Exploring the Roles and Challenges of ChatGPT in Empowering HI Professionals

ChatGPT is an advanced language model that uses deep learning to understand and generate human-like responses. It has been trained on a huge amount of text data, which helps it learn the patterns and meanings of language. The model is built on a special architecture called transformer networks, which allows it to process and make sense of natural language.  

During its training, ChatGPT goes through two main stages. First, it learns by predicting the next word in a sentence, which helps it understand how words fit together. Then, it is fine-tuned on specific tasks and data to improve its responses. 

ChatGPT exhibits remarkable benefits, as it comprehends and generates natural language responses with a high degree of fluency and coherence. It can understand queries, infer meaning, and provide contextually relevant answers. However, it has limitations. ChatGPT may occasionally produce incorrect or nonsensical responses, struggle with ambiguous queries, or exhibit sensitivity to input phrasing. 

Applications in Health Information and Informatics 

The advent of the next digital revolution brings artificial intelligence (AI) technologies like ChatGPT that have the potential to revolutionize workflows in healthcare delivery. These technologies offer immense support - from empowering patients with access to comprehensive health data to assisting caregivers in creating instructions and summarizing visits for referrals, coding, billing, and research. With access to such powerful tools, health systems can analyze healthcare data to facilitate timely, cost-effective, and informed decision-making, ultimately improving the quality of care. 

AI technologies are revolutionizing healthcare with a range of advanced applications. These include health education to provide individuals with information on medical conditions, treatments, and preventive measures, as well as facilitating symptom checking for preliminary guidance. AI also improves clinical documentation by capturing patient-provider conversations and auto filling missing data elements. Partnerships between Epic, Microsoft, UC San Diego Health, UW Madison, WI, and Stanford Health Care enable generative AI to draft responses to patient questions through the patient portal.  

Additionally, collaborations such as Cerner Enviza with the Food and Drug Administration aim to develop AI tools to enhance protocols and drug safety research. The applications of AI extend to various workflow and administrative tasks, including predictive models for appointment no-shows, customer service chatbots, and algorithms for error prevention and calculations. Embracing AI empowers healthcare organizations to optimize workflows, enhance efficiency, and ultimately deliver improved healthcare outcomes, driving innovation and benefiting the populations served. 

AI/ChatGPT have the potential to greatly support healthcare delivery, education, services, and technology advancements. Even if your organization hasn't yet embraced AI and ChatGPT, getting started is easy and can help with time-consuming tasks. For instance, ChatGPT can assist in drafting non-sensitive emails, updating language in policies and procedures, creating job descriptions and templates, improving data visualizations, and outlining disciplinary actions.  

These tasks, though tedious, are made easier with ChatGPT. It's a great starting point, especially for leaders, to leverage its power for minor, non-healthcare specific use cases. Online resources provide guidelines for asking detailed questions and achieving the best results.  

Following are guidelines drawn from human resources/talent management expert Laurie Ruettimann for effectively employing ChatGPT in the mentioned use cases:  

• Clearly define the scope and limitations to maintain patient confidentiality in the healthcare delivery system. 
• Use specific keywords and phrases for more accurate results. 
• Break down complex questions into smaller parts. 
• Be patient and willing to experiment. 
• Adopt a journalistic approach and ask follow-up questions. 

Here are examples of these guidelines in action: 

Ethical Considerations 

The use of ChatGPT in healthcare raises important ethical concerns regarding privacy, data security, and confidentiality similar to those associated with electronic health records (EHRs). In addition to these concerns, there are additional challenges related to ChatGPT, including bias, misinformation, cheating, plagiarism, job displacement, power dynamics, accountability, and data inaccuracies. 

Ensuring the protection of privacy and secure handling of sensitive health information are crucial considerations. ChatGPT has the potential to retain and learn from user input, which may pose risks of exposing personal information. It is important to note that ChatGPT does not comply with current privacy laws such as the Health Insurance Portability and Accountability Act of 1996, making it noncompliant and illegal for handling patient data. Seeking guidance from healthcare professionals and adhering to organizational privacy policies and regulations is advised. 

OpenAI has faced legal challenges related to privacy law violations and unauthorized data usage without consent. Data security is of utmost importance, requiring strict measures such as encryption, secure authentication, and authorized access. De-identifying protected health information (PHI) and minimizing bias in AI language models are crucial steps to protect patient privacy and foster trust. 

The White House emphasizes the significance of public assessment, aligning with the principles of the AI Bill of Rights and Risk Management Framework. DEF CON 31, a hacker convention, will include a public assessment of ChatGPT and other AI platforms. Meanwhile, efforts are being made to address bias and algorithmic discrimination in AI technologies. 

Bias in ChatGPT and similar models can arise from various factors, including training data, algorithms, use cases, and policy decisions, which can perpetuate stereotypes and incorrect assumptions. Overall, the responsible utilization of ChatGPT mandates safeguarding privacy, upholding data security, and fostering accountability, all while proactively tackling biases and healthcare-related challenges. Health information (HI) professionals have a responsibility to ensure ethical use of chatbots and AI in healthcare settings. They should also use their own professional judgment and expertise to evaluate the benefits and risks of using chatbots and AI for health purposes. 

In the 2023 article Leveraging ChatGPT and Generative AI in Healthcare Analytics, author Lisa Eramo  writes about ethical principles and guidelines that are relevant for the use of ChatGPT and AI in healthcare. The article states that health information professionals should adhere to the AHIMA Code of Ethics and follow the best practices for data governance, privacy, and compliance when using ChatGPT and AI tools. Additionally, the article suggests that HI professionals should collaborate with other stakeholders, such as clinicians, researchers, and patients to ensure ethical applications of ChatGPT and AI in healthcare analytics. 

To this end, healthcare professionals should apply human oversight, preserve patient confidentiality, mitigate bias, ensure technical dependability, and emphasize transparency in their utilization of ChatGPT. 

Future Directions and Challenges 

The future of health information management (HIM) and informatics holds immense potential with the application of AI. AI technologies can revolutionize patient engagement, precision medicine, and population health by leveraging various data sources and creating innovative experiences. 

One exciting avenue is the development of AI-powered conversational agents such as chatbots, voice assistants, and digital coaches. These intelligent systems can engage in natural language conversations with patients, answering their questions, providing feedback, and motivating them to adhere to their care plans. By enabling personalized and interactive communication, these AI-driven tools can enhance patient engagement and improve health outcomes. 

Furthermore, AI can facilitate gamified and immersive experiences that promote patient learning and behavior change. By incorporating game elements and virtual reality, AI technologies can create engaging and interactive platforms to educate patients, encourage healthy habits, and facilitate positive behavioral modifications. 

Another vital application of AI in HIM is the integration and analysis of diverse health data sources. AI algorithms can combine EHRs, genomic data, wearable device information, social media data, and environmental sensor data to create a comprehensive and dynamic patient profile. This integration enables precision medicine by tailoring treatment plans to an individual patient’s unique characteristics and needs.  

Additionally, AI-powered analytics can analyze large-scale population health data to identify patterns, predict disease outbreaks, and inform public health interventions. 

In conclusion, the integration of AI technologies and ChatGPT in healthcare workflows has shown potential benefits and challenges. ChatGPT can enhance healthcare documentation, communication, and clinical decision-making while improving patient experiences.However, accuracy, privacy, and bias pose challenges when implementing AI in healthcare. 

Responsible and ethical implementation is crucial, prioritizing patient privacy, data security, diverse training datasets, and addressing biases and errors transparently. Ongoing discussions, updates, and collaboration among healthcare professionals, researchers, policymakers, and AI developers are necessary for ethical development, validation, and regulation of AI systems in healthcare. 

By responsibly implementing AI technologies, such as ChatGPT, we can improve patient outcomes, enhance healthcare delivery, and create a more patient-centered and efficient healthcare system. 

Shannon H. Houser, PhD, MPH, RHIA, FAHIMA, is a professor in the department of health services administration at the University of Alabama at Birmingham in Birmingham, AL.  

Gerard Pappalardo, RN, MSN, CCM, CCDS, CDIP, CCS, is a clinical nurse of health services organization, quality operations, and clinical data registries at the Mount Sinai Health System, in New York City, NY. 

Megan Pruente, MPH, RHIA, is the director of professional services at Harris Data Integrity Solutions in Carlsbad, CA.