Skip Navigation
Skip to contents

JMD : Journal of Movement Disorders



Page Path
HOME > J Mov Disord > Volume 16(2); 2023 > Article
Potential Benefits and Perils of Incorporating ChatGPT to the Movement Disorders Clinic
Andres Deikcorresp_iconorcid
Journal of Movement Disorders 2023;16(2):158-162.
Published online: May 24, 2023

Parkinson’s Disease and Movement Disorders Center, Department of Neurology, University of Pennsylvania, Philadelphia, PA, USA

Corresponding author: Andres Deik, MD, MSEd, FAAN Parkinson’s Disease and Movement Disorders Center, Department of Neurology, University of Pennsylvania, 330 S 9th Street, Philadelphia, PA 19107, USA / Tel: +1 (215) 829-7049 / Fax: +1 (215) 829-6606 / E-mail:
• Received: April 12, 2023   • Revised: April 18, 2023   • Accepted: April 21, 2023

Copyright © 2023 The Korean Movement Disorder Society

This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License ( which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.

  • 89 Download
  • 2 Web of Science
  • 3 Crossref
ChatGPT (; Chat Generative Pre-trained Transformer) is an artificial intelligence (AI) language model created by the AI lab OpenAI (OpenAI Inc., San Francisco, CA, USA) that can generate contextually relevant text on many subjects. ChatGPT understands and answers to natural language input and is trained on massive amounts of text data, including books, articles, and websites [1].
Several versions of this Generative Pre-trained Transformer (GPT) have been released (the latest iteration, GPT-4, became public mid-March 2023), and it is now one of the largest language models ever devised. Given its versatility and speed, reports of the application of ChatGPT in healthcare have risen steeply in 2023 [2-9], and the fields of neurology [10] and movement disorders are not excluded from this revolution [11]. In this viewpoint, the potential benefits and shortcomings of integrating ChatGPT to the movement disorders clinic will be discussed, as well as possible future directions.
There are numerous ways in which ChatGPT can assist the movement disorders clinician [7]. In the author’s opinion, the following are the most relevant and impactful (Figure 1).
Expediting administrative work
ChatGPT’s largest potential may lie within its ability to rapidly generate documents that are necessary for routine movement disorders care [6]. The largest demand for these documents probably relates to letters of medical necessity for specialty medications (now used commonly for the management of patients with Parkinson disease), but many other documents of varying complexity can also be generated. Indeed, clinicians can now instantly obtain letters that, for example, justify the use of botulinum toxin injections for facial dystonia, support the use of the new generation of VMAT-2 inhibitors to treat chorea or explain why a patient with a deep brain stimulator should not walk through a metal detector. This remarkable capability has the potential to reduce both physician burnout and administrative times, ultimately benefitting the patient.
Generating texts for patient education
Using ChatGPT, movement disorders specialists can now instantly provide their patients either a printed document, a text message or an e-mail that is written in lay terms and in their native tongue and explaining a vast variety of topics, from levodopa titration to interpreting genetic testing results. ChatGPT’s grammatically accurate translating capabilities can be particularly helpful in multiethnic countries or for movement practices that offer telemedicine overseas [9] Movement specialists can also decide whether to create these documents before the clinic visit or in real-time; regardless, this advancement will probably boost clinic efficiency and patient safety while reducing the amount of post-visit patient inquiries.
Data synthesis
Like in other medical fields, movement specialists often spend significant time searching online databases before making decisions in unfamiliar situations. ChatGPT can synthesize this information by relevance and within a context provided by the clinician. For instance, a movement specialist could inquire about published treatments on the management of relatively rare conditions, like episodic ataxia, or ask for a comparison of prices among genetic testing companies. Being able to automate these searches could allow them to offer answers to their patients while still facing the patient, which can be particularly helpful for rural practices lacking easy access to subspecialized movement disorders care [12].
Expanding the differential diagnosis
Movement specialists may sometimes be able to categorize a patient’s phenomenology while still not being confident on the actual diagnosis. This is especially true in the dynamic field of neurogenetics, given its expanding number of overlapping genotype-phenotype correlations. ChatGPT can help put together a differential diagnosis based on signs and symptoms identified by the specialist. By training the model with data from sites like OMIM ( and PubMed (, ChatGPT can aid this task. It is important, however, that clinicians use the generated document only as a guide and apply sound clinical judgement when interpreting the results.
As enticing as including ChatGPT in movement disorders care may seem, certain caveats must be taken into consideration. The following are some issues to keep in mind:
The potential for data breaches or unauthorized access to protected health information (PHI) is perhaps the largest concern at this time when utilizing this technology for movement disorders care [13]. Whereas ChatGPT does not store or process PHI, it is plausible this technology could be exploited by third parties with egregious purposes. Examples of sensitive information in movement disorders practice that could be misused are abundant, the following being just a select few:
  • • Entering the genetic status of a patient harboring a mutation in the Huntingtin gene

  • • Creating a letter that details a Parkinson patient’s legal or financial struggles secondary to impulse control disorders

  • • Preparing a dossier for an ataxia patient that is planning on applying for permanent disability

The Health Insurance Portability and Accountability Act (HIPAA) is an American federal law designed to protect the privacy and security of individuals’ PHI. It is possible that HIPAA could be amended with the rise of AI language models [14], or that healthcare facilities will each establish their own guidelines based on HIPAA’s broad principles [8]. Until either of these scenarios takes place, it is important that movement specialists utilize their best judgment and consider avoiding entering PHI into ChatGPT.
ChatGPT’s fallibility stems from the complexity of the field of movement disorders and the limited amount of training data available to the model. If a clinician were to ask the ChatGPT for treatment recommendations for a patient with a relatively unusual condition, it may provide an answer that may be factually correct, but could be incomplete, and not necessarily generalizable. Even for common movement disorders like Parkinson Disease (PD), ChatGPT could generate a list of signs and symptoms, when asked, that is not exhaustive. ChatGPT can also generate incorrect bibliography and clinicians are urged to always assess these for validity and appropriateness [12].
ChatGPT’s responses are only as adequate as the data it is trained on, and it is susceptible to inherent gender, racial or cultural bias within the data. For instance, if the training data contains a disproportionate number of examples related to PD (compared to other movement disorders), the model may be more likely to generate responses related to Parkinson when prompted with a movement disorder-related query. Furthermore, if the training data is biased towards a particular demographic, such as middle-aged white males (who are more likely to be diagnosed with PD), the model may also inaccurately suggest that Parkinson is the most prevalent movement disorder across all age groups. This bias may lead to incorrect conclusions and inappropriate treatment decisions [4,13].
At the time this manuscript was written, ChatGPT had a September 2021 “cutoff date”. 3 This means that the model only had access to information that was available up until that point in time and couldn’t generate text based on more recent events. This can impact movement disorders practice in relation to other rapidly changing fields, like, again, the field of neurogenetics. If, for example, a clinician wanted to consult the tool on genotype-phenotype correlations for a recently described mutation, the model may not be able to generate this response given its novelty. The model may also not be able to comment on a recently recognized side effect to a medication, or on the results of clinical trials that were published after the “cutoff date”. Clinicians should be aware of this limitation and should be urged to cross reference other search engines whenever trying to make decisions based on very recent data. Hopefully, technology may overcome this limitation in the future, allowing the tool to generate results based on current information.
Patients’ access to the technology
Could ChatGPT become the new “Dr. Google”? Ever since the popularization of internet search engines in the early 1990s, patients have been able to query for differential diagnoses and treatment recommendations based on their self-identified symptoms. ChatGPT now offers patients the ability to curate and categorize the results from these queries, as well as to pre-specify the complexity of the language of the output text [15]. Whereas it is likely that the tool will often be correct in its predictions, patients need to be counseled that the tool should not replace medical care, and that there are situations in which the phenomenology is ambiguous (for example, dystonic tremor vs. essential tremor, or early chorea vs. ataxia) and in which the tool is no match for the movement specialist’s trained eye.
One can only speculate on the following steps in the exponential growth of this technology, and how it will transform future movement disorders care. In the author’s opinion, the next big step will be the integration of AI with the different electronic medical record (EMR) platforms available. It is difficult to predict all the downstream consequences of this strategic partnership, but it is reasonable to envision a future in which AI-powered EMRs could use voice recognition software to “listen in” to the movement disorders clinic visit, and document the clinic note autonomously. Clinicians could, of course, edit the final product to their liking, particularly since the movement examination will need to be entered manually until reliable, AI-powered video interpretation becomes a reality.
It is also possible an AI-powered EMR could generate diagnostic and/or therapeutic suggestions, based on the data compiled within the clinic note and in the context of all previous encounters, even those at outside institutions. An automated review of a patient’s test results in the context of their medication utilization history could allow the detection of health trends and expedite the introduction of proactive interventions. Furthermore, data from one patient could be compared to aggregated, deidentified data from others within the same movement disorders practice (or other practices), which could help recognize outliers in disease progression and medication response. This could help to discern, for example, patients with early multiple systems atrophy from those with idiopathic PD.
ChatGPT could also revolutionize telemedicine delivery. Telemedicine has become an indispensable tool since its rapid and widespread adoption during the COVID-19 pandemic, and will likely continue to have a role in the post-pandemic era. Movement disorders specialists are particularly sensitive to the need for telemedicine to persist, as many of their patients (like those with advanced parkinsonism, chorea, or ataxia) have gait dysfunction limiting their ability to attend in-person visits. Moreover, many of them are elderly, frail, and immunocompromised, increasing their risk of complications from infectious diseases.
To enhance telemedicine care, ChatGPT could be integrated with remote patient monitoring systems, both wired and wireless, and prompted to periodically inform the movement specialist about any concerning changes in symptoms. ChatGPT’s accurate interpretation of the data from these monitors could also help confirm diagnostic impressions, especially in patients who are exclusively assessed over video, and chart disease progression. ChatGPT could also ask patients about both motor and non-motor symptoms ahead of their telemedicine appointments, remind patients about these appointments and assist them with scheduling issues. Lastly, patients could be granted access to sanctioned chatbots that could be trained to answer simple health-related questions, and to refer the patient to the specialist when appropriate.
One last potential future benefit of this technology could be the automatization of prior authorization and appeal processes. Just like this technology has been found to be capable of passing medical board examinations [16,17], it could answer specific questions raised by payers regarding a patient’s comorbidities, prior therapeutic trials, or contraindications, among others. It could also forward the responses back to payers autonomously, or with minimal supervision.
Time will tell whether any of these predictions come to pass. As competitors to ChatGPT come online (like Google’s Bard,, it is possible that other tools could dominate the market, or, perhaps, some tools will prove to be better suited for some tasks. It also remains to be seen whether these innovations will strengthen the clinician-patient relationship and reduce physician burnout, or whether they will produce medico-legal headaches that will, ultimately, hinder its widespread adoption. In the meantime, the power of text-generating AI tools is undeniable, and, given the mismatch between the pace of technological progress and the implementation of regulation, caution is warranted when employing them to generate documents involved in movement disorders care [18,19].

Ethics Statement

This manuscript is not the product of a study that was submitted to the University of Pennsylvania’s Institutional Review Board. Informed patient consent was not necessary or acquired for this work.

Conflicts of Interest

The author declares that there are no conflicts of interest relevant to this work. However, over the previous 12 months, the author has served as site Principal Investigator for clinical trials and received research support from Insightec, Teva Pharmaceuticals, Cerevel Therapeutics, Prevail Therapeutics, Addex Pharmaceuticals and Lundbeck Therapeutics. The author has also received publishing royalties from a UpToDate and Elsevier.

Funding Statement


Figure 1.
Applications of ChatGPT (Chat Generative Pre-trained Transformer; OpenAI Inc., San Francisco, CA, USA) in the movement disorders clinic. The figure illustrates the four most significant ways that ChatGPT can simplify movement disorders practice, listed in clockwise order starting from the left upper quadrant. These ways include: 1) the quick redaction of various types of documents required for movement disorders clinical practice, such as medical excuses, summaries of medical records, letters of medical necessity, and letters certifying fitness to drive, among many others, required by patients, payers, and other entities; 2) the generation of patient-oriented documents in plain language (and in multiple languages) by movement specialists, where they can explain, for example, the significance of a diagnosis, the medications used for its treatment, the side effects of such medications and their titration schedules; 3) the efficient synthesis and summarization of the literature on movement disorders, based on the specialist’s parameters; and 4) the utilization of ChatGPT’s vast source data to expand the differential diagnosis in cases where a patient’s history and examination fail to clearly define an etiology. Of note, whereas this figure was assembled manually by the author, the five individual images were generated by Microsoft Bing’s Image Creator (, which is powered by OpenAI’s DALL-E ( Image was generated on April 12, 2023.
  • 1. Stokel-Walker C, Van Noorden R. What ChatGPT and generative AI mean for science. Nature 2023;614:214–216.ArticlePubMedPDF
  • 2. Cascella M, Montomoli J, Bellini V, Bignami E. Evaluating the feasibility of ChatGPT in healthcare: an analysis of multiple clinical and research scenarios. J Med Syst 2023;47:33.ArticlePubMedPMCPDF
  • 3. Vaishya R, Misra A, Vaish A. ChatGPT: is this version good for healthcare and research? Diabetes Metab Syndr 2023;17:102744.ArticlePubMed
  • 4. Sallam M. ChatGPT utility in healthcare education, research, and practice: systematic review on the promising perspectives and valid concerns. Healthcare (Basel) 2023;11:887.ArticlePubMedPMC
  • 5. Ferres JML, Weeks WB, Chu LC, Rowe SP, Fishman EK. Beyond chatting: the opportunities and challenges of ChatGPT in medicine and radiology. Diagn Interv Imaging 2023;Mar. 14. [Epub]. Available from: Article
  • 6. Ali SR, Dobbs TD, Hutchings HA, Whitaker IS. Using ChatGPT to write patient clinic letters. Lancet Digit Health 2023;5:e179–e181.ArticlePubMed
  • 7. Rao A, Pang M, Kim J, Kamineni M, Lie W, Prasad AK, et al. Assessing the utility of ChatGPT throughout the entire clinical workflow. medRxiv [Preprint] 2023;[cited 2023 April 2]. Available at: Article
  • 8. Will ChatGPT transform healthcare? Nat Med 2023;29:505–506.ArticlePubMedPDF
  • 9. Patel SB, Lam K. ChatGPT: the future of discharge summaries? Lancet Digit Health 2023;5:e107–e108.ArticlePubMed
  • 10. Boßelmann CM, Leu C, Lal D. Are AI language models such as ChatGPT ready to improve the care of individuals with epilepsy? Epilepsia 2023;Mar. 3. [Epub]. Available from: Article
  • 11. Moisset X, Ciampi de Andrade D. Neuro-ChatGPT? Potential threats and certain opportunities. Rev Neurol (Paris) 2023;Mar. 21. [Epub]. Available from: Article
  • 12. Johnson D, Goodman R, Patrinely J, Stone C, Zimmerman E, Donald R, et al. Assessing the accuracy and reliability of AI-generated medical responses: an evaluation of the Chat-GPT model. Res Sq 2023;Feb. 28. [Epub]. Available from: Article
  • 13. Masters K. Ethical use of artificial intelligence in health professions education: AMEE Guide No.158. Med Teach 2023;Mar. 13. [Epub]. Available from: Article
  • 14. Rosenbloom ST, Smith JRL, Bowen R, Burns J, Riplinger L, Payne TH. Updating HIPAA for the electronic medical record era. J Am Med Inform Assoc 2019;26:1115–1119.ArticlePubMedPMCPDF
  • 15. Hopkins AM, Logan JM, Kichenadasse G, Sorich MJ. Artificial intelligence chatbots will revolutionize how cancer patients access information: ChatGPT represents a paradigm-shift. JNCI Cancer Spectr 2023;7:pkad010.ArticlePubMedPMCPDF
  • 16. Gilson A, Safranek CW, Huang T, Socrates V, Chi L, Taylor RA, et al. How does ChatGPT perform on the United States Medical Licensing Examination? The implications of large language models for medical education and knowledge assessment. JMIR Med Educ 2023;9:e45312.ArticlePubMedPMC
  • 17. Mbakwe AB, Lourentzou I, Celi LA, Mechanic OJ, Dagan A. ChatGPT passing USMLE shines a spotlight on the flaws of medical education. PLOS Digital Health 2023;2:e0000205.ArticlePubMedPMC
  • 18. Kitamura FC. ChatGPT Is Shaping the Future of Medical Writing but Still Requires Human Judgment. Radiology 2023;307:e230171.ArticlePubMed
  • 19. Zheng H, Zhan H. ChatGPT in scientific writing: a cautionary tale. Am J Med 2023;Mar. 9. [Epub]. Available from: Article

Figure & Data



    Citations to this article as recorded by  
    • (How) ChatGPT—Artificial Intelligence Thinks It Can Help/Harm Physiatry
      Jakub Jačisko, Viktor Veselý, Ke-Vin Chang, Levent Özçakar
      American Journal of Physical Medicine & Rehabilitation.2024; 103(4): 346.     CrossRef
    • Can ChatGPT diagnose my collapsing dog?
      Samira Abani, Steven De Decker, Andrea Tipold, Jasmin Nicole Nessler, Holger Andreas Volk
      Frontiers in Veterinary Science.2023;[Epub]     CrossRef
    • Bias and Inaccuracy in AI Chatbot Ophthalmologist Recommendations
      Michael C Oca, Leo Meller, Katherine Wilson, Alomi O Parikh, Allison McCoy, Jessica Chang, Rasika Sudharshan, Shreya Gupta, Sandy Zhang-Nunes
      Cureus.2023;[Epub]     CrossRef

    Comments on this article

    Add a comment

    JMD : Journal of Movement Disorders