Analysis of Patient Education in Trauma Surgery: Is ChatGPT-4 a Useful Supplement to Physician-Led Informed Consent?

Status

Active, not recruiting

Phases

Unknown

Study type

Interventional

Source

DRKS

Registry ID

DRKS00039898

Enrollment

124

Registered

2026-04-08

Start date

2025-05-01

Completion date

Unknown

Last updated

2026-08-03

For informational purposes only — not medical advice. Sourced from public registries and may not reflect the latest updates. Terms

Conditions

S42.0 S52.5 S42.2 S82.5 S82.6 S92.0 S82.1 S52.0

Interventions

Group 1: Participants in the control group undergo the standard preoperative informed consent process. After providing written study consent, participants complete a pre-intervention questionnaire ass

German adaptation: ZUF-8). Group 2: Participants in the intervention group receive an additional LLM-assisted information session prior to the standard physician-led informed consent consultation. Aft

Eligibility

Sex/Gender

All

Age

18 Years to No maximum

Inclusion criteria

Inclusion criteria: (1) Age =18 years. (2) Indication for elective orthopaedic trauma surgery. (3) Ability to participate in an informed consent consultation and to complete a questionnaire. (4) Written informed consent for study participation.

Exclusion criteria

Exclusion criteria: (1) Emergency indication. (2) Cognitive impairment precluding study participation (e.g., advanced dementia, severe psychiatric disorders, significant cognitive limitations). (3) Insufficient German language proficiency to adequately understand the content of the informed consent process and questionnaires. (4) Refusal to participate in the study.

Design outcomes

Primary

Measure	Time frame
Post-consultation perceived understanding (composite score), operationalized as the mean of four self-rated 10-point items assessing understanding of the diagnosis, planned procedure, risks and complications, and postoperative care. Baseline ratings of the same four domains are obtained before the consent process to enable adjusted between-group analyses and pre–post comparisons. The primary analysis is performed using analysis of covariance (ANCOVA) with the post-consultation composite score as the dependent variable, study group as the fixed factor, and baseline score, age, and prior AI exposure as covariates. Assessment timepoint: immediately after the physician-led informed consent consultation. Method: self-reported questionnaire	—

Measure

Time frame

Post-consultation perceived understanding (composite score), operationalized as the mean of four self-rated 10-point items assessing understanding of the diagnosis, planned procedure, risks and complications, and postoperative care. Baseline ratings of the same four domains are obtained before the consent process to enable adjusted between-group analyses and pre–post comparisons. The primary analysis is performed using analysis of covariance (ANCOVA) with the post-consultation composite score as the dependent variable, study group as the fixed factor, and baseline score, age, and prior AI exposure as covariates. Assessment timepoint: immediately after the physician-led informed consent consultation. Method: self-reported questionnaire

—

Secondary

Measure	Time frame
(1) Overall comprehensibility of the informed consent process, rated on a 10-point scale. (2) Overall satisfaction with the informed consent process, rated on a 10-point scale. (3) Informed decision-making – participants' self-assessment of whether they feel sufficiently informed to make a well-founded decision regarding the planned procedure, rated on a 10-point scale. (4) Trust in physician-led informed consent alone, rated on a 10-point scale. (5) Trust in AI-only informed consent, rated on a 10-point scale. (6) Trust in medically supervised AI-assisted informed consent, rated on a 10-point scale. (7) Concerns regarding AI integration into the informed consent process, assessed using multiple-choice predefined categories (including lack of personal conversation/human contact, potential errors/misunderstandings, lack of empathy, insufficient explanation of complex issues, data privacy/security concerns). (8) Global patient satisfaction, measured using the validated CSQ-8 (German adaptation: ZUF-8). All secondary endpoints are assessed immediately after the consultation via self-reported questionnaire. Secondary analyses are exploratory; no adjustment for multiple testing is applied. (1) Physician-rated evaluation of ChatGPT-generated patient information documents in the intervention group, assessed across five dimensions: factual accuracy, completeness, linguistic clarity/comprehensibility, and empathy (each rated on 4-point ordinal scales), as well as hallucination occurrence (yes/no). Ratings are assigned independently by two board-certified trauma surgeons blinded to each other's assessments and to participant outcome data. (2) Descriptive analysis of patient–LLM question behaviour, including categorization of submitted questions by topic using a predefined content framework.	—

Measure

Time frame

(1) Overall comprehensibility of the informed consent process, rated on a 10-point scale. (2) Overall satisfaction with the informed consent process, rated on a 10-point scale. (3) Informed decision-making – participants' self-assessment of whether they feel sufficiently informed to make a well-founded decision regarding the planned procedure, rated on a 10-point scale. (4) Trust in physician-led informed consent alone, rated on a 10-point scale. (5) Trust in AI-only informed consent, rated on a 10-point scale. (6) Trust in medically supervised AI-assisted informed consent, rated on a 10-point scale. (7) Concerns regarding AI integration into the informed consent process, assessed using multiple-choice predefined categories (including lack of personal conversation/human contact, potential errors/misunderstandings, lack of empathy, insufficient explanation of complex issues, data privacy/security concerns). (8) Global patient satisfaction, measured using the validated CSQ-8 (German adaptation: ZUF-8). All secondary endpoints are assessed immediately after the consultation via self-reported questionnaire. Secondary analyses are exploratory; no adjustment for multiple testing is applied. (1) Physician-rated evaluation of ChatGPT-generated patient information documents in the intervention group, assessed across five dimensions: factual accuracy, completeness, linguistic clarity/comprehensibility, and empathy (each rated on 4-point ordinal scales), as well as hallucination occurrence (yes/no). Ratings are assigned independently by two board-certified trauma surgeons blinded to each other's assessments and to participant outcome data. (2) Descriptive analysis of patient–LLM question behaviour, including categorization of submitted questions by topic using a predefined content framework.

—

Countries

Germany

Contacts

Public ContactPhilipp Zehnder

Technische Universität München, Klinikum rechts der Isar, Klinik und Poliklinik für Unfallchirurgie

philipp.zehnder@mri.tum.de08941408361

Outcome results

None listed

Source: DRKS (via WHO ICTRP) · Data processed: Aug 10, 2026