The Effects of an AI created Question Prompt List on Preoperative Anxiety and Patient Satisfaction with Physician Informed Consent: A Pilot Study

Introduction

The anxiety of patients within and related to the United States healthcare system is a well studied and yet poorly addressed reality. Anxiety related to cost, access, the fear of our gross and again bodies, represent the fears of many patients. Patient fears related to procedural and surgical interventions are similarly plentiful. The fear of anesthesia, pain, potential complications, and even the procedure itself have been well documented for decades.

Preoperative patient anxiety is frequently addressed through administration of an anxiolytic medication, or simple distraction, but neither of these interventions attempt to address the source of a patient’s anxiety. There is evidence that increasing the amount of preoperative information given to patients can aid in reducing overall anxiety, but there is also evidence that more in depth consultations do not significantly decrease perioperative anxiety.

Patient’s oftentimes have difficulty raising questions, or knowing which questions to ask, even when prompted from healthcare providers. More active patient participation contributes to improved health outcomes and quality of care, so it is important to understand the factors affecting a patients ability to communicate with healthcare providers.

Question prompt lists (QPLs) have been proven to be effective, particularly in the oncological setting, at increasing patient-physician communication and patient engagement due to helping a patient effectively address their concerns with a health care worker.

A question prompt list is a list of questions, varying in length depending on the context, designed with the assistance of patients and healthcare workers who specialize in a particular field being discussed. QPLs have proven to be effective at increasing patient engagement during initial consultations after new cancer diagnoses, resulting in better patient outcomes and satisfaction.

Although the benefits of QPLs are well researched in multiple areas of healthcare, fewer articles explore the use of QPLs in surgical consultations. QPLs that were mailed to patients prior to outpatient surgical consultations proved to be ineffective in improving patient outcomes, but prior QPL research in the oncological setting states multiple benefits when it is given to the patient during their interaction with a physician. The effects on having a physician provide a QPL on the day of surgery during an informed consent have yet to be thoroughly explored. One barrier to the implementation of QPLs at the bedside for procedural and surgical consents could be the perception that patients do not need them. Although health care workers do their best to prompt questions from patients, typically methods of doing so have proven to be ineffective. Standards methods of prompting questions, such as asking if the patient has any question, yield significantly fewer questions than the presence of a QPL. Patients with lower levels of literacy also frequently ask fewer questions, which is not frequently assessed prior to procedures.

The creation of a QPL for a large number of surgical procedures and interventions may also pose a barrier to the creation and implementation of QPLs in a perioperative setting. Although QPLs are typically created through the consultation of patients and healthcare professionals, the use and effectiveness of LLM artificial intelligence, such as chatGPT, as an aid in creating a QPL has yet to be explored.

OpenAI’s chatGPT, specifically the gpt-4o model, is a large language model machine learning service with a large focus on natural language processing. Since the explosion of chatGPT many fields have experimented with its effectiveness as a teaching tool. In healthcare specifically, researchers have looked into the software’s ability to assist in analyzing patient medical histories and interpreting big data to better assist in patient care and patient outcomes. We have yet to see research on the implementation of AI as a communication aid between patients and health care providers.

Our team has developed an application to assist in this study that does just that. The application takes human speech, an informed consent for surgery in our case, analyzes the content, and procedurally generates a question prompt list based on the content of the informed consent. This application could hypothetically be used to create a QPL for any verbally conveyed information in a healthcare setting, such as hospital discharge instructions or new medication teaching, but in this study we will solely focus on surgical and procedural informed consent.

Present Study

In the present study, we will use quantitative methods to examine whether the utilization of our application, iQPL, results in increased question asking compared to our control group, and whether these patient's experienced a significant reduction in perioperative anxiety as a result.

We hypothesize this AI powered application will effectively create a list of 20 common, relevant questions based on the given procedural consent; the presence of which will result in an overall increase in questions patients ask their consenting physicians.

Much like non-AI QPLs, we hypothesize this increased engagement and communication will result in higher levels of patient satisfaction and lower levels of pre-operative anxiety compared to patients who do not use the AI assisted QPL.

The Application

If you would like to know a little bit about how the application works, keep reading. Otherwise, skip to the next section.

iQPL is a web application designed with React and NodeJS. iQPL levererages the OpenAI gpt-4o API for audio analysis and chat functionality. By delivering a finely tuned prompt to the gpt-4o API, we are able to get a list of 20 questions that the average person could ask after hearing the given audio input. The audio input in this case is a procedural consent, but it could hypothetically be any sort of audio input.

If the application functions well during this experiment, we could hypothetically use it to help patients ask questions during any sort of interaction, from new medication teaching to discussing test results.

The application is extremely simple to use. Before the consent is given, you click a large 'Start Recording' button on the home screen.

After the consent, you click the 'Finish Recording' button. The application will now start building the QPL, this can take up to 10 seconds.

The QPL appears on the screen in the patient's preferred language. Selecting one of the choices will have the question read out loud. Or the patient can merely peruse the list and ask any questions they might have.