University of Rochester Medical Center

Kathleen Fear

Through this article, Kathleen Fear discusses the transformative potential of generative AI in healthcare. She highlights how this technology can process unstructured data to improve patient care, reduce clinician burnout and enhance operational efficiency. At the University of Rochester Medical Center, generative AI tools are being developed to streamline tasks such as chart summarization and data extraction, ultimately creating a more connected and effective healthcare system.

For thousands of years, physicians have written down observations of their patients, treatment plans and outcomes; from pictographs to papyri to modern medical records, collecting patient data has long been a core part of medical practice. However, capturing data is just a first step. The real challenge lies in effectively using that information to improve patient care, which has remained elusive due to the complexity and fragmentation of healthcare data.

Medical documentation was primarily handwritten for centuries, making it difficult to aggregate, analyze and share information. The advent of computers and digital records marked a significant leap forward, allowing more efficient data storage and retrieval. Yet, despite these advancements, much of the valuable information in patient records remains locked in unstructured text like notes, letters and reports that are not easily accessible through traditional data processing methods or dispersed across multiple encounters or even across records at different healthcare institutions.

Further, the vast amount of data we collect comes at a high human cost. Clinicians spend an unreasonable amount of time on documentation, often at the expense of their well-being. Studies have shown that doctors can spend up to two hours on EHR-related tasks for every hour of direct patient care. This administrative burden not only leads to burnout but also detracts from the quality time that could be spent with patients.

"Generative AI is not just a technological advancement; it represents a paradigm shift in healthcare, turning centuries of collected data into actionable insights and significantly reducing the burden of documentation for clinicians"

This is where generative AI has the potential to transform healthcare. Generative AI is artificial intelligence that can create new content, like text, music, images, or other data. While generative AI tools like chatGPT and DALL-E have garnered much attention for their human-like chat responses and ability to mimic creativity, an underappreciated aspect of generative AI is its strength in processing unstructured data. It can sift through vast amounts of text, identify critical data points and extract meaningful information that might be overlooked. This capability is compelling in healthcare, where important details are often buried in narrative notes scattered throughout a patient's chart. Clinicians must manually trawl through that chart to create a complete picture of a patient’s health.

At the University of Rochester Medical Center (URMC), we are working to harness and apply the power of generative AI to improve access to and usability of data in the EHR. One of these projects focuses on helping primary care providers prepare for patients’ visits. Providers need to quickly understand the patient’s recent care journey, including any hospital visits, specialist consultations, or changes in medication since the last appointment. Traditionally, this would involve clicking around several areas in the EHR to discover and review the appropriate information, a time-consuming and often frustrating task. We are building chart summarization tools using generative AI that present a concise and customizable overview of what’s changed for the patient since they were last seen, enabling the provider to catch up efficiently and focus more on the patient.

Some information that is important to providers can be complicated to find. Some patients communicate frequently with their providers via the EHR's messaging tools and essential clinical information can be buried in these messages. In particular, medication changes and updates may occur in these exchanges, but those changes may not always be formally recorded in the patient’s medication history. We have developed a tool to identify these message changes, compare them to the patient’s medication list and surface any discrepancies with the provider so that the history can be updated without needing to track every interaction.

Chart review and updates for direct patient care are not the only time-consuming tasks; they are manual work in which clinicians are engaged. At URMC, we are also tackling the issue of extracting data from the EHR for submission to clinical and research registries. In many cases, this process involves extensive manual review of patient charts and data entry into registry forms, which can take an hour or more for each case entered. To alleviate this burden, URMC is developing generative AI tools to augment and sometimes automate the process, allowing nurses to focus more on patient care and quality improvement rather than rote administrative tasks.

The ability of generative AI to unlock and make sense of unstructured data is not just a technological advancement; it represents a paradigm shift in how we approach healthcare. We now have the tools to turn the wealth of data we've collected for centuries into actionable insights that can improve patient outcomes, enhance operational efficiency and ultimately transform healthcare delivery. In the face of a crisis of clinician burnout, generative AI can significantly reduce the burden of documentation and administrative tasks, providing much-needed relief to healthcare providers.

The journey of data in healthcare reflects our ongoing quest to capture and utilize information better. With generative AI, we are finally able to harness the full potential of this data, unlocking critical insights that were previously hidden and bridging the gaps that have long plagued our healthcare system. As we continue to develop and refine these AI tools, the future of healthcare looks brighter, more connected, and more effective than ever before.