Home Health Designing trust into genAI to maximize benefits for healthcare organizations

Designing trust into genAI to maximize benefits for healthcare organizations

by News7

Artificial Intelligence foundation models are evolving rapidly throughout the healthcare ecosystem. System integration plays an indispensable role in ensuring using generative AI results in safety, security and trustworthiness.

Further, having a domain-specific AI model integrate effectively and responsibly with the broader healthcare system is a critical element of ensuring a trusted AI environment.

Srini Iyer is senior vice president and chief technology officer at Leidos Health & Civil Sector. At the HIMSS24 Global Conference & Exhibition in March in Orlando, Leidos and Google will address the ongoing challenge of achieving trust and security with genAI by showcasing their collaboration on the Medical Pathways Language Model 2 (MedPaLM2), highlighting use cases to show the criticality of designing trust into genAI for maximizing the benefits to healthcare organizations.

We sat down with Iyer to get a sneak preview of his HIMSS24 educational session entitled “The Impact of Domain-Specific Models on Health AI.”

Q. What is the overarching focus of your session? Why is it important to health IT leaders at hospitals and health systems today?

A. Generative AI models represent a huge change in the field of AI. Specifically, the impact of AI on healthcare highlights the advantages and potential of using AI models trained on medical data for various tasks within the healthcare domain. This session will emphasize the potential of domain-specific AI to revolutionize healthcare by delivering more accurate, efficient and cost-effective care.

According to the June 2023 Gartner Healthcare Provider Research Panel Survey, a majority of the respondents (85%) believe AI large language models will have a significant to disruptive impact on healthcare, with 14% rating it a moderate impact.

There are several use cases of interest to health IT leaders at hospitals and health systems. Top among them are automated data analytics, document auto-generation, and EHR search and summarization. They should be interested in this topic for several reasons:

Improved accuracy and relevance. Healthcare domain-specific AI models, like Med-PaLM 2, are trained on vast amounts of medical data, enabling them to understand and respond to complex medical questions with greater accuracy and relevance compared to generic AI models.
Better patient outcomes. More accurate analysis of medical data can lead to faster diagnoses and better treatment plans.
Streamlined workflows and administrative tasks. AI can automate routine tasks, freeing up healthcare professionals to focus on important patient care.
Increased efficiency. Domain-specific AI models require less data and training time than traditional AI models, making them more scalable and cost-effective to implement. This can be particularly beneficial for smaller hospitals and health systems with limited resources.

In the next few years, more than half of the generative AI models used by enterprises will be domain specific, up from 1% today. Domain-specific AI can act as a valuable assistant to healthcare professionals, providing them with instant access to relevant medical information and insights, ultimately improving decision-making and patient care.

Q. What is one of the main learnings you would like your HIMSS24 session attendees to walk away with?

A. In a short period of a few months, with a small team, Leidos developed a successful Med-PaLM 2 Proof of Concept to validate trustworthy genAI in healthcare, demonstrating how trust and security can be seamlessly integrated into genAI systems to maximize benefits for healthcare organizations.

We selected a use case that focuses on the top three needs from healthcare provider executives. Medical professionals play a critical role in providing quality care, but their time is often challenged by administrative tasks like completing complex medical reports.

Agencies like the VA, SSA and CMS require detailed documentation, yet report generation places a significant burden on clinicians, impacting both efficiency and accuracy. The healthcare private sector also faces the same challenges.

We got better responses and our accuracy improved when we used vector store. These are ideal for generative AI applications because they allow one to search for relationships between unstructured data points and help LLMs remember these relationships over time.

There were challenges we encountered and addressed as we worked through this project:

Length and complexity. Reports can be extensive, requiring navigation through intricate sections and fields, demanding considerable time and attention.
Information overload. Clinicians may need to consult various sources and references to complete these reports accurately, often adding to the time burden.
High error potential. The sheer volume of information and complexity of sections can increase the risk of errors, potentially impacting patient care and reimbursement.

Q. What is another learning you would like session attendees to walk away with?

A. People with AI skills are hard to find and often expensive. Building generative AI skills within a company is a journey, not a destination. We were able to get our teams hands-on experience to learn skills involved around developing, training and deploying models.

We were able to collect many lessons learned along the way. AI platforms and tools are still maturing; applying these evolving tools to support your specific use case requires experimentation, in-depth knowledge and patience.

We had early access to some of these domain-specific models and we knew going in that documentation for these rapidly evolving tools was limited. Our developers had to work with product teams and go through an iterative process to determine the right path forward.

Having access to good data is critical to the success of these healthcare projects. This can be a big challenge for health IT, as we must deal with PII/PHI and HIPAA compliance. This limits the access to real-world data, which means we need to lean on synthetic data or de-identified data.

As an early adopter of implementing foundational models in healthcare, we are cautiously optimistic we can address some of our critical healthcare challenges to improve patient safety, resulting in better outcomes for our patients.

The session, “The Impact of Domain-Specific Models on Health AI,” is scheduled for March 12, 3:00-4:00 p.m. in room W208C at HIMSS24 in Orlando. Learn more and register.

Follow Bill’s HIT coverage on LinkedIn: Bill Siwicki

Email him: [email protected]

Healthcare IT News is a HIMSS Media publication.

Source : Healthcare IT News

You may also like