Improving Lean Methodology and Clinical Analysis at UCSF and ZSFG

If there's a tool that can extend your workforce like AI, we don't have to have five different patient navigators looking at all the charts and doing chart review.

We can surface bottlenecks to administrators to help them solve it with a click of a button now.

If the AI is not solving a real problem that you've defined, it's probably not the right tool for you.

Hello, and welcome to Healthcare AI Pioneers.

I'm your host, Dr. Jesse Pines.

This is a podcast that explores the latest developments, innovations, strategies, and trends in healthcare AI, with a focus on what clinicians, administrators, hospitals, and health systems are actually implementing and how that's going in the real world.

In this episode, we're going inside Zuckerberg San Francisco General Hospital, the safety net academic medical center at the heart of San Francisco's public health system, to explore an ambitious and novel application of AI in hospital operations, specifically embedding AI directly into performance improvement work to tackle inpatient length of stay.

The team has combined classic lean methodologies, the structured iterative approach to operational improvement borrowed from manufacturing, with an LLM that reads unstructured clinical data, the notes, the narratives, and the free text that structured databases can't capture.

The model identifies gaps in the care continuum, suggests countermeasures, and predicts how many length of stay days a given initiative might save.

In other words, it's an AI tool designed to help a resource-constrained hospital prioritize where to focus its improvement energy.

Today, I'm joined by two people who are leading this work.

But first, as always, we begin with a recent headline.