OpenRounds Editorial

Healthcare AI Daily Briefing

Thursday, October 8, 2026

New to OpenRounds? Get the free daily email or see how we work.

Medical Imaging

Fusion MRI model yields internal AUC, lower external AUC for small HCC.

Journal article

In a multicenter cohort of 296 lesions (209 training/internal test, 87 external validation), the interpretable fusion model achieved an AUC of 0.913 in the internal test set versus 0.791 for radiomics‑only. Although this advantage did not reach statistical significance in the external validation set owing to the limited external sample size, the model remains a promising noninvasive adjunct. [1]

Research & Evidence

MINT outperforms task-specific models and physicians on minute-scale forecasts.

Preprint

Electronic health record models are traditionally trained on years, days, or hours, lacking the minute‑scale resolution needed for bedside decisions. MINT, a minute‑scale model for emergencies, was pretrained on 766,733 visits from five health systems (10 hospitals, 16 years). It outperformed task‑specific models in 21 of 25 comparisons and physicians in forecasting respiratory‑support escalations, showing minute‑scale risk explanations. [2]

Operations & Workflow

AI aims to return two hours of admin time, improving clinician‑patient connection.

Podcast from Beyond the Chart: Exploring the AI Frontier for Oncology

The interview describes a study finding that clinicians spend two hours on administrative work for each hour of patient face‑time, and says AI tools that automate documentation could return that time, letting physicians give patients their undivided attention and strengthen the clinical connection. [3]

Policy & Governance

Bias found in Optum acuity program when cost used as proxy for sickness.

Podcast from KFF's The Business of Health with Chip Kahn

The interview describes that a guest researcher found bias in an Optum acuity program that used cost as a proxy for sickness, giving African-American patients systematically lower scores despite equal need. The bias emerged in a postmarket assessment, underscoring the shared responsibility of providers and manufacturers to monitor for such issues. [4]

Sources

Development and Validation of an Interpretable Machine Learning Model Based on Gd-EOB-DTPA-enhanced MRI for Evaluating Small HCC (≤2 cm): A Multicenter Cohort Study. · Academic radiology Original source
Learning the electronic health record at the minute-scale · medRxiv Original source
How Earlier Palliative Care Could Change Everything with Dr. Justin Baker · Beyond the Chart: Exploring the AI Frontier for Oncology Original source
AI: To Regulate, or Not to Regulate? · KFF's The Business of Health with Chip Kahn Original source

Fusion MRI model yields internal AUC, lower external AUC for small HCC.

MINT outperforms task-specific models and physicians on minute-scale forecasts.

AI aims to return two hours of admin time, improving clinician‑patient connection.

Bias found in Optum acuity program when cost used as proxy for sickness.

Sources

Get the free healthcare AI briefing