Chethan Sarabu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Chethan Sarabu has a total of 19 co-authors. Here are the top co-authors by number of shared publications:
- Catherine M. DesRoches
Harvard Medical School, Open Notes Labs, Beth Israel Deaconess Medical Center , Boston, MA 02215, United States
3 shared publications - Liz Salmi
Beth Israel Deaconess Medical Center, Boston, MA
3 shared publications - Amanda Norris
Open Notes Labs, Beth Israel Deaconess Medical Center , Boston, MA 02215, United States
2 shared publications - Betsy Lowe
Patient and Family Advisory Council, Beth Israel Deaconess Medical Center, Boston, Massachusetts
2 shared publications - Grayson Schultz2 shared publications
- Jody Naimark
Department of Family Medicine, Winchester Hospital, Winchester, MA, United States
2 shared publications
Latest Publications
- Open Access
- 29 July 2026
- Published by JMIR Publications Inc. in JMIR AI
What Patients Want in a Visit Summary Generated by Ambient Scribes: A Structured Co-Design Method (Preprint)
- 29 July 2026
- Published by JMIR Publications Inc. in JMIR AI
Open AccessWhat Patients Want in a Visit Summary Generated by Ambient Scribes: A Structured Co-Design Method (Preprint)
- Open Access
- 29 July 2026
- Published by JMIR Publications Inc. in JMIR Preprints
What Patients Want from an Ambient Scribe: A Structured Co-Design Method (Preprint)
BACKGROUND AI-powered ambient scribes are rapidly being adopted in healthcare settings, with growing interest in their potential to generate patient-facing visit summaries. While research has demonstrated that patients benefit from reading clinical notes, prior visit summaries were designed primarily to satisfy regulatory requirements rather than patient needs. Most existing ambient scribe research focuses on clinician perspectives, particularly regarding documentation burden and burnout, leaving patient preferences largely unexamined. OBJECTIVE This study aimed to engage patients as co-designers to understand what they want and need from an ambient scribe. METHODS Between October 2024 and April 2025, the research team conducted five structured virtual engagement activities with 16 members of the OpenNotes Lab Insight Panel, including 14 patients and 2 clinicians. Panelists were diverse across race, age, gender, disability status, health literacy, language, geography, and educational background. Activities involved observation of role-played clinical visits (a chronic conditions follow-up and a potential cancer diagnosis visit), review of ambient scribe–generated clinical notes and visit summaries, and structured discussion using a modified 1-2-4-all method. A Rapid Assessment Procedure (RAP) was used to analyze transcripts, meeting chats, and participant notes across all five sessions, with iterative coding by four team members and findings shared back to panelists. RESULTS Nine themes emerged and were grouped into three categories. Development themes included: Practicality (patients wanted clear tasks, timelines, defined responsibilities, and contingency plans); Language & Tone (preferred reading level ranged from 4th to 8th grade; patients wanted empathetic, personalized, non-judgmental language, particularly for sensitive visits); Accessibility (patients wanted multi-modal access including audio, print, and digital formats, with screen reader compatibility and portal shareability); and Clarity & Accuracy (panelists identified factual errors and omissions, including misattribution of anxiety as a diagnosis rather than a situational response, and wanted patient-reported information reflected). Implementation themes included Transparency (unanimous expectation of disclosure that content was AI-generated) and Patient Choice (universal desire for opt-out options and preference portability). Additional themes addressed workflow improvements, logistical challenges around note finalization timing, and desired outcome measures. CONCLUSIONS Patients primarily want ambient scribe–generated visit summaries to function as practical care navigation tools with clear action items, transparent AI disclosure, and flexible formats responsive to both patient context and visit type. Realizing this potential will require collaboration among ambient scribe developers, EHR vendors, health systems, and patients, rather than layering new tools onto existing inefficient workflows.
- 29 July 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessWhat Patients Want from an Ambient Scribe: A Structured Co-Design Method (Preprint)
BACKGROUND AI-powered ambient scribes are rapidly being adopted in healthcare settings, with growing interest in their potential to generate patient-facing visit summaries. While research has demonstrated that patients benefit from reading clinical notes, prior visit summaries were designed primarily to satisfy regulatory requirements rather than patient needs. Most existing ambient scribe research focuses on clinician perspectives, particularly regarding documentation burden and burnout, leaving patient preferences largely unexamined. OBJECTIVE This study aimed to engage patients as co-designers to understand what they want and need from an ambient scribe. METHODS Between October 2024 and April 2025, the research team conducted five structured virtual engagement activities with 16 members of the OpenNotes Lab Insight Panel, including 14 patients and 2 clinicians. Panelists were diverse across race, age, gender, disability status, health literacy, language, geography, and educational background. Activities involved observation of role-played clinical visits (a chronic conditions follow-up and a potential cancer diagnosis visit), review of ambient scribe–generated clinical notes and visit summaries, and structured discussion using a modified 1-2-4-all method. A Rapid Assessment Procedure (RAP) was used to analyze transcripts, meeting chats, and participant notes across all five sessions, with iterative coding by four team members and findings shared back to panelists. RESULTS Nine themes emerged and were grouped into three categories. Development themes included: Practicality (patients wanted clear tasks, timelines, defined responsibilities, and contingency plans); Language & Tone (preferred reading level ranged from 4th to 8th grade; patients wanted empathetic, personalized, non-judgmental language, particularly for sensitive visits); Accessibility (patients wanted multi-modal access including audio, print, and digital formats, with screen reader compatibility and portal shareability); and Clarity & Accuracy (panelists identified factual errors and omissions, including misattribution of anxiety as a diagnosis rather than a situational response, and wanted patient-reported information reflected). Implementation themes included Transparency (unanimous expectation of disclosure that content was AI-generated) and Patient Choice (universal desire for opt-out options and preference portability). Additional themes addressed workflow improvements, logistical challenges around note finalization timing, and desired outcome measures. CONCLUSIONS Patients primarily want ambient scribe–generated visit summaries to function as practical care navigation tools with clear action items, transparent AI disclosure, and flexible formats responsive to both patient context and visit type. Realizing this potential will require collaboration among ambient scribe developers, EHR vendors, health systems, and patients, rather than layering new tools onto existing inefficient workflows.
- Top Cited Papers
- 22 May 2025
- Published by Massachusetts Medical Society in NEJM AI
- Vol. 2 (6)
People Overtrust AI-Generated Medical Advice despite Low Accuracy
- 22 May 2025
- Published by Massachusetts Medical Society in NEJM AI
- Vol. 2 (6)
Top Cited PapersPeople Overtrust AI-Generated Medical Advice despite Low Accuracy
- Open Access
- 6 March 2025
- Published by Oxford University Press (OUP) in JAMIA Open
- Vol. 8 (2) , ooaf021
A proof-of-concept study for patient use of open notes with large language models
Objectives: The use of large language models (LLMs) is growing for both clinicians and patients. While researchers and clinicians have explored LLMs to manage patient portal messages and reduce burnout, there is less documentation about how patients use these tools to understand clinical notes and inform decision-making. This proof-of-concept study examined the reliability and accuracy of LLMs in responding to patient queries based on an open visit note. Materials and Methods: In a cross-sectional proof-of-concept study, 3 commercially available LLMs (ChatGPT 4o, Claude 3 Opus, Gemini 1.5) were evaluated using 4 distinct prompt series—Standard, Randomized, Persona, and Randomized Persona—with multiple questions, designed by patients, in response to a single neuro-oncology progress note. LLM responses were scored by the note author (neuro-oncologist) and a patient who receives care from the note author, using an 8-criterion rubric that assessed Accuracy, Relevance, Clarity, Actionability, Empathy/Tone, Completeness, Evidence, and Consistency. Descriptive statistics were used to summarize the performance of each LLM across all prompts. Results: Overall, the Standard and Persona-based prompt series yielded the best results across all criterion regardless of LLM. Chat-GPT 4o using Persona-based prompts scored highest in all categories. All LLMs scored low in the use of Evidence. Discussion: This proof-of-concept study highlighted the potential for LLMs to assist patients in interpreting open notes. The most effective LLM responses were achieved by applying Persona-style prompts to a patient’s question. Conclusion: Optimizing LLMs for patient-driven queries, and patient education and counseling around the use of LLMs, have potential to enhance patient use and understanding of their health information.
- 6 March 2025
- Published by Oxford University Press (OUP) in JAMIA Open
- Vol. 8 (2) , ooaf021
Open AccessA proof-of-concept study for patient use of open notes with large language models
Objectives: The use of large language models (LLMs) is growing for both clinicians and patients. While researchers and clinicians have explored LLMs to manage patient portal messages and reduce burnout, there is less documentation about how patients use these tools to understand clinical notes and inform decision-making. This proof-of-concept study examined the reliability and accuracy of LLMs in responding to patient queries based on an open visit note. Materials and Methods: In a cross-sectional proof-of-concept study, 3 commercially available LLMs (ChatGPT 4o, Claude 3 Opus, Gemini 1.5) were evaluated using 4 distinct prompt series—Standard, Randomized, Persona, and Randomized Persona—with multiple questions, designed by patients, in response to a single neuro-oncology progress note. LLM responses were scored by the note author (neuro-oncologist) and a patient who receives care from the note author, using an 8-criterion rubric that assessed Accuracy, Relevance, Clarity, Actionability, Empathy/Tone, Completeness, Evidence, and Consistency. Descriptive statistics were used to summarize the performance of each LLM across all prompts. Results: Overall, the Standard and Persona-based prompt series yielded the best results across all criterion regardless of LLM. Chat-GPT 4o using Persona-based prompts scored highest in all categories. All LLMs scored low in the use of Evidence. Discussion: This proof-of-concept study highlighted the potential for LLMs to assist patients in interpreting open notes. The most effective LLM responses were achieved by applying Persona-style prompts to a patient’s question. Conclusion: Optimizing LLMs for patient-driven queries, and patient education and counseling around the use of LLMs, have potential to enhance patient use and understanding of their health information.