An interventional study of LLM-Generated Medical Note and Control (Original Medical Note) in Musculoskeletal Disease and Orthopedic Patients, sponsored by University of Texas at Austin. Not yet recruiting. Open to participants aged 18 Years to 89 Years. Per ClinicalTrials.gov, last updated 2026-05-22.
Sponsored by University of Texas at Austin · Not applicable, Interventional, and Health services research
The goal of this clinical trial is to learn how patients feel when reading their medical notes. The study compares reading the original doctor's note with reading a simpler, Artificial Intelligence (AI)-generated version written in plain language in adults receiving musculoskeletal specialty care.
The main questions the study aims to answer are:
Researchers will compare patients who read their original clinic note with patients who read an Artificial Intelligence (AI)-generated plain-language summary to see whether simpler language changes patient understanding, trust, or emotional responses.
Participants will:
Background As patient access to electronic medical records becomes more widespread, understanding how individuals emotionally respond to their clinical documentation has become increasingly important. A qualitative study analyzing 600 medical encounter notes authored by 138 clinicians identified distinct patterns of language reflecting clinicians' attitudes toward patients. Five categories of negative language were described (questioning credibility, disapproval, stereotyping, labeling patients as "difficult," and unilateral decision-making), alongside six categories of positive language (compliments, approval, self-disclosure, minimizing blame, personalization, and collaborative decision-making). Prior research has demonstrated that access to clinical notes ('open notes') can influence patient experience, with potential benefits including improved understanding, greater trust, enhanced perceived quality of care, and increased engagement in self-care and health management. These benefits appear particularly pronounced with patients with lower education attainment or those from ethnic minority groups. Despite this, fewer than half of clinicians routinely discuss shared notes with patients during visits.
Rationale Language choices within clinical documentation may meaningfully shape how patients perceive their diagnosis, their relationship with healthcare providers, and their emotional well-being. For example, in a survey study of 100 healthy companions of orthopedic hand surgery patients, participants evaluated 19 commonly used medical terms and their synonyms using the Self-Assessment Manikin (SAM). Terms such as "pain" was rated more negatively than alternatives such as "discomfort" or "ache", while "rupture" elicited a more negative response than "tear" or "defect". These findings suggest that frequently used clinical terminology may also unintentionally elicit negative emotional reactions. Patients may be particularly vulnerable to language-related distress when reading their own medical records. The tone, complexity, and structure of these notes, especially in surgical specialties, may impact patient trust and overall comfort, and, in some cases, may unintentionally erode trust or intensify anxiety. Emerging evidence suggests that plain-language summaries might mitigate these effects. In a small study of 20 participants, artificial intelligence was used to generate plain-language medical notes, which were perceived as more useful, supportive of the patient-clinician relationship, and empowering for patients' understanding in their health. Although the importance and benefits of accessible visit notes have been well established, there remains limited empirical evidence regarding the optimal tone, structure, and language of medical documentation from the patient perspective.
Hypotheses
Primary hypothesis:
There are no factors associated with patient ratings of trust and experience with the clinician (TRECS), including whether patients viewed their original medical record entry or an LLM-generated plain language summary based on a curated version of the record (where PHI is removed) prior to the visit and other patient factors (demographics, mental health clusters).
Secondary hypotheses:
There are no factors associated with emotional responses to viewing one's own medical record entry, including whether patients view the original entry or an LLM-generated plain language summary based on a curated version of the record (where PHI is removed) prior to the visit and other patient factors (demographics, mental health clusters).
Qualitative hypothesis:
What themes does an LLM identify in patient verbatim comments regarding viewing their medical record entry or an LLM plain language summary?
657 studies on the registry are indexed under Musculoskeletal Diseases; 159 are open to participants now.
This study's planned enrollment of 135 is above the median of 73 across 438 interventional studies indexed under Musculoskeletal Diseases.
Browse Musculoskeletal Diseases studies →University of Texas at Austin is the lead sponsor of 319 studies on the registry; 78 are open to participants now.
Of its 10 completed or terminated interventional studies of FDA-regulated products, 5 (50%) have results posted.
Counted across the registry records on this site, refreshed daily.
Exclusion criteria:
- Any impairment precluding completion of a survey on a tablet
Participants randomized to the intervention arm will review a plain-language summary generated from a curated version of their prior musculoskeletal clinic note using a large language model (LLM). All protected health information (PHI) and identifying information will be removed prior to LLM processing. The summary will be designed to simplify medical terminology and improve readability while maintaining the original clinical meaning of the note. The specific LLM used has not yet been finalized but will likely consist of the most current version of ChatGPT and/or Perplexity available through institutionally approved platforms at the time of the study. Participants will review the summary on an iPad prior to their clinic visit and complete questionnaires assessing trust, experience, and emotional responses.
Behavioral: LLM-Generated Medical Note
Participants randomized to the control arm will review the original clinical note from their prior musculoskeletal clinic visit. The note will be presented in its original format without language simplification. Participants will review the note on an iPad prior to their clinic visit and complete questionnaires assessing trust, experience, and emotional responses.
Behavioral: Control (Original Medical Note)
The intervention consists of presenting participants with an LLM-generated plain-language summary of a prior musculoskeletal clinic note. The summary will be produced from a curated version of the original documentation after removal of all protected health information (PHI), macros, and administrative content. Physical therapy notes will be excluded, and normal examination or imaging findings may be simplified (e.g., "Exam otherwise normal"). The LLM will be instructed to preserve clinical meaning while reducing jargon and improving readability. The summary will be concise, neutral in tone, and written in patient-friendly language without adding new medical information or altering clinical recommendations. The specific LLM platform is not yet finalized but will likely use the most current version of ChatGPT and/or Perplexity available through institutional access.
Participants randomized to the control arm will review the original clinical note from their prior musculoskeletal clinic visit. The note will be presented in its original format without language simplification. Participants will review the note on an iPad prior to their clinic visit and complete questionnaires assessing trust, experience, and emotional responses.
Trust and Experience with the Clinician Scale (TRECS-7)
The Trust and Experience with the Clinician Scale (TRECS-7) is a validated 7-item scale that measures patients' trust in and experience with their clinician during a medical consultation. Designed to minimize ceiling effects, it enables more sensitive detection of variation in patient experience across different clinical interactions (Brinkman et al.). Each of 7 statements is scored from 0-4 (strongly disagree, disagree, neutral, agree, strongly agree), resulting in a total score between 0 and 28. Higher scores indicate greater perceived trust in the clinician. Source: Brinkman N, Looman R, Jayakumar P, Ring D, Choi S. Is It Possible to Develop a Patient-reported Experience Measure With Lower Ceiling Effect? Clin Orthop Relat Res. 2025 Apr 1;483(4):693-703. Time Frame: Measured once, immediately following consultation with the musculoskeletal specialist
Time frame: Immediately after visit
Emotional response to medical note
Emotional responses to viewing one's own medical record will be assessed using 0-100 sliding scales measuring emotional comfort, anxiety, perceived clinician caring, and perceived impact on communication with the clinician. Higher or lower scores will reflect participants' position between paired emotional anchors including sad/happy, worried/at ease, doctor is caring/doctor is not caring, uncomfortable/comfortable, and affects communication/does not affect communication.
Time frame: Immediately after intervention
Themes identified by the LLM in verbatim text
Themes identified from participant verbatim comments regarding their experience reading the study material. Participants will respond to the free-text prompt: "How did you feel reading your medical records before the visit with the clinician? (Please elaborate)" Responses will be analyzed using a large language model (LLM) to identify recurring themes related to understanding, emotional reactions, perceived trust, clarity of information, comfort or distress, and anticipated impact on communication with the clinician.
Time frame: Collected immediately after intervention/control
No study locations are listed for this record.
Documents are hosted by the registry — open the source record to download them.
Plan to share: No
This study is not yet recruiting, as verified in May 2026. You cannot join it, but the record below documents what was studied.
Get an email when the registry record changes — status, dates, results — or when someone posts here.
Sign in to followQuestions and observations about this study, from anyone following it. Not medical advice, and not a channel to the study team — their contact details are on the registry record.
Sign in to join the discussion. Reading takes no account; posting does. You choose a display name, and a pseudonym is the default.
Nothing here yet. If you are running this trial, taking part in it, or weighing whether to, this is the place to say so.
University of Texas at Austin