Getty Images/iStockphoto

How generative AI in healthcare is helping cut admin burden

Over 90% of healthcare workers feel optimistic about the promise of generative AI in healthcare to alleviate administrative burdens, according to a new survey.

Administrative burden is a widespread issue across the healthcare industry, driven by the rising demands of healthcare documentation and regulatory requirements. However, generative AI in healthcare could be part of the solution.

A new survey conducted by the Harris Poll and Google Cloud reveals that over 90% of healthcare workers feel optimistic about the potential of generative AI (GenAI) to alleviate administrative burdens.

In a recent press conference, Aashima Gupta, global director of global healthcare solutions at Google Cloud, shared results from the survey and led a panel of subject matter experts in a discussion about the promise of GenAI in healthcare.

Drafting bedside shift reports

Michael Schlosser, M.D., MBA, senior vice president of care transformation and innovation at HCA Healthcare, underscored the connection between administrative burden, provider burnout and care quality.

"Doctors and nurses and other care team members all live in this world of all of these additional administrative tasks that take them away from their day-to-day work of focusing on patients and providing care," Schlosser said. "Our thesis has been the more administrative burden we remove, the higher the quality and the higher experience the care will be."

HCA Healthcare has tapped GenAI for several administrative tasks, including bedside shift reports. Also known as nurse handoffs, these reports allow outgoing nurses to communicate essential patient information to the incoming shift to ensure continuity of care.

"We partnered with the Google research team because this is not a simple use case," Schlosser said. "This isn't just about creating a summary of the chart; that really doesn't achieve the goal. It's much more nuanced than that."

Instead of summarizing patient EHR data, the project focused on developing a model that could think like a nurse, identifying key items that need attention and summarizing critical events while highlighting essential information for the next shift.

We believe that GenAI and AI overall is transforming how healthcare professionals access and use information to make powerful decisions confidently.
Helen WatersExecutive vice president and COO, Meditech

"We had to teach a model and create a structure that would sit around it, enabling it to understand what key items need attention and what the critical summary of events is," Schlosser explained. "This way, the next shift knows exactly what to focus on to ensure continuity in care delivery."

In the pilot group, 90% of the nurses rated the model as helpful enough to replace their current clinical documentation process for handoffs.

According to Schlosser, end-user input is key to the model's success.

"We collected extensive data from our nurses, who constantly provided feedback on the model's performance," he said.

The team is scaling the model and aiming to implement it in five hospitals by the end of 2024.

"We'll learn from this cohort and make necessary adjustments before a broader rollout in 2025," Schlosser stated.

Automating prior authorization

Prior authorization is one of the largest administrative burdens the industry faces today. According to a 2023 AMA survey, healthcare staff dedicate approximately 12 hours each week to completing prior authorizations, with some employees working solely on these tasks.

What's more, 94% of physicians report that the prior authorization process has caused delays in accessing necessary care.

Highmark Health, a nonprofit healthcare company and integrated delivery network, has automated about 30% of its prior authorizations using generative AI.

"Once you automate the authorization process, a lot of the process-related issues that lead to denials start going away rapidly," said Tony Farah, M.D., executive vice president and chief medical and clinical transformation officer at Highmark.

In addition to automating prior authorizations, Highmark has implemented a utilization management strategy known as gold carding, which exempts providers with a history of efficient, high-quality care from some prior authorization requirements.

Overall, gold carding combined with prior authorization automation has reduced provider administrative costs by 85%.

"Once they see the patient or interact with a patient, the provider is able to achieve this approval process within seconds versus days or weeks sometimes, which has a negative impact on patient care," Farah explained.

Summarizing health records

Several clients of EHR vendor Meditech have adopted an AI tool to explore and summarize medical records.

"It's a very successful endeavor, generating a lot of interest from physicians, nurses and other users," said Helen Waters, executive vice president and COO of Meditech. "They feel they can quickly navigate dense records and identify critical information for treatment, protocols or prescribing."

Similar to HCA Healthcare, Meditech has also piloted generative AI to draft documentation for nurse handoffs. The summaries follow the SBAR (Situation, background, assessment, recommendation) format, which nurses use widely for care transitions.

"We believe that gen AI and AI overall is transforming how healthcare professionals access and use information to make powerful decisions confidently," Waters remarked. "Clinicians come to this field to make a difference in the lives of patients. Feeling confident in their decisions and not seeing technology as a barrier is critical as we look to the future."

Meditech is exploring future generative AI use cases for advanced directives, prior authorizations and claims processing.

"Our work with Google has yielded great results, and we look forward to continued success and learning across institutions," she said.

Hannah Nelson has been covering news related to health information technology and health data interoperability since 2020.

Dig Deeper on Clinical documentation