����ֱ��

Subscribe on: | | | | |

On this episode of The Digital Patient, Dr. Joshua Liu, Co-founder & CEO of ����ֱ��, and colleague, Alan Sardana, chat with Dr. Juan "JC" Rojas, Associate Chief Medical Information Officer at Rush University Medical Center, about "Where AI Performs Best in Clinical Care Today, Evaluating AI like a New Drug or Device, Using AI to Identify High-Risk Patients When Data is Incomplete, and more..." Click the play button to listen or read the show notes below.
‍

Audio:

‍

Guest(s):

• Dr. Juan C. Rojas, Associate Chief Medical Information Officer at Rush University Medical Center
• Dr. Joshua Liu, Co-founder & CEO at ����ֱ��

‍

‍Episode 208 - Show Notes:

[00:00:07] Episode preview

[00:05:16] Why ICU medicine showed healthcare as a team sport powered by data, technology, and trust.

[00:05:44] How consumer AI like Netflix inspired Dr. Rojas to ask why predictive intelligence wasn’t helping critically ill patients.

[00:06:13] The fellowship project that sparked his informatics career: predicting ICU readmissions to prevent premature discharges.

[00:06:43] Why Dr. Rojas believes AI should function as “augmented intelligence,” supporting clinicians—not replacing them.

[00:08:05] Where AI performs best in healthcare: short-term, physiologic predictions like sepsis and ICU deterioration.

[00:08:51] Why human-driven outcomes like readmissions and length of stay break many predictive models.

[00:10:10] The case for evaluating AI like a drug or device—measuring real-world value, alert burden, and workflow impact.

[00:11:49] The hidden barrier to health equity analytics: inconsistent definitions of race, ethnicity, language, and gender.

[00:12:45] Why health system averages can hide disparities—and how stratified data reveals where care is failing.

[00:14:36] Inside Rush’s AI governance playbook: silent testing, local validation, and equity checks before deployment.

[00:17:34] How Rush uses AI to identify high-risk patients for social needs screening when data is incomplete.

[00:19:24] Why bias and/or imperfect data should not halt progress, as long as health systems actively work to improve fairness over time.

[00:22:37] How ignoring data bias is like skipping hand hygiene.

[00:26:57] Attribution as a major challenge in analytics and why agreeing on definitions before surfacing performance data is paramount.

[00:28:48] The informaticist as a bicycle frame between clinicians and data teams, translating needs across clinical and technical worlds.

[00:33:34] Recounting a failed AI rollout where overlapping scores created confusion, showing how trust and clarity matter more than model accuracy.

[00:37:37] Why AI adoption often fades over time, citing alert fatigue, workflow disruption, and unnoticed system changes that break models.

[00:43:22] Principles for localizing AI models, including silent testing, predefined success thresholds, and strong clinical champions.

‍

Fast 5 Lightning Round:

1. What is your favorite book or book you’ve gifted the most?
   Not a book, but the Acquired podcast.
2. If you could instantly master any skill, what would it be?
   "Becoming a full-stack software engineer capable of building production-grade software."
3. Would you rather have Super strength, super speed, or the ability to read people’s minds?
   Mind reading.
4. What is something in healthcare you believe others might find insane?
   That healthcare data rights should belong to patients.
5. What is the last movie or TV show you saw?
   The Diplomat.

‍

The Digital Patient has been recognized as Feedspot's . Thank you to our listeners for making this happen!