Search
Search
Close this search box.

ChatGPT and the Future of Health Care

In the past year, discussion of digital health and, specifically, the impact of artificial intelligence (AI) has made headlines. Digital health refers to the use of technology, such as mobile devices, sensors, and software, to improve health care and health outcomes (Table). Digital health also represents a cultural change from an authoritarian (provider to patient) relationship to a democratic (shared decision-making) one.1 Within the health care industry, however, excitement is being balanced with caution about how AI can create efficient clinical workflows and improve personalized patient-centered care.

Table. Types of Digital Health

Artificial intelligence (AI), is a branch of computer science that includes development of smart machines that can perform tasks that typically require human intelligence.2
Machine learning (ML)‌ is a branch of AI that uses statistical techniques for fitting models to data and learning by training models with data.3
Deep learning (DL) is a more complex form of ML that can complete extremely complex tasks such as speech recognition through the use of natural language processing (NLP) and interpreting radiologic images. Deep learning, specifically NLP, is what allows programs to analyze clinical notes, prepare reports (eg, on radiology examinations), transcribe patient interactions, and conduct conversational AI.4
Large language models (LLMs) are trained to predict the likelihood of a given sequence of words based on the context of the words that precede it. These models can generate unique sequences of words never observed previously by the model that are plausible based on natural human language.4

ChatGPT

Open AI’s ChatGPT has become the most impactful LLM because of its ability to perform a wider array of natural language tasks. ChatGPT became mainstream after the November 2022 launch of the most recent version, which is unique in its ability to be conversational in an almost human way. The technology responds to multiple languages generating refined and sophisticated responses based on advanced modeling.5


Continue Reading

According to a ChatGPT generated response:

 “GPT (Generative Pre-trained Transformer) refers to a type of language model developed by OpenAI that can be used in health care applications such as clinical documentation, medical diagnosis, drug discovery, chatbots, and patient monitoring. GPT models have been exposed to a vast amounts of text data, including books, articles, and other written material, which allows them to learn the patterns and structures of language and generate responses. In health care, GPT models have been used for clinical documentation, medical diagnosis, drug discovery, and patient monitoring. Although GPT has numerous potential applications, there are also several challenges associated with its use in health care, such as data privacy, bias, and ethical concerns. Therefore, it is essential to carefully evaluate the utility and implications of GPT in health care practice, research, and patient care.”

The more context you provide the LLM, the better the answer; however, there is much room for improvement. Large language models have been known to hallucinate. This means that there are times when the LLM is fabricating information. Hallucinated facts are addressed directly in the OpenAI ChatGPT technical report: “Despite its capabilities, GPT-4 has similar limitations to earlier GPT models: it is not fully reliable [eg, can suffer from “hallucinations”], has a limited context window, and does not learn from experience. Care should be taken when using the outputs of GPT-4, particularly in contexts where reliability is important.”6

Practice Workflows and Patient Care

Digital health will improve clinical practice and patient care in many ways. Clinical practice and patient care have evolved from only being delivered within the walls of a clinical office or hospital to being available everywhere and anytime. The day when patients can schedule their medical appointments, share health data from their mobile devices with their providers, and manage their care to improve health outcomes is already here. As digital health and AI continue to evolve, task automation with personalized care will be more commonly integrated into practice.

For patients, digital health will aid in learning about their conditions, making lifestyle changes, and tracking health data. For providers, task automation of documentation, letter writing, and prior authorizations for medical devices and medications will improve the coordination of care. Furthermore, AI and DL models can help improve compliance with medical prescriptions and provide personalized medical treatment for various neurological disorders.8 ChatGPT and other LLMs also have the potential to enhance diagnostics, predict disease risk and outcome, and facilitate drug discovery among other areas in translational research.5

Diagnostic Imaging Clinical Decision-Making and Interpretation

Computed tomography (CT) scans and magnetic resonance imaging (MRI) revolutionized how we diagnose and treat various diseases by changing how to risk stratify patients, the workup process, and deciding on the next course of treatment. In the same vein, deep learning (DL) algorithms within radiology are changing the way we practice and screen patients. Findings from recent studies suggest that DL algorithms may be as good as health care providers in diagnostic imaging interpretation.9 Although these studies have not been proven in clinical practice, researchers are rapidly moving towards the day when that will be a reality. Findings from a systematic review and meta-analysis found DL models had higher sensitivity, specificity, and accuracy on optical coherence tomography (OCT) scans and retinal fundus photographs (FRP) for diabetic retinopathy (DR), age-related macular degeneration (AMD), and glaucoma.10 The same study found the DL algorithms had high sensitivity and specificity in identifying chest pathology on CT scans and chest radiography.10 The study authors also found that DL had high diagnostic accuracy to identify breast cancer on mammography, ultrasonography, and digital breast tomosynthesis.10 While much research is still needed in this space, use of AI in radiologic imaging is leading the way in augmenting the work of radiologists.

Risk Prediction and Stratification of Diseases

Managing and mitigating risk factors for cancer, cardiovascular disease, diabetes, and other chronic conditions can be a challenge. Recent findings suggest that AI modeling can aid in risk stratifying patients for stroke, cardiovascular disease, and diabetes complications, and is sometimes better than the traditional methods currently used such as atherosclerotic cardiovascular disease (ASCVD) or CHA2DS2-VASc scores.11,12,13 Furthermore, AI models can make health care more effective and efficient through early detection, diagnosis, risk stratification, severity prediction, mortality prediction, management, and decision-making in acute respiratory distress syndrome (ARDS).14 Progress also is being made in cancer screening. Through the use of machine learning (ML) and natural language processing (NLP), electronic health records can be scanned to identify high-risk individuals for cancer by using symptoms and previous imaging and laboratory studies to screen patients promptly, thereby improving early cancer diagnosis rates.15

Research

Health care research is one of the areas that will be greatly affected by AI. Areas of drug discovery, disease pattern recognition, and image interpretation are being studied. The drug discovery cycle will be significantly impacted and improved through the clinical trial design process. As noted by Harrer and colleagues, the major ways to infuse AI into the clinical trial design pipeline include cohort composition, patient recruitment, and patient monitoring.7 This will allow for acceleration of drug development from phase to phase and also allow for innovation and growth. Harrer et al concluded that “AI technology first needs to be tested alongside the existing technology it aims to complement or replace, and the added value must be demonstrated and benchmarked in an explainable, ethical, repeatable, and scalable way — not only to users but also to regulatory bodies. Following this approach, AI may be adopted into the clinical trial ecosystem step-by-step, making trials faster, while at the same time hopefully lowering failure rates and R&D costs.” 7

AI and Clinical Research

Another way AI will affect research is by increasing the number of pathways a researcher can explore with the existing constraints of time and money. For example, in the bioinformatics and genomic space, AI will enable the extraction of information from scientific journals, summarization and interpretation of the content, as well as the output of runnable code. Unlocking the ability for anyone to verify and reproduce a genomic experiment in potentially minutes. Even in the instances where the output of the model in question is less than correct, it can serve as a starting place for a researcher and potentially help solve the cold start problem that researchers face every day. Overall, increasing the efficiency of researchers and allowing them to focus on the science rather than the software will increase the efficacy of researchers.

Conclusion

Health care will likely continue to be shaped by advances in technology with an increased focus on the implementation of digital health and AI and their outcomes. Changes in the roles and responsibilities of health care providers may also occur as well as the development of new professions and career pathways to support the integration of technology into clinical care.16 While AI and digital health will not and should not replace health care professionals, roles will shift to embed technology in the day-to-day work of clinicians and augment patient care. It is imperative to continue with having human-in-the-loop in the training of AI technologies.

Several challenges remain in use of AI in health care such as data privacy, bias, and ethical concerns. Further research is needed to understand the best use and implications of ChatGPT in health care practice, research, and patient care.5

Dipu Patel, DMSc, MPAS, PA-C, is professor and vice chair for innovation at the University of Pittsburgh Doctor of PA Studies Program; Vishal Patel isChief Technology Officer and co-founder of Tinybio.

References

  1. What is Digital Health? The Medical Futurist. February 28, 2021. Accessed June 13, 2023. https://medicalfuturist.com/wiki/digital-health/
  2. Schroer A. What is artificial intelligence (AI)? Built In. Updated May 19, 2023. Accessed June 13, 2023. https://builtin.com/artificial-intelligence
  3. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):94-98. doi:10.7861/futurehosp.6-2-94
  4. Kung TH, Cheatham M, Medenilla A, et al. Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models. PLOS Digit Health. 2023;2(2):e0000198. doi:10.1371/journal.pdig.0000198
  5. Sallam M. ChatGPT Utility in healthcare education, research, and practice: systematic review on the promising perspectives and valid concerns. Healthcare (Basel). 2023;11(6):887. doi:10.3390/healthcare11060887
  6. OpenAI. GPT-4 Technical Report; 2023. https://cdn.openai.com/papers/gpt-4.pdf
  7. Harrer S, Shah P, Antony B, Hu J. Artificial intelligence for clinical trial design. Trends Pharmacol Sci. 2019;40(8):577-591. doi:10.1016/j.tips.2019.05.005
  8. Will ChatGPT transform healthcare? Nat Med. 2023;29(3):505-506. doi:10.1038/s41591-023-02289-5
  9. Liu X, Faes L, Kale AU, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019;1(6):e271-e297. doi:10.1016/S2589-7500(19)30123-2
  10. Aggarwal R, Sounderajah V, Martin G, et al. Diagnostic accuracy of deep learning in medical imaging: a systematic review and meta-analysis. NPJ Digit Med. 2021;4(1):65. doi:10.1038/s41746-021-00438-z
  11. Ding L, Liu C, Li Z, Wang Y. Incorporating artificial intelligence into stroke care and research. Stroke. 2020;51(12):e351-e354. doi:10.1161/STROKEAHA.120.031295
  12. Deep-learning AI may predict heart disease morbidity from chest x-ray. Mass General Hospital. January 10, 2023. Accessed June 13, 2023. https://advances.massgeneral.org/cardiovascular/article-external.aspx?id=1156
  13. Shao H, Shi L, Lin Y, Fonseca V. Using modern risk engines and machine learning/artificial intelligence to predict diabetes complications: a focus on the BRAVO model. J Diabetes Complications. 2022;36(11):108316. doi:10.1016/j.jdiacomp.2022.108316
  14. Rashid M, Ramakrishnan M, Chandran VP, et al. Artificial intelligence in acute respiratory distress syndrome: a systematic review. Artif Intell Med. 2022;131:102361. doi:10.1016/j.artmed.2022.102361
  15. Hunter B, Hindocha S, Lee RW. The role of artificial intelligence in early cancer diagnosis. Cancers (Basel). 2022;14(6):1524. doi:10.3390/cancers14061524.
  16. Lu M. Which jobs will be most impacted by ChatGPT? Visual Capitalist. Published April 30, 2023. Accessed June 13, 2023. https://www.visualcapitalist.com/cp/which-jobs-artificial-intelligence-gpt-impact/
  17. Haupt CE, Marks M. AI-generated medical advice-GPT and beyond. JAMA. 2023;329(16):1349-1350. doi:10.1001/jama.2023.5321
  18. Lee P, Bubeck S, Petro J. Benefits, limits, and risks of GPT-4 as an AI chatbot for medicine. N Engl J Med. 2023;388(13):1233-1239. doi:10.1056/NEJMsr2214184
  19. Giattino C, Mathieu E, Broden J, Roser M. Artificial Intelligence. Our World in Data. December 6, 2022. Accessed June 13, 2023. https://ourworldindata.org/artificial-intelligence
  20. ‌Sacks D. The give-to-get model for AI startups. March 29, 2023. Accessed June 13, 2023. https://sacks.substack.com/p/the-give-to-get-model-for-ai-startups
  21. Rim TH, Lee CJ, Tham YC, et al. Deep-learning-based cardiovascular risk stratification using coronary artery calcium scores predicted from retinal photographs. Lancet Digit Health. 2021;3(5):e306-e316. doi:10.1016/S2589-7500(21)00043-1
  22. Lin A, Kolossváry M, Motwani M, Išgum I, Maurovich-Horvat P, Slomka PJ, Dey D. Artificial intelligence in cardiovascular imaging for risk stratification in coronary artery disease. Radiol Cardiothorac Imaging. 2021;3(1):e200512. doi:10.1148/ryct.2021200512
  23. Beutel G, Geerits E, Kielstein JT. Artificial hallucination: GPT on LSD? Crit Care. 2023;27(1):148. doi:10.1186/s13054-023-04425-6

This article originally appeared on Cancer Therapy Advisor