Informatics Grand Rounds with Dr. Leslie Lenert

Why It Matters

Understanding the true impact of computer‑mediated care versus AI hype is essential for designing safe, unbiased workflows that improve detection of hidden health issues while preventing new inequities.

Key Takeaways

• Early computer-patient interactions improve disclosure of stigmatized issues.
• Kiosk-based IPV screening in EPIC increased detection tenfold.
• AI outperforms physicians in many tasks; human+AI helps top performers.
• AI-generated papers risk provenance loss and embed hidden biases.
• Predictive models reflecting wrong values can worsen health disparities.

Summary

In this Grand Rounds, Dr. Leslie Lenert reviews the evolution of computer‑patient interfaces, from Warner Slack’s early experiments to today’s e‑health and AI‑driven tools, and examines how these technologies reshape clinical workflows.

He highlights that patients disclose sensitive information—psychiatric symptoms, sexual behavior, drug use—more readily to a computer than a human, a phenomenon independent of AI. Using this insight, his team built an EPIC‑based kiosk that lets women aged 18‑49 complete an intimate‑partner‑violence (IPV) questionnaire privately. In a stepped‑wedge trial of 9,000 women across 15 clinics, overall screening rose and the kiosk mode produced a ten‑fold increase in IPV detection, identifying about 60 at‑risk patients.

Lenert then contrasts the historic “fundamental theorem of informatics” (human + computer > human) with recent evidence that large language models (LLMs) often surpass physicians alone, and that human‑AI teaming only benefits the highest‑performing clinicians. He notes a surge of 3.2 AI‑vs‑human comparison papers per day, and warns that AI‑generated images and manuscripts embed hidden value judgments, as illustrated by a biased IPV screening illustration and a predictive expenditure model that under‑serves minorities.

The talk urges clinicians and administrators to distinguish genuine informatics workflow gains from AI hype, to safeguard data provenance, and to align predictive models with equitable health values. Without such safeguards, rapid AI adoption could amplify bias and erode scientific rigor.

Original Description

This session with Dr. Leslie Lenert will discuss a Human-Led AI-driven learning health system. #healthai #johnshopkins #healthinformatics https://bids.jhmi.edu/about/events-also-past-grand-rounds/

This session covers:

-How the rise of artificial intelligence is changing concepts of the Learning Health System

-The threats to science and the potential for dehumanization of medicine through AI

-How we can enhance human-AI collaboration in a learning health system through living laboratories and digital twin models