Nashville, Tenn.-based HCA Healthcare, the nation's largest health system, continues to embrace artificial intelligence to improve its clinical and administrative operations.
Becker's recently did a Q&A with Mangesh Patil, chief analytics officer of HCA Healthcare, about the opportunities — and challenges — for AI at the 186-hospital system.
Question: What is a successful AI project you've been involved with recently at HCA Healthcare?
Mangesh Patil: It's tough to choose just one, but I'll highlight a few ongoing projects.
One standout project is Timpani, developed by HCA Healthcare's Care Transformation and Innovation team in collaboration with Palantir. When asked what their biggest challenge is, our nursing leaders identified the staffing and scheduling process. Unlike other industries, demand cannot be controlled in a hospital. Timpani tackles this directly, using staff input (unique preferences, skills and experience), system information, volume statistics and forecasting to generate automated, transparent and balanced schedules and provide intelligent staffing decision support once schedules are posted.
This technology provides better insight into labor, productivity and other critical staffing processes to help us make better forecasting decisions for HCA Healthcare's hospitals, helping to ensure we have the right clinicians at the right place at the right time to enable excellent care for our patients. Key to the Timpani development process was and continues to be elevating the voices of our nursing leaders to design this solution, one that meets their needs and enables them and their dynamic care teams to better serve patients.
We're also exploring large language and vision models to develop efficient and reliable workflows in documentation, shift handoffs, radiology, case management, supply chain, and revenue cycle. For instance, the nurse handoff use case was featured in the Google Next keynote last April. Although we see immediate value, we're ensuring careful adoption through our change management and responsible AI programs.
I'm excited and optimistic about the transformational potential of these use cases and the positive impact they will bring to healthcare.
Q: What is the biggest challenge or obstacle you've faced of late in implementing AI?
MP: Generative AI offers significant opportunities to transform healthcare, not just hospitals, but the entire ecosystem. Its ability to analyze unstructured text data — like physician notes, patient surveys, and call records — bridges the gap between intuitive and systematic workflows and allows us to look qualitatively beyond the coding systems originated in billing. Further combining this data with longitudinal patient records, images, connected devices, videos, floor plans, location tags, care team schedules, and talent profiles, gives us a comprehensive and multimodal knowledge graph for building interconnected and intelligent hospital systems.
However, there are a few challenges:
Data operations: AI models are only as good as the data they consume. Curating and maintaining high-quality data requires modern technology platforms and processes to bring multimodal data together from organizational silos into a centralized location where data scientists can discover, access, understand, and analyze it within an agile DevOps cycle.
Scale: We have some of the largest datasets in the healthcare industry to train AI, but building and operationalizing AI models for 186 hospitals, 2,400 ambulatory care sites, and 300,000 colleagues requires some serious AI engineering chops and production thinking. While we embrace a "fail fast" model through extensive prototyping, we never lose sight of the production scale. It also means saying "no" to unproven AI technologies until they meet our quality standards.
Q: What is unique about AI in healthcare compared to the other industries you've worked in?
MP: Before joining HCA Healthcare, I spent about 10 years developing AI for hospitality, media, and entertainment. I often joke about the contrast between hospitality and hospitals, but from a data science and analytics perspective, there's not a dramatic difference between the two. Both industries require operational efficiency, synchronization, and resilience while maintaining a high standard of care or experience for guests and patients. The math is quite reusable.
However, healthcare does come with added complexity due to its regulatory environment and the impact of decisions on human life. This can slow things down, but there's no comparison to the fulfillment that comes from improving human life. The impact of our work on people's lives is direct and, in my case, explainable to my kids.
It's also fascinating to see the intellectual curiosity and eagerness of clinical experts to use data and AI. For someone with deep expertise in their field, I find it remarkable how willing they are to embrace AI.