Effectiveness of a Large Language Model-Based Educational Tool on Intraocular Lens Options

Stanford University

Status: Invitation-only

Conditions: Cataract Surgery Experience; Cataract Extraction; Eye Disorders; Intraocular Lens; Cataract Surgery; Cataract and IOL Surgery

Treatments: Other: LLM-based Education

Study type: Interventional

Funder types: Other

Identifiers

NCT07317661

82807

Details and patient eligibility

About

Patients with cataracts disease need to choose what type of artificial lens will go into their eye prior to surgery date. Some lenses are standard and are usually covered by insurance. Other "premium" lenses have various benefits such as reducing the need for glasses but usually require out-of-pocket costs.

The combined busy outpatient clinic and complexity of artificial lens choices in the ever-changing world of cataract surgery tends to lead patients confused about their available lens options. There is an abundance of educational material present in premium lenses, however these are limited by accessibility and are standardized at single educational levels.

Therefore in the present study, we want to test whether giving patients a short LLM powered AI-guided explanation from Custom GPT from OpenAI of lens options prior to their consultation with their doctor can improve visit efficiency, physician explanation and patient understanding of lens options. We will compare two groups: standard of care versus standard of care plus AI education.

The LLM in this study is intended to provide supplemental information about premium intraocular lens(IOLs) options to study participants, and is no means supposed to replace a health care professional in the diagnosis, cure, treatment, and/or mitigation of disease. Study is analogous to giving a verified health pamphlet to a patient for them to view and learn different IOL options, in other words, facilitating patient understanding of their options.

The LLM will be trained by several health care professionals and MD specialists to provide sufficient instructions. Sources will include verified online resources and MD information.

The investigators hope to learn if a large language model-based educational tool can improve visit efficiency, physician explanation and patient understanding of intraocular lens options. New knowledge of this study could guide how cataract counseling is delivered in the future and may help clinics spend more time on individualized questions instead of repeating generic information.

Enrollment: 70 estimated patients

Sex: All

Ages: 18+ years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• Age 18 or older
• Presenting for cataract evaluation or preoperative cataract counseling in the ophthalmology clinic
• Able to provide informed consent
• English-speaking
• No prior cataract surgery in either eye (so that all patients are making a first-eye IOL decision)

Exclusion criteria

• Any cognitive impairment or hearing impairment that prevents meaningful counseling or survey completion
• Urgent ocular condition requiring immediate attention that would override routine cataract counseling (for example, acute retinal detachment)
• Patient declines or is unable to complete the brief post-visit survey
• Has ocular conditions that would impact eligibility of non-monofocal lens options

Trial design

Primary purpose: Treatment

Allocation: Randomized

Interventional model: Parallel Assignment

Masking: Single Blind

---

70 participants in 2 patient groups

LLM-based Education + Standard of Care

Experimental group

Description:

* Before seeing the fellow, the participant will listen to a short, structured LLM powered AI-delivered educational session with Custom GPT (10 minutes or less). The intractable AI script explains standard monofocal IOLs and premium options (toric, extended depth of focus, multifocal, light adjustable lens), including benefits, trade-offs, and out-of-pocket costs. * The AI module may allow the patient to ask clarifying questions within scope of that script. This AI session is not currently part of standard care and is considered the experimental intervention. * The participant takes a patient satisfaction (CSQ-8) after their clinical visit with the fellow and attending

Treatment:

Other: LLM-based Education

Standard of Care

No Intervention group

Description:

* The participant skips the AI module and proceeds directly to routine fellow and attending counseling, which reflects current standard of care practice. * The participant takes a patient satisfaction (CSQ-8) after their clinical visit with the fellow and attending physician