Assessing ChatGPT Responses to Commonly Asked Questions About Hypoglossal Nerve Stimulation

ABSTRACT

Objective

This study aimed to explore the accuracy and comprehensiveness of 10 frequently asked questions posed to ChatGPT, an online chatbot, regarding obstructive sleep apnea (OSA) and hypoglossal nerve stimulation (HNS) therapy.

Methods

Ten questions were formulated after extensive literature review alongside the guidance of the senior author (R.N.), a board‐certified otolaryngologist specializing in HNS therapy. We employed ChatGPT version 3.5 to answer each question, limiting responses to one paragraph. No follow‐up queries were posed. The precision and thoroughness of responses were graded via the grading scale proposed by Mika et al. (within the context of hip arthroplasty) and the DISCERN criteria. Each response was graded by three independent, blinded reviewers (L.L., D.M., and R.N.).

Results

On average, the responses produced by ChatGPT were satisfactory but typically required moderate clarification (inter‐rater reliability [IRR] 0.533). Of the 10 responses, only one response was deemed excellent not requiring clarification. Two of the responses were deemed satisfactory requiring minimal clarification and the remaining seven responses were all deemed satisfactory requiring moderate clarification. No response was deemed unsatisfactory requiring substantial clarification. The average DISCERN score was 41.0 (IRR = 0.513), suggesting that the written health information was “average.”

Conclusions

The responses produced by ChatGPT were fairly accurate, but often benefited from further clarification to ensure a comprehensive understanding. Ultimately, it seems that ChatGPT currently functions as a powerful resource to aid patients in further understanding topics related to OSA and HNS therapy, but should not replace direct consultation with knowledgeable healthcare providers.

Summary

• On average, ChatGPT provides adequate responses to questions regarding hypoglossal nerve stimulation; however, the answers would benefit from further clarification to capture nuanced details.

• This study adds to the current literature by providing a systematic evaluation of ChatGPT's accuracy and readability in addressing patient inquiries about OSA and hypoglossal nerve stimulation, highlighting both its utility and limitations as a patient education tool.

1. Introduction

Artificial intelligence (AI) is becoming increasingly integral to healthcare and technology sectors, enhancing diagnostics and decision‐making. One prominent example is ChatGPT, developed by OpenAI, an online chatbot designed to respond to various user inquiries [1]. Now, with over 100 million users, this platform is popular for its potential application in various fields, including health care.

In today's digital age, the vast availability of medical information paves the way for resources like ChatGPT to become valuable tools for acquiring medical knowledge. Recent studies have demonstrated ChatGPT's utility in answering an array of otolaryngology‐related medical issues [2, 3, 4, 5], pertinent as patients increasingly turn to online resources for medical guidance [6, 7]. At the same time, it is important to understand the extent and limitations of ChatGPT's capabilities within the context of health‐related patient queries [8]. Commonly encountered disorders like obstructive sleep apnea (OSA), estimated to affect 22% of men and 17% of women [9], may be a common topic for the chatbot due to its widespread prevalence. Similarly, innovative treatment options, like the hypoglossal nerve stimulator (HNS) developed by Inspire Medical Systems (Minneapolis, MN), are becoming increasingly relevant, with patients potentially turning to ChatGPT for a better understanding of its role in OSA management. With the growing influence of AI, understanding ChatGPT's success in accurately and comprehensively responding to health‐related inquiries becomes paramount. In this study, we assess ChatGPT's ability to provide accurate and comprehensive answers to 10 common questions related to OSA and HNS therapy by scoring each response using three (L.L., D.M., and R.N.) independent, blinded reviewers. We predict that ChatGPT will be able to address these inquiries with notable accuracy, comprehensiveness, and clarity.

2. Methods

The methodology of this cross‐sectional analysis is similar to Mika et al.'s study on total hip arthroplasty [10]. We reviewed websites of 10 reputable healthcare institutions, including Johns Hopkins, Mayo Clinic, Harvard Medical School, and Yale Medicine, along with the Inspire Medical Systems Frequently Asked Questions page, to identify common patient questions about HNS implantation. Guided by the senior author (R.N.), a board‐certified otolaryngologist specializing in HNS, we formulated the 10 most frequently encountered questions from our literature review on HNS therapy.

On March 8, 2024, we submitted the questions to the ChatGPT 3.5 online interface at https://chat.openai.com/. This model was chosen for its accessibility, popularity, and cost‐free availability to patients. Each query was presented with the prompt: “I am a patient. Provide a one‐paragraph response to the following question and make sure it includes all pertinent information,” followed by the question. ChatGPT's responses were recorded after the initial queries without repetition or follow‐up questions.

Following Mika et al.'s methodology [10], each response was classified into one of four categories: “excellent response not requiring classification,” “satisfactory response requiring minimal clarification,” “satisfactory response requiring moderate clarification,” and “unsatisfactory response requiring substantial clarification,” numbered 1 through 4. “Excellent” responses were succinct, comprehensive, and accurate. “Satisfactory” responses needing minimal clarification were accurate but missed some nuances. Those needing moderate clarification had valuable insights but included outdated or irrelevant details. “Unsatisfactory” responses were imprecise or overly general, requiring significant clarification to avoid misconceptions.

Three authors (L.L., D.M., and R.N.) independently and blindly evaluated ChatGPT's responses, using Mika et al.'s grading scale [10]. Additionally, responses to each of the 10 questions were assessed using the DISCERN instrument, which classifies written health information into five levels: excellent (63–80), good (51–62), average (39–50), poor (28–38), and very poor (27 or less) [11]. Inter‐rater reliability (IRR) was determined using intraclass correlation coefficients for both the Mika et al. scale and the DISCERN instrument to gauge evaluator consistency.

To assess the clarity and readability of ChatGPT's responses, we used the Flesch‐Kincaid Grade Level (FKGL), SMOG index, Coleman–Liau index, Gunning Fog index, and Automated Readability Index. Readability scores were generated using the wordcalc calculator (https://www.wordcalc.com/readability/), after converting each ChatGPT response into plaintext.

3. Results

The mean score of ChatGPT's responses using Mika et al.'s grading scale was 3 (IRR = 0.533) [10], indicating “satisfactory requiring moderate clarification.” The mean DISCERN score was 41.0 (IRR = 0.513), rated as “average.” Assessments for each question and response are shown in Table 1. The average readability score across five metrics was 13.4 years of education, akin to a first‐year college student (Table 2). ChatGPT's responses are in Appendix 1. Similarities were noted among DISCERN evaluations, often scoring low on providing details of additional sources. While ChatGPT typically omits references, it can provide them if asked. Additionally, 70% of responses advised consulting healthcare providers or emphasized regular follow‐ups, suggesting that ChatGPT acknowledges the importance of direct consultation with healthcare professionals. Likewise, including this within the response may further encourage patients to seek care from a healthcare provider rather than rely solely on online information.

Table 1.

Mean scores of responses produced by ChatGPT.

Question	Mean Mika et al. score	Mean DISCERN score
What is obstructive sleep apnea?	1	48.7
What are the risks of untreated obstructive sleep apnea?	2	41.7
What treatment options exist for obstructive sleep apnea?	3	46
How does the hypoglossal nerve stimulator work?	2	44.7
What determines if I can get a hypoglossal nerve stimulator to treat my obstructive sleep apnea?	3	38.7
What does surgery look like for hypoglossal nerve stimulator implantation?	3	34.3
What are the possible risks of hypoglossal nerve stimulator implantation?	3	40.7
What does recovery following hypoglossal nerve stimulator implantation look like?	3	39.7
What restrictions will I have with the hypoglossal nerve stimulator implanted to treat my sleep apnea?	3	36.7
Why would I have to get my hypoglossal nerve stimulator implant replaced?	3	38.7
Averaged Means	3	41.0

Note: A Mika et al. score of 1 corresponds to “excellent response not requiring clarification,” a score of 2 corresponds to “satisfactory response requiring minimal clarification,” a score of 3 corresponds to “satisfactory response requiring moderate clarification,” and a score of 4 corresponds to an “unsatisfactory response requiring substantial clarification.” All means were rounded to the nearest whole number.

Table 2.

Individual readability score of the responses produced by ChatGPT.

Test	Score
FKGL	12
SMOG	12
Coleman–Liau Index	12
Gunning Fog Index	19
Automated Readability Index	12

Abbreviations: FKGL, Flesh‐Kincaid Grade Level; SMOG, Simple Measure of Gobbledygook index.

4. Discussion

As AI becomes integrated into society, more patients may use AI for medical inquiries. Therefore, understanding AI's capabilities and limitations is crucial. By paralleling Mika et al.'s study on hip arthroplasty [10], we examined ChatGPT's accuracy and comprehensiveness in answering questions about OSA and HNS therapy.

Our findings indicate that ChatGPT generally provided accurate and detailed responses, albeit often requiring minor to moderate clarification. With an average DISCERN score of 48, ChatGPT's responses suggest an average quality. ChatGPT appears valuable for basic patient education on OSA and HNS therapy, although its information may not be exhaustive or tailored to individual needs, often requiring further evidence‐based data for comprehensive understanding. While not a substitute for professional medical advice, ChatGPT can complement patient education for OSA and HNS therapy. To support these claims, our analyses compare ChatGPT responses with literature on OSA and HNS implantation.

4.1. Comprehensive Analysis

Multiple studies have demonstrated ChatGPT's efficacy in addressing patient inquiries in otolaryngology [2, 3, 4], corroborating our findings. While ChatGPT is relatively accurate and thorough, it often demands a high reading level for full comprehension. However, inconsistencies in evaluating ChatGPT's responses exist, with varying measures like the Flesch Reading Ease Score, Gunning–Fog Index, and DISCERN questionnaire. Developing a validated framework incorporating these metrics would enable more consistent assessment [4, 12, 13]. Lechien et al. introduced and validated the Artificial Intelligence Performance Instrument (AIPI) to evaluate ChatGPT's clinical patient management, but it focuses on diagnostic and management abilities rather than overall response accuracy and comprehensibility [14].

This study found ChatGPT's responses typically require a college‐level reading comprehension, exceeding recommended levels for patient educational content. Similar findings in other studies demonstrate ChatGPT's default to higher reading levels [4, 12]. Authorities like the National Institute of Health, the US Department of Health and Human Services, and the American Medical Association recommend materials at a sixth to seventh‐grade reading level to improve health literacy and address disparities [15]. Therefore, while ChatGPT offers valuable insights, its complexity highlights the need to tailor content to meet established health literacy standards, ensuring accessibility and equity in patient education. Finally, a categorized analysis of each question posed to ChatGPT can be found in Appendix 2.

4.2. Limitations

Our study is not without limitations. We analyzed only one chatbot and did not ask for further clarifications, unlike real‐world patient interactions. Furthermore, ChatGPT's responses are sensitive to query phrasing, with subtle variations producing different answers [16]. Each query to ChatGPT, despite even when identical prompts are used, may introduce variability in both content and phrasing. This adaptability may provide a more personalized response but may also lead to inconsistencies in what information is provided. Additionally, we used the no‐cost version (3.5) of ChatGPT, assuming most patients would avoid the premium 4.0 version. Future studies should compare the accuracy and comprehension of responses between versions 3.5 and 4.0. Finally, the rapid evolution of AI presents a fundamental challenge in evaluating models like ChatGPT. By the time this study is published, new iterations of language models may introduce substantial changes, necessitating further examination of similar findings. While our analysis provides a thorough assessment of ChatGPT's responses at the time of data collection, the specific gaps and areas requiring clarification may shift as the model continues to evolve. This underscores the importance of ongoing assessment of AI‐generated medical information to ensure its accuracy, reliability, and relevance in patient education.

5. Conclusion

AI platforms like ChatGPT are transforming patient access to health information. While promising, understanding AI's limitations is crucial. Our study shows that ChatGPT provides fairly accurate and comprehensive information but cannot replace direct consultations with healthcare providers. Instead, it should complement patient education. Further research is needed to fully understand AI's potential and limitations in healthcare.

Author Contributions

Mihai A. Bentan: conception, data collection, data analysis, manuscript composition, proofing. Lawrance Lee, David Moffatt, and Ryan Nord: data collection, data analysis, manuscript composition, proofing.

Disclosure

The authors have nothing to report.

Ethics Statement

There are no applicable ethical considerations given that the performed research did not involve human participants, human data, or human tissue in this study.

Conflicts of Interest

Ryan Nord is a consultant for Inspire Medical Systems, Nyxoah, and SIESTA medical. The authors declare no conflicts of interest.

Appendix 1. Responses Produced by ChatGPT

1.1. Questions

Question 1: “What is obstructive sleep apnea?”

ChatGPT's response: “Obstructive sleep apnea (OSA) is a common sleep disorder characterized by repeated interruptions in breathing during sleep due to the collapse or blockage of the upper airway. These interruptions, known as apneas, can lead to decreased oxygen levels in the blood and fragmented sleep patterns. Symptoms of OSA include loud snoring, gasping or choking during sleep, excessive daytime sleepiness, morning headaches, and difficulty concentrating. Risk factors for OSA include obesity, older age, male gender, family history, and certain anatomical factors such as a thick neck or narrow airway. Untreated OSA can have serious health consequences, including hypertension, heart disease, stroke, and impaired cognitive function, underscoring the importance of timely diagnosis and management by a healthcare professional. Treatment options range from lifestyle modifications, such as weight loss and positional therapy, to medical interventions like continuous positive airway pressure (CPAP) therapy or surgery, tailored to the individual's needs and severity of the condition. Regular follow‐up and compliance with treatment are crucial for effectively managing OSA and improving overall health and quality of life.”

Question 2: “What are the risks of untreated obstructive sleep apnea?”

ChatGPT's response: “Untreated obstructive sleep apnea poses significant risks to both physical and mental health. Physically, it can lead to hypertension, heart disease, stroke, type 2 diabetes, and increased risk of accidents due to daytime sleepiness. Mental health can also be affected, with untreated sleep apnea linked to depression, anxiety, and decreased cognitive function. Additionally, untreated sleep apnea can strain relationships due to loud snoring and disrupted sleep patterns. Seeking treatment is crucial to mitigate these risks and improve overall quality of life.”

Question 3: “What treatment options exist for obstructive sleep apnea?”

ChatGPT's response: “For obstructive sleep apnea (OSA), treatment options vary depending on the severity of the condition and individual health factors. Continuous Positive Airway Pressure (CPAP) therapy is a common approach, which involves wearing a mask that delivers a steady stream of air to keep the airways open during sleep. Other treatments include dental devices that adjust the position of the jaw to prevent airway blockage, surgery to remove excess tissue blocking the airway, and lifestyle changes such as weight loss, avoiding alcohol and sedatives before bed, and altering sleep positions. It's crucial for patients to consult with healthcare professionals to determine the most suitable treatment plan tailored to their specific needs and health status.”

Question 4: “How does the hypoglossal nerve stimulator work?”

ChatGPT's response: “The hypoglossal nerve stimulator, also known as Inspire therapy, is a treatment option for obstructive sleep apnea (OSA) in patients who are unable to tolerate continuous positive airway pressure (CPAP). It works by delivering mild stimulation to the hypoglossal nerve during sleep, which controls the movement of the tongue and other throat muscles. This stimulation helps to keep the airway open by preventing the collapse of soft tissues in the throat, thus reducing the frequency of breathing pauses characteristic of sleep apnea. The device consists of a small pulse generator implanted under the skin in the upper chest area, along with a sensing lead positioned in the chest to detect breathing patterns and a stimulation lead placed around the hypoglossal nerve. During sleep, the device automatically senses breathing patterns and delivers stimulation to the hypoglossal nerve, thereby promoting airway patency and improving breathing efficiency without the need for a mask or continuous airflow. It is important for patients considering this therapy to undergo thorough evaluation and consultation with a sleep specialist to determine if they are suitable candidates and to understand the potential benefits and risks associated with the procedure.”

Question 5: “What determines if I can get a hypoglossal nerve stimulator to treat my obstructive sleep apnea?”

ChatGPT's response: “Whether you're eligible for a hypoglossal nerve stimulator to treat obstructive sleep apnea (OSA) typically depends on several factors including your specific medical history, severity of your OSA, previous treatment attempts, and anatomy of your upper airway. Generally, candidates for this treatment have moderate to severe OSA, have not found relief from other therapies such as continuous positive airway pressure (CPAP), and have anatomical features that make them suitable for the device. An evaluation by a sleep medicine specialist or otolaryngologist will assess these criteria to determine if a hypoglossal nerve stimulator is an appropriate option for managing your OSA.”

Question 6: “What does surgery look like for hypoglossal nerve stimulator implantation?”

ChatGPT's response: “Surgery for hypoglossal nerve stimulator implantation typically involves general anesthesia and positioning the patient on their back. The surgeon makes an incision in the neck to access the hypoglossal nerve and then implants a small device beneath the skin near the collarbone, which connects to a lead that is carefully placed around the hypoglossal nerve. This device delivers mild electrical impulses to the nerve during sleep, helping to keep the airway open and alleviate obstructive sleep apnea symptoms. The procedure generally takes a few hours, and patients typically stay in the hospital overnight for monitoring before being discharged with instructions for postoperative care and device usage.”

Question 7: “What are the possible risks of hypoglossal nerve stimulator implantation?”

ChatGPT's response: “Hypoglossal nerve stimulator implantation carries several potential risks, including but not limited to infection at the surgical site, adverse reactions to anesthesia, bleeding, and damage to surrounding structures such as blood vessels or nerves. Additionally, there may be discomfort or pain following the procedure, difficulty swallowing or speaking, temporary or permanent weakness or paralysis of the tongue, and potential for device malfunction or displacement necessitating further intervention. It's crucial for patients to thoroughly discuss these risks with their healthcare provider and weigh them against the potential benefits of the procedure before making an informed decision.”

Question 8: What does recovery following hypoglossal nerve stimulator implantation look like?

ChatGPT's response: “Recovery following hypoglossal nerve stimulator (HNS) implantation typically involves several stages. Initially, patients may experience mild discomfort, swelling, and bruising around the surgical site, which generally subsides within a few weeks. Patients are advised to follow post‐operative care instructions provided by their healthcare team, which may include pain management strategies and dietary modifications. Over the following weeks and months, individuals typically undergo a gradual adjustment period as they adapt to the device, during which improvements in sleep quality and reduction in sleep apnea symptoms are observed. Regular follow‐up appointments with healthcare providers are essential to monitor progress, optimize device settings, and address any concerns or complications that may arise. It's important for patients to maintain open communication with their healthcare team throughout the recovery process to ensure optimal outcomes.”

Question 9: What restrictions will I have with the hypoglossal nerve stimulator implanted to treat my sleep apnea?

ChatGPT's response: “After the implantation of a hypoglossal nerve stimulator to manage your sleep apnea, certain restrictions will apply to ensure optimal healing and effectiveness of the device. Initially, you may be advised to refrain from strenuous activities and heavy lifting for a specified period to prevent strain on the surgical site. It's important to follow your healthcare provider's guidance regarding post‐operative care and gradually resume normal activities as advised. Additionally, you may need to avoid certain electromagnetic devices, such as MRI machines, which could interfere with the functioning of the implant. It's crucial to maintain regular follow‐up appointments with your healthcare team to monitor your progress and address any concerns that may arise during your recovery and ongoing management of sleep apnea with the hypoglossal nerve stimulator.”

Question 10: Why would I have to get my hypoglossal nerve stimulator implant replaced?

ChatGPT's response: “There are several reasons why you might need to have your hypoglossal nerve stimulator implant replaced. Over time, the device may experience mechanical failure or wear and tear, leading to decreased effectiveness or complete malfunction. Additionally, changes in your health status, such as weight fluctuations or shifts in anatomy, could necessitate adjustments or replacements to ensure optimal functioning of the device. Furthermore, advancements in technology might prompt the replacement of older models with newer ones offering improved features or capabilities. Regular follow‐up appointments with your healthcare provider are essential to monitor the performance of the implant and determine if replacement or adjustments are necessary to maintain its efficacy in treating your sleep apnea.”

Appendix 2. Categorized Analyses on Questions

Obstructive Sleep Apnea

Question 1: “What is obstructive sleep apnea?”

GRADE: Excellent Response Not Requiring Clarification

The chatbot provided a concise, comprehensive overview of OSA, addressing key aspects such as risk factors, treatment options, and the implications of untreated OSA all via patient‐friendly language. Estimated to affect 22% of men and 17% of women, OSA involves recurrent upper airway collapse leading to nocturnal oxygen desaturation and sleep disruption [9]. ChatGPT's response correctly identified risk factors like age, gender, obesity, cardiovascular conditions, anatomical differences, and family history [17, 18] while highlighting potential risks of untreated OSA, including hypertension, diabetes, heart failure, stroke, depression, and anxiety [18, 19]. The chatbot also correctly details primary treatments, emphasizing CPAP therapy as the first‐line treatment, often accompanied by lifestyle modifications for weight reduction [20, 21].

Question 2: “What are the risks of untreated obstructive sleep apnea?”

GRADE: Satisfactory Response Requiring Minimal Clarification

ChatGPT's response comprehensively addresses the risks associated with untreated OSA, covering both physical and psychological aspects. It highlights major physical complications such as increased risk of hypertension, heart disease, stroke, diabetes, and motor vehicle accidents [18, 22]. Likewise, psychological repercussions including depression, anxiety, and cognitive decline were also noted [19, 23, 24]. Additionally, the indirect effects on interpersonal relationships, such as insomnia symptoms and daytime fatigue for bed partners, were noted as well [25, 26]. The response could be expanded to include untreated OSA's effects on work performance [27, 28] and social interactions [29], offering a more comprehensive view. Similarly, addressing the correlation between OSA severity and the degree of risk for adverse outcomes like cardiovascular disease [30] and depression [19] is noteworthy.

Question 3: “What treatment options exist for obstructive sleep apnea?”

GRADE: Satisfactory Response Requiring Moderate Clarification

The chatbot correctly noted that OSA treatment is tailored based on severity and individual factors, such as anatomy, body mass index (BMI), and surgical eligibility. CPAP therapy is correctly identified as the first‐line treatment, complemented by alternatives like dental devices, surgical interventions, and lifestyle modifications. However, CPAP therapy is limited by patient adherence, ranging from 20% to 80% [20, 31, 32]. For patients where CPAP fails, the chatbot mentions alternatives like oral appliance therapy and uvulopalatopharyngoplasty, but omits other therapies such as myofunctional therapy, bariatric surgery, or newer innovations such as Inspire's HNS, approved by the United States Food and Drug Administration (FDA) in 2014 and implanted in over 50,000 patients worldwide [33]. HNS provides respiratory synchronized stimulation of the hypoglossal nerve [32], improving OSA with fewer complications compared to upper airway surgery [34]. Multiple studies have demonstrated its efficacy [35, 36, 37], underscoring its increasing role as a viable treatment option that should be acknowledged.

Preoperative Counseling on Hypoglossal Nerve Stimulators

Question 4: “How does the hypoglossal nerve stimulator work?”

GRADE: Satisfactory Response Requiring Minimal Clarification

ChatGPT provided a comprehensive explanation of HNS therapy, detailing how the device functions and how it benefits those who cannot tolerate CPAP. The response can be improved by explicitly stating that HNS is active at night and inactive during the day, clarifying that it does not affect daytime upper airway activity. Likewise, it should clarify that patients use a remote to turn the device on and off and adjust stimulation intensity, emphasizing patient control and the importance of patient adherence in therapy success. Additionally, mentioning that most patients do not awaken due to the stimulation may help reduce patient concerns. Hofauer et al. found that 86.3% of patients did not awaken from the stimulation and patients became increasingly accustomed to stimulation with continued use [38]. These details enhance the patient's understanding of the device's function and impact on daily life.

Question 5: “What determines if I can get a hypoglossal nerve stimulator to treat my obstructive sleep apnea?”

GRADE: Satisfactory Response Requiring Moderate Clarification

ChatGPT provided a foundational understanding of HNS therapy eligibility, hinting at a minimum severity necessary for treatment, failing CPAP, and suitable anatomic features. However, it omitted that these anatomic characteristics are assessed through a separate procedure. Drug‐induced sleep endoscopy (DISE), performed under general anesthesia, is required for determining HNS eligibility through identifying contraindications such as complete concentric collapse at the velum [39]. Other anatomical factors, such as lateral oropharyngeal collapse, are still under investigation [39, 40]. Additionally, ChatGPT overlooked criteria such as the OSA severity, BMI ≤ 40 kg/m² and that > 75% of sleep apnea must be obstructive [33]. Finally, the response overlooked the role of HNS therapy in pediatric patients with Down syndrome. Given its success in managing OSA in these patients, this omission overlooks a relevant aspect of treatment diversity and success [41, 42].

Question 6: “What does surgery look like for hypoglossal nerve stimulator implantation?”

GRADE: Satisfactory Response Requiring Moderate Clarification

While the chatbot accurately detailed the use of general anesthesia, procedure duration, and the need for a neck incision, it omitted the chest wall incisions needed for device implantation. Originally, HNS implantation involved three incisions: one for the stimulation cuff via a modified submandibular approach, another for the internal pulse generator (IPG) on the anterosuperior chest wall, and a third for the respiratory sensing lead on the inferolateral chest wall [43]. The introduction of a two‐incision technique, as detailed by Kent et al., allows for the placement of the respiratory sensing lead through the anterosuperior chest wall incision, thereby eliminating the need for the third incision [44]. This technique, FDA‐approved in March 2021, maintains therapeutic effectiveness, reduces operative time, and does not increase the risk of complications [44, 45, 46]. Finally, the chatbot overlooks the increasingly common practice of same‐day discharge due to the procedure's safety and low complication risk [35, 36, 37, 47].

Question 7: “What are the possible risks of hypoglossal nerve stimulator implantation?”

GRADE: Satisfactory Response Requiring Moderate Clarification

ChatGPT's response successfully covers common adverse events in HNS implantation. The STAR trial found that most (88%) nonserious events, such as sore throat, pain, and muscle soreness, occurred within 30 days postoperatively, with a low risk (< 2%) of life‐threatening events [35], as further corroborated by Thaler et al. [37] However, the response should have emphasized that many adverse events are temporary and resolve over time. Woodson and colleagues' 5‐year ADHERE analysis found all initially reported cases of tongue weakness, postoperative discomfort independent of incision, intubation effects, headache, and other postoperative symptoms to have resolved by 12 months, with only 2.6% (n = 1) reporting continued incision discomfort [36]. Additionally, ChatGPT omitted known complications like pneumothorax and stimulation discomfort. Nord et al. found pneumothorax incidence to be 0.21% in a large data set [48], while the STAR trial found stimulation discomfort to decrease over time, dropping from an initial 40% to 8% at 12 months and 5% at 48 months [35, 36]. Naturally, not all implanted with HNS may demonstrate a significant improvement of their OSA; therefore, the need for further or adjunct treatment to effectively manage their OSA is a possible complication. Finally, ChatGPT should have highlighted the potential need for device explantation in many cases of infection. Ali et al. found that all three reports of infection required device explantation [43]. Likewise, analyses of the Manufacturer and User Facility Device Experience (MAUDE) database found infection to be the leading cause of explantation (64.7% and 63.3%) [49, 50]. Emphasizing the transient nature of many complications and the risk of serious infection would provide a clearer risk profile for patients.

Postoperative Counseling on Hypoglossal Nerve Stimulators

Question 8: What does recovery following hypoglossal nerve stimulator implantation look like?

GRADE: Satisfactory Response Requiring Moderate Clarification

ChatGPT's response provided an initial understanding of common postoperative symptoms like temporary discomfort, swelling, and bruising [35, 36, 37], and advised patients to follow care instructions and communicate with their healthcare team. However, other typically experienced postoperative symptoms such as sore throat from intubation, nausea from anesthesia, and muscle soreness were omitted [35]. Including these would offer a more comprehensive understanding of the postoperative experience. The response should also clarify immediate postoperative restrictions: the ipsilateral arm is typically suspended in a sling for the first 24 h, and strenuous exercise should be avoided for a month. While ChatGPT mentioned dietary modifications, patients can usually resume a normal diet immediately [51]. Additionally, a clearer timeline of follow‐up appointments would be helpful: a postoperative visit typically occurs 1–2 weeks after implantation, device activation at 4–6 weeks, and efficacy testing several months later via a home sleep test. Providing these details would better inform patients about the recovery process and their follow‐up schedule.

Question 9: What restrictions will I have with the hypoglossal nerve stimulator implanted to treat my sleep apnea?

GRADE: Satisfactory Response Requiring Moderate Clarification

ChatGPT's response provided a basic overview of postoperative restrictions, particularly physical limitations. It would have benefitted from detailing the aforementioned specific immediate postoperative restrictions and including long‐term limitations, such as depth limits for scuba diving or hyperbaric chamber use [52]. The response correctly mentioned avoiding certain electromagnetic devices like magnetic resonance imaging (MRI), but newer IPG models (3028) are MRI‐compatible assuming proper pre‐imaging guidelines are followed, such as deactivating the device and ensuring the MRI's field strength is within limits [33]. Importantly, ChatGPT omitted the contraindication for diathermy, which can damage the implant and surrounding tissue, requiring both patients and surgeons to be aware of this limitation and avoid the use of diathermy in future operations [52]. Finally, while airport security devices have not been shown to interfere with implants [53], these implants can trigger metal detectors [54]. Per the United States Transportation Security Agency (TSA), patients should inform the TSA about the presence of their implant before security screening [55].

Question 10: Why would I have to get my hypoglossal nerve stimulator implant replaced?

GRADE: Satisfactory Response Requiring Moderate Clarification

ChatGPT's response outlines reasons for HNS replacement, such as device malfunction, weight changes, and upgrading to MRI‐compatible models. Alapati et al. reported a 4.6% failure rate at an average of 2 years postimplantation, mostly (68.8%) due to respiratory sensing lead damage [56]. However, the response should include device migration as a cause for malfunction. Islam et al. found a 1.1% migration rate, all requiring reoperation, while analysis of the MAUDE database found that 16.4% of sense lead malfunctions were due to migration [57]. ChatGPT also overlooked infection as a cause for reoperation. Ali et al. reported a 1.4% infection rate, all requiring explantation [43], while Certal et al. reported a 0.5% incidence managed with explantation [58]. Finally, the response omitted other causes for reoperation, such as IPG battery replacement (lasting ~11 years) and device erosion/extrusion [41, 42].

Data Availability Statement

The responses produced by ChatGPT can be found within the supplementary files.