Outpatient Care
Conditions
Brief summary
The goal of this clinical trial is to learn how the use of a large language model (LLM) based tool affects outpatient clinical care in adult patients attending general hospital outpatient clinics. The main questions it aims to answer are: Does the use of an LLM-based tool affect the efficiency of outpatient visits? Does the use of an LLM-based tool affect the experience of doctors and patients during outpatient care? Researchers will compare outpatient visits supported by an LLM-based tool to standard outpatient visits without such a tool, to see whether and how the tool influences the care process and the experiences of doctors and patients. Participants will: Take part in outpatient visits that may or may not involve an LLM-based tool, depending on their assigned group Complete a short questionnaire about their visit experience after the consultation
Interventions
OTHERLarge Language Model Based Tool
A large language model based tool is introduced into the outpatient consultation workflow to support the consultation and documentation process.
OTHERWorkflow Support for Large Language Model Tool Integration
Additional workflow support is provided to integrate the output of the large language model based tool into the consultation process, approximating a more integrated deployment of the tool.
Sponsors
Tsinghua University
Study design
Allocation
RANDOMIZED
Intervention model
CROSSOVER
Primary purpose
HEALTH_SERVICES_RESEARCH
Masking
SINGLE (Outcomes Assessor)
Intervention model description
This is a cluster-randomized crossover trial in which the unit of randomization is the doctor-session (i.e., an individual outpatient doctor on a given clinic day). Each participating doctor is assigned, in randomized order, to each of the three study conditions across multiple clinic sessions, so that every doctor contributes sessions to all three groups (crossover at the doctor level). Patients are nested within doctor-sessions. Each patient is allocated to a single study group, determined by the group assignment of the doctor-session in which their visit occurs, and therefore receives only one condition (no crossover at the patient level). At the time of registering for a clinic appointment, patients are not aware of which group their chosen doctor's session has been assigned to; this becomes known only after registration.
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
Doctors: 1. Licensed physicians providing outpatient consultations at a participating study hospital 2. Expected to complete a sufficient number of outpatient clinic sessions during the study period 3. Provides written informed consent Patients: 1. Age 18 years or older 2. Attending an outpatient consultation with a participating doctor 3. Able to interact with the tool using an internet-connected device such as a smartphone 4. Provides written informed consent
Exclusion criteria
Patients: 1. Psychiatric conditions, unstable vital signs, or other medical situations considered unsuitable for AI-based interaction 2. Declines to provide informed consent
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Duration of the Outpatient Consultation | During the outpatient visit | Time of the outpatient consultation, measured in milliseconds |
| Doctor-Reported Efficiency of the Consultation | Immediately after the consultation | Doctor's self-rated efficiency of the consultation, measured on a 5-point Likert scale (1 = very low to 5 = very high), with higher scores indicating higher perceived efficiency. |
| Doctor-Reported Satisfaction With the Consultation Process | Immediately after the consultation | Doctor's satisfaction with the consultation process, measured on a 5-point Likert scale (1 = very low to 5 = very high), with higher scores indicating higher satisfaction. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Doctor-Reported Efficiency of Obtaining Patient Information | Immediately after the consultation | Doctor's self-rated efficiency in obtaining the patient's clinical information (such as symptoms, history, prior examinations) during the consultation, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate higher efficiency. |
| Doctor-Reported Cognitive Effort in Clinical Decision-Making | Immediately after the consultation | Doctor's self-rated cognitive effort invested in clinical decision-making during the consultation, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate greater effort. |
| Doctor-Reported Burden of Clinical Documentation | Immediately after the consultation | Doctor's self-rated burden of completing the outpatient medical record for the consultation, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate greater burden. |
| Doctor's Intention to Continue Using the Tool | Within 1 week after the participating doctor completes all enrolled consultations | Doctor's intention to continue using the large language model based tool in routine practice, measured on a 5-point Likert scale (1 = strongly unwilling to 5 = strongly willing); higher scores indicate stronger intention. |
| Patient Trust in the Physician | Immediately after the consultation | Patient's level of trust in the physician after the visit, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate greater trust. |
| Patient Satisfaction With the Visit | Immediately after the consultation | Patient's satisfaction with the visit, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate greater satisfaction. |
| Patient-Perceived Physician Attentiveness | Immediately after the consultation | Patient-perceived attentiveness of the physician during the visit, assessed by a multi-item measure and reported as a composite score on a 1-5 scale; higher scores indicate greater perceived attentiveness. |
| Patient Satisfaction With the AI Pre-Consultation (Arm 2 and Arm 3 ) | Immediately after the consultation | Patient's satisfaction with the AI-based pre-consultation interaction, measured on a 5-point Likert scale (1 = very low to 5 = very high); higher scores indicate greater satisfaction. Assessed only in Arm 2 and Arm 3. |
| Patient's Intention to Use AI Pre-Consultation in the Future (Arm 2 and Arm 3) | Immediately after the consultation | Patient's intention to use AI-based pre-consultation again in the future, measured on a 5-point Likert scale (1 = strongly unwilling to 5 = strongly willing); higher scores indicate stronger intention. Assessed only in Arm 2 and Arm 3. |
Outcome results
None listed