence (AI) is rapidly reshaping healthcare and the competencies expected of graduating medical students, yet AI curricula and competency recommendations for undergraduate medical education (UME) remain fragmented. We conducted a PRISMA-ScR scoping review to map and synthesize proposed AI competencies for UME by performing a global search of PubMed, Embase, Web of Science, and ERIC without language restrictions and from database inception through July 28, 2025. Verbatim competency-relevant text was extracted and decomposed into discrete statements and classified using domains, competencies, or learning objectives. Statement frequencies were summarized to characterize recurring areas of emphasis, underrepresented topics, and cross-domain relationships. Of 4071 records identified, 54 studies from 22 countries met inclusion criteria. From 564 eligible statements, we synthesized a taxonomy comprising seven domains (AI ethics; AI law and regulation; AI professionalism in healthcare; clinical applications of AI; critical appraisal of AI output; research and innovation in AI; theory and foundations of AI) spanning 37 competencies and 170 learning objectives. Sources were predominantly editorial/opinion, with recurring emphasis on ethicolegal oversight, critical appraisal of AI outputs, and foundational understanding of AI methods and data. This synthesis provides a structured inventory to inform curriculum planning and future stakeholder-based refinement, prioritization, and evaluation.

  PublisherOA PDFDOI

• The Persona Paradox: Medical Personas as Behavioral Priors in Clinical Language Models

  ArXiv.org · 2026-01-08

  articleOpen access

  Persona conditioning can be viewed as a behavioral prior for large language models (LLMs) and is often assumed to confer expertise and improve safety in a monotonic manner. However, its effects on high-stakes clinical decision-making remain poorly characterized. We systematically evaluate persona-based control in clinical LLMs, examining how professional roles (e.g., Emergency Department physician, nurse) and interaction styles (bold vs.\ cautious) influence behavior across models and medical tasks. We assess performance on clinical triage and patient-safety tasks using multidimensional evaluations that capture task accuracy, calibration, and safety-relevant risk behavior. We find systematic, context-dependent, and non-monotonic effects: Medical personas improve performance in critical care tasks, yielding gains of up to $\sim+20\%$ in accuracy and calibration, but degrade performance in primary-care settings by comparable margins. Interaction style modulates risk propensity and sensitivity, but it's highly model-dependent. While aggregated LLM-judge rankings favor medical over non-medical personas in safety-critical cases, we found that human clinicians show moderate agreement on safety compliance (average Cohen's $κ= 0.43$) but indicate a low confidence in 95.9\% of their responses on reasoning quality. Our work shows that personas function as behavioral priors that introduce context-dependent trade-offs rather than guarantees of safety or expertise. The code is available at https://github.com/rsinghlab/Persona\_Paradox.

  PublisherOA PDF

• What Should the AI Era Doctor Know? A Scoping Review of Proposed Artificial Intelligence Competencies for Medical Education

  Research Square · 2026-03-13

  preprintOpen access
  PublisherOA PDFDOI

• The Persona Paradox: Medical Personas as Behavioral Priors in Clinical Language Models

  arXiv (Cornell University) · 2026-01-08

  preprintOpen access

  Persona conditioning can be viewed as a behavioral prior for large language models (LLMs) and is often assumed to confer expertise and improve safety in a monotonic manner. However, its effects on high-stakes clinical decision-making remain poorly characterized. We systematically evaluate persona-based control in clinical LLMs, examining how professional roles (e.g., Emergency Department physician, nurse) and interaction styles (bold vs.\ cautious) influence behavior across models and medical tasks. We assess performance on clinical triage and patient-safety tasks using multidimensional evaluations that capture task accuracy, calibration, and safety-relevant risk behavior. We find systematic, context-dependent, and non-monotonic effects: Medical personas improve performance in critical care tasks, yielding gains of up to $\sim+20\%$ in accuracy and calibration, but degrade performance in primary-care settings by comparable margins. Interaction style modulates risk propensity and sensitivity, but it's highly model-dependent. While aggregated LLM-judge rankings favor medical over non-medical personas in safety-critical cases, we found that human clinicians show moderate agreement on safety compliance (average Cohen's $κ= 0.43$) but indicate a low confidence in 95.9\% of their responses on reasoning quality. Our work shows that personas function as behavioral priors that introduce context-dependent trade-offs rather than guarantees of safety or expertise. The code is available at https://github.com/rsinghlab/Persona\_Paradox.

  PublisherDOI

• Principles and implementation strategies for equitable and representative academic partnerships in global health informatics research

  Journal of the American Medical Informatics Association · 2025-01-17 · 3 citations

  articleOpen access

  OBJECTIVE: Developing equitable, sustainable informatics solutions is key to scalability and long-term success for projects in the global health informatics (GHI) domain. This paper presents key strategies for incorporating principles of health equity in the GHI project lifecycle. MATERIALS AND METHODS: The American Medical Informatics Association (AMIA) GHI Working Group organized a collaborative workshop at the 2023 AMIA Annual Symposium that included the presentation of five case studies of how principles of health equity have been incorporated into projects situated in low-and-middle-income countries and with Indigenous communities in the U.S. and best practices for operationalizing these principles into other informatics projects. RESULTS: We present five principles: (1) Inclusion and Participation in Ethical, Sustainable Collaborations; (2) Engaging Community-Based Participatory Research Approaches; (3) Stakeholder Engagement; (4) Scalability and Sustainability; (5) Representation in Knowledge Creation, along with strategies that informatics researchers may use to incorporate these principles into their work. DISCUSSION: Presented case studies and subsequent focus groups yielded key concepts and strategies to promote health equity that may be operationalized across GHI projects. CONCLUSION: Equitable, sustainable, and scalable GHI projects require intentional integration of community and stakeholder perspectives in project development, implementation, and knowledge creation processes.

  PublisherOA PDFDOI

• Gaps and Pathways to Success in Global Health Informatics Academic Collaborations: Reflecting on Current Practices

  JMIR Medical Informatics · 2025-10-23 · 1 citations

  articleOpen access

  Unlabelled: Academic global health informatics (GHI) projects are impactful collaborations between institutions in high-income and low- and middle-income countries (LMICs) and play a crucial role in enhancing health care services and access in LMICs using eHealth practices. Researchers across all involved organizations bring unique expertise to these collaborations. However, these projects often face significant obstacles, including cultural and linguistic barriers, resource limitations, and sustainability issues. The lack of representation from LMIC researchers in knowledge generation and the high costs of open-access publications further complicate efforts to ensure inclusive, accessible, and collaborative scholarship. This viewpoint describes present gaps in the literature on academic GHI collaborations and describes a path forward for future research directions and successful research community development. Key recommendations include centering community-based participatory research, developing post-growth solutions, and creating sustainable funding models. Addressing these challenges is essential for fostering effective, scalable, and equitable GHI interventions that improve global health outcomes.

  PublisherDOI

• Medical School Admission in the Age of Generative AI (Preprint)

  2025-07-09

  preprintSenior author

  <sec> <title>UNSTRUCTURED</title> Storytelling, a timeless art form, is central to the human identity and serves as a bridge across time and place. In medical school admissions, the personal statement prompts applicants to share their story and connect with admission committees through this deeply human medium. However, generative AI tools like ChatGPT blur the line between genuine self-expression and algorithmic assistance, raising concerns about the authenticity of these narratives. This challenge raises questions about the personal statement's role in evaluating candidates and the ethical implications of AI in the medical school admission process. As admission committees adapt to this technology, they must balance fairness, accessibility, and authenticity in evaluating the personal statement while also preparing for its eventual replacement. </sec>

  PublisherDOI

• The impact of climate change on vulnerable populations in pediatrics: opportunities for AI, digital health, and beyond—a scoping review and selected case studies

  Pediatric Research · 2025-01-29 · 10 citations

  articleOpen access

  Climate change critically impacts global pediatric health, presenting unique and escalating challenges due to children's inherent vulnerabilities and ongoing physiological development. This scoping review intricately intertwines the spheres of climate change, pediatric health, and Artificial Intelligence (AI), with a goal to elucidate the potential of AI and digital health in mitigating the adverse child health outcomes induced by environmental alterations, especially in Low- and Middle-Income Countries (LMICs). A notable gap is uncovered: literature directly correlating AI interventions with climate change-impacted pediatric health is scant, even though substantial research exists at the confluence of AI and health, and health and climate change respectively. We present three case studies about AI's promise in addressing pediatric health issues exacerbated by climate change. The review spotlights substantial obstacles, including technical, ethical, equitable, privacy, and data security challenges in AI applications for pediatric health, necessitating in-depth, future-focused research. Engaging with the intricate nexus of climate change, pediatric health, and AI, this work underpins future explorations into leveraging AI to navigate and neutralize the burgeoning impact of climate change on pediatric health outcomes. IMPACT: Our scoping review highlights the scarcity of literature directly correlating AI interventions with climate change-impacted pediatric health that disproportionately affects vulnerable populations, even though substantial research exists at the confluence of AI and health, and health and climate change respectively. We present three case studies about AI's promise in addressing pediatric health issues exacerbated by climate change. The review spotlights substantial obstacles, including technical, ethical, equitable, privacy, and data security challenges in AI applications for pediatric health, necessitating in-depth, future-focused research.

  PublisherOA PDFDOI

• Abstract 66: Diagnostic Accuracy of ChatGPT4.o for TIA or Stroke Using Patient Symptoms and Demographics

  Stroke · 2025-01-30

  article

  Introduction: Many patients may not recognize the initial symptoms of TIA or stroke and delay seeking urgent medical care, leading to missed treatments and worse outcomes. Diagnostic decision support (DDSS) systems may help these patients recognize and act on early symptoms of stroke or TIA. Large language models (LLMs) are available and utilized by the public. We evaluated the efficacy of GPT 4.o, a recent LLM, for the prediction of stroke or TIA from data collected at presentation among patients admitted to an emergency department TIA/Stroke observational unit (ED-OU). Methods: 1466 patients admitted to the ED-OU for suspected TIA in a large, urban, academic ED, from 3/2013 - 2/2020 were included. A thorough history and physical were obtained, including presenting symptoms, symptom time course, vital signs, and past medical history. The outcome was a discharge diagnosis of stroke, TIA, or an alternate diagnosis, from a consulting neurologist and confirmed with neuroimaging. For a random sample of 500 records demographics and symptom data were entered into GPT 4.o via OpenAI’s Application Programming Interface (API). We used a validated prompt that requests the 5 most likely diagnoses. The sensitivity of the GPT 4.o diagnoses for the neurologist’s diagnosis was calculated as a match in the top 1 (M1), top 3 (M3) or top 5 (M5). Sensitivity to a combined diagnosis of TIA or stroke was also calculated. Results were compared to (1) a random sample of 100 patients from the same dataset manually entered into the ChatGPT4.o interface by a research assistant, and (2) results from evaluation of ChatGPT 4.0 with data from a study of patients requesting urgent primary care, who entered their clinical data into a DDSS app. Results: The 500 cases included 257 with TIA, 73 with stroke, and 170 with other diagnoses. Table 1 shows results of diagnostic matches. Diagnostic lists of 18.4% of cases had no match between GPT4.o and the neurologist’s diagnosis. 5.4% of cases had no clear neurologist diagnosis. GPT4 sensitivity just for combined diagnosis of TIA or Stroke was 98.8%. Conclusions: DDSS like ChatGPT/GPT4.o have potential to aid patients’ prompt recognition of TIA or stroke symptoms which could shorten time to care. To better define usability, accuracy and safety of DDSS, we are studying direct data collection from patients in the ED or urgent primary care, including stroke patients or their companions, and evaluating other DDSS tools including symptom checkers.

  PublisherDOI

• Open Source Software in Healthcare: International Case Series from the IMIA Open Source Working Group

  Studies in health technology and informatics · 2024-01-25 · 1 citations

  articleOpen access

  In this case series, we demonstrate how open-source software has been widely adopted as the primary health information system in many low- and middle-income countries, and for government-developed applications in high-income settings. We discuss the concept of Digital Global Goods and how the general approach of releasing software developed through public funding under open-source licences could improve the delivery of healthcare in all settings through increased transparency and collaboration as well as financial efficiency.

  PublisherOA PDFDOI

Frequent coauthors

• Joaquín A. Blaya

  54 shared

• Darius Jazayeri

  46 shared

• Kenneth D. Mandl

  Harvard University

  41 shared

• Christodoulos Stefanadis

  Athens Medical Center

  38 shared

• Bryant T. Karras

  Briggs & Stratton (United States)

  36 shared

• Michael M. Wagner

  University of Pittsburgh

  36 shared

• Jeremy U. Espino

  University of Pittsburgh

  36 shared

• Lisa Trigg

  36 shared

Education

• M.D., General Medicine, Cardiology and Knowledge Based Systems

  Edinburgh University

• Other, Clinical Decision Making

  MIT and the New England Medical Center

• M.D.

  Harvard Medical School

  2006

Awards & honors

• Marie Skłodowska-Curie Fellowship from the European Union Ho…