In an AMCP 2024 panel discussion, the current and future impact of artificial intelligence (AI) on managed care across health care management and delivery systems was explored through detailed case studies and examples to identify practical implementation and strategies supporting patient care and operational efficiencies.

Health care must be analyzed at a local level and supported by technology. Once an abstract sci-fi concept, AI is now a tangible tool. The panel stressed that the time to think about and incorporate AI is now.

“At the end of the day, we're all in the business of trying to make the best decisions for patients, and to do that you need accurate information about what medications people are taking,” said Sean Jeffery, PharmD, BCGP, FASCP, AGSF, Professor of Pharmacy Practice, UConn School of Pharmacy, Director of Pharmacy at Hartford HealthCare Integrated Care Partners. “All the work that we do in plan design, or managing patients through protocols, or health systems best practices doesn't matter if somebody can't access, take, or figure out how to use their medications.” Many issues exist in the current medication reconciliation process. Standard approaches are often labor-intensive, and organizations won’t provide financial support to expand them. When data is found, it often isn’t acted upon, and when it is, action taken could be ineffective at best and cause patient harm or readmissions at worst. Too often, available electronic health record (her) tools fall short.

He shared a story from his work in Connecticut tracking medication reconciliation, where he met a patient who had found a unique system for ensuring his medication adherence: Cool Whip lids. The gentleman laid out his pills on the plastic lids on his kitchen counter as if they were clock faces so that he didn’t miss any of his AM and PM pills. Next to the pills, he also kept a wooden court mallet and a large pair of pliers. “Anybody want to guess what the gavel and the pliers are for?” asked the panelist. “The gavel was his pill crusher and the pliers were his pill splitter. Pretty creative! I'd like to take that creativity that he has and try to play that forward now into the role of AI and what we might be able to do with it.”

The panelist shared health care AI-related insights from a recent conference with about 100 local experts such as state officials, health system leaders, data scientists, lawyers, pharmacists, physicians, social workers, nurses, and more.

“One of the takeaways is as we all start approaching the deployment of these new tools, we have to be mindful of testing. Testing is going to be very important so that you understand the veracity of what your model is able to do,” he said. “And the testing leads to trust. Ultimately, for us to be successful, we as developers and designers of these tools have to trust what we're developing. Our employees have to trust what they're going to be using, and our consumers have to trust the results.”

The panelist was introduced to an AI tool in 2023 that might help alleviate the sometimes painstaking process of monitoring and verifying a patients’ medication usage.

An important part of understanding AI is knowing that AI relies on a constant diet of new data in order to train itself and constantly improve, shared the panel. However, AI is consuming provided data at an alarming rate. New sources of data are being explored for potential use, including Youtube videos and AI-generated content. Sourcing quality training data is going to be essential in the future of AI’s use in managed care.

“Your data is what's going to differentiate a generic AI application from something that is built and best suited for your business,” said panelist Jay Rajda, MD, MBA, Physician Executive of Global Healthcare at Amazon Web Services, who introduced AI cloud services available and in development at Amazon AWS.

“As a cloud services provider, these compute resources, storage, databases, things like that are available over the internet,” he said. “There is a significantly improved security in the cloud. This is the same infrastructure that militaries and banks use.”

AWS may be a strong contender for supporting advancements in health care. The panelist shared that they seek to enable access and delivery of person-centered health care, driving improved outcomes at a lower cost, and accelerating the digitization and utilization of health care data. The panelist highlighted that performance, resilience, and security are all necessary ingredients in creating innovation.

In a case study, intelligent health care document processing was used to process insurance claims, lab reports, clinical notes, and medical imaging reports. This tool funneled data into Amazon Textract, Amazon Comprehend Medical, and AWS HealthLake. Amazon Textract extracted information from structural and unstructured medical documents. Amazon Comprehend Medical linked to standard medical ontologies like ICD-10-CM, RCNorm, and SNOMED CT. AWS HealthLake, a successor to Amazon HealthLake, created a complete view of a patient's medical history.

Another successful case study was shared where Elevance Health enabled intelligent claims processing using AWS ML services. Elevance Health was troubled that it took an average of 20 minutes per claim to manually extract sensitive information from claims forms. The solution? Amazon Textract digitized and automated the claims process using optical character recognition to extract data from claims forms and applies machine learning to index and classify each document.

The benefits reaped were clear. The new process extracts and digitizes data to quickly process thousands of claims each day. Now, 80% of the claims-processing workflow is automated and is expected to reach 90% or higher.

An exciting potential use for AWS-managed services is timely fraud detection. The idea shared would build custom fraud detection algorithms using Amazon Sagemaker, a service that allows nonspecialist data scientists to build, train, and deploy high-quality machine learning models for any use case with fully managed infrastructure, tools, and workflows.

The panel also discussed the option to rapidly deploy a fraud detection solution using Amazon Fraud Detector–a fully managed service that enables the deployment of customized fraud detection solutions based on historic data without writing code. The prospect of procuring a ready-to-deploy fraud detection solution from one of AWS’ many trusted partners is also appealing.

In another case study, Sun Life used SageMaker to successfully accelerate the launch of new fraud detection capabilities. Sun Life, a Canadian financial services company that provides health and life insurance and asset management services to individuals and corporations, found themselves struggling with disparate data sets in legacy systems that led to delays in the timely introduction of new capabilities. As a solution, they migrated their on-premises Hortonworks/Teradata environment to a scalable data lake platform using Amazon S3, Amazon EMR, Amazon Glue, AWS Lambda, and Amazon SageMaker.

Because of this case study, the data lake has helped launch a new fraud recommendation engine in Canada. They reduced the time spent acting on fraud leads from months to days and this has led to next best offer capabilities in Asia.

Other successful case studies were discussed including the use of integrated, human-centered health care solutions being utilized by Cambia and the use of AWS supporting MetroPlusHealth to reach out to at-risk people.

To operationalize AI to create value, Shawn Wang, Chief AI Officer at Elevance Health, explained that we must empower stakeholders to deliver better experiences. This means more time for associates to exercise creativity and potential projects, improved member experiences, and giving providers more time with patients.

Generative AI (GenAI) seeks to automate call summaries and offers real-time sentiment analysis and issue resolution insights. It also is a knowledge management tool for agents to address over 200 types of member queries.

GenAI can be leveraged in a 3-tier approach. The largest tier, tier 3, is general capabilities like GenAI for associate use. The second tier is custom capabilities and touches on business-specific solutions. The final and top tier is transformational and includes game-changing strategic initiatives.

The 3-tier approach requires a “recipe” that responsibly integrates AI into the fabric of a business to create impact at scale. AI must be run as a business; this means knowing your customer base, creating both top-down and bottom-up alignment, and moving a business from the passenger to a driver's seat.

Success also hinges on focusing on the critical few. This means focusing on areas where AI has the most opportunities, seeking simple solutions, and addressing the first-mile and last-mile challenges. Then, AI solutions can be managed as products. AI must become a team sport with multi-disciplinary players to create a supply and demand loop from innovation to scale. This workflow integration is key to creating value.

In the closing panelist discussion, Joseph Honcz, RPh, MBA, President at C4i, discussed the process of transitioning from a company policy against open-source AI models to using AI tools effectively.

“When the entire company has access to a model and they're able to start to explore in a safer environment, it's like the training wheels to figure out how to make these work,” said Dr Jeffery.

While the panelists acknowledged that predicting AI development is difficult, the rate at which AI is projected to evolve is shocking. Since AI models have been tested and passed medical licensing exams, the panel host asked if clinicians should be worried about being replaced by AI.

“I don’t see a complete replacement of any clinicians with AI,” said Dr Rajda. “What I do see are tremendous opportunities for impact…it will make clinicians much more productive and efficient and allow them to practice on the top of their license. What's going to happen, I believe, is not AI replacing clinicians, but clinicians who use AI will be more effective and therefore replace clinicians that don’t.”

Reference

Jeffrey S, Rajda J, Wang S, Honcz J. From theory to practice: a panel discussion on deploying artificial intelligence in managed care. Presented at: AMCP 2024; April 17, 2024; New Orleans, LA