Generative AI is a hot topic in the tech world, a field that has accelerated thanks to the massive strides that AI engines like GPT-4, Bard, and Bing AI have made throughout 2023.
More and more industries are utilizing generative AI to streamline their workflows, either for individual workers, or in a broader and more systematic way. Most of the discussion around generative AI has focused on text and image generation; however, this can be applied to acute care in revolutionary ways to reduce clinician burden from documentation, reducing alert fatigue and improving patient experience with virtual assistants. But how does this new technology work, and how is it different from previous AI models?
Previous iterations of AI have focused on machine learning, in which an AI is trained to detect patterns and make predictions based on those patterns. Generative AI, on the other hand, is the next step where the AI engine now has the ability to create new content. Generative AI has many applications, but one of the most interesting is seen in large language models (LLMs). This is what is used in services like ChatGPT.
Generative AI in Acute Care
With large language models, the medical industry can move toward automation for one of its most time-consuming tasks: documentation. At every level, the need to document dominates medical care and is often a point of frustration for both clinicians and patients. A 2022 paper published in BMC Nursing, found that American nurses spend a staggering 25%-41% of their time on documentation1. Large language models offer a new way to ensure that documentation still happens, while doctors and nurses engage directly with their patients. A nurse can speak with their patient while an AI autonomously notates the conversation and directly adds the notes to the EHR, just like having an assistant take dictation. Unlike a simple voice to text service, and thanks to the added nuance of large language model AI, the documentation goes beyond pure dictation and is more like the documentation that a nurse might notate: Only gathering relevant facts that might need to be referenced later.
Today, between 72% and 99% of clinical alarms are false. This problem has been so prevalent that, since 2014, the Joint Commission has focused on alarm fatigue as a national patient safety goal. A solution that employs AI for in-room monitoring to prevent falls and pressure ulcers can reduce alarm fatigue by up to 95%. Furthermore, the same solution can make fall prevention predictive by providing a 31 to 65-second warning ahead of a bed exit for high fall-risk patients. It can also reduce falls by over 50% by utilizing edge AI capability on LIDAR sensors. Solutions like these enhance clinician engagement by reducing stress, and by preventing never events in hospitals.
Similarly, an AI virtual assistant in the patient room can be the first line of communication to meet patient needs and improve patient engagement. Using large language models, the assistant can have a human-like, iterative conversation with a patient depending on what they ask. For example, if a patient communicates that they are in pain, the virtual assistant can clarify further by asking where the pain is, how severe, how long the pain has been present, and so on. The AI automatically documents the interaction in the EHR and issues a notification to the assigned nurse, turning what might have taken hours to be addressed with standard rounding into a 2-minute conversation. This virtual assistant can help patients to be more comfortable as well, by responding to food and drink orders. Because the assistant is integrated with the patient’s medical record, the AI can even ensure that the food and drink requested is suitable for the patient according to their health needs and dietary restrictions. In this way, a virtual assistant empowered with generative AI works like a switchboard operator for the hospital floor: Each request or question is considered and then directed to who needs to respond. Each interaction is automatically documented, relieving the burden on busy nurse teams to delegate tasks as they come.
These are just a few applications that generative AI will have in healthcare. The horizon is ever expanding with new ways that we can make our medical care more and more efficient and human-focused while using advanced tech tools. These applications of AI are active today with AI-powered solutions protecting patients and supporting nursing teams via autonomous patient monitoring, virtual nursing integrations, and wireless vitals monitoring. Learn more about how AI is improving patient and clinician experience in acute care here.
- https://bmcnurs.biomedcentral.com/articles/10.1186/s12912-022-00811-7#:~:text=Even%20though%20the%20actual%20time,%25%20%5B16%2C%2017%5D.
- https://pubmed.ncbi.nlm.nih.gov/24153215/