Sometimes, it pays to be number two.

Beth Israel Lahey Health has long been the state’s second-largest hospital system, the second biggest employer, the second most fill-in-the-blank. When compared to Mass General Brigham, owner of Massachusetts’ two biggest hospitals, Beth Israel’s place in the pecking order was always clear.

But Mass General Brigham has faced a number of challenges in recent years, from losing its relationship with Dana-Farber Cancer Institute to a unionization campaign by its primary care physicians amid discontent among wide swaths of its doctors. In many ways, Beth Israel has benefited from its competitor’s issues: it’s struck its own collaboration with the renowned Dana-Farber and has recruited dozens of primary care clinicians — one-third of whom came from MGB alone in the last year.

Continue to STAT+ to read the full story…

© Ben Pennington for The Boston Globe

arXiv:2511.20680v1 Announce Type: cross Abstract: Despite high performance on clinical benchmarks, large language models may reach correct conclusions through faulty reasoning, a failure mode with safety implications for oncology decision support that is not captured by accuracy-based evaluation. In this two-cohort retrospective study, we developed a hierarchical taxonomy of reasoning errors from GPT-4 chain-of-thought responses to real oncology notes and tested its clinical relevance. Using breast and pancreatic cancer notes from the CORAL dataset, we annotated 600 reasoning traces to define a three-tier taxonomy mapping computational failures to cognitive bias frameworks. We validated the taxonomy on 822 responses from prostate cancer consult notes spanning localized through metastatic disease, simulating extraction, analysis, and clinical recommendation tasks. Reasoning errors occurred in 23 percent of interpretations and dominated overall errors, with confirmation bias and anchoring bias most common. Reasoning failures were associated with guideline-discordant and potentially harmful recommendations, particularly in advanced disease management. Automated evaluators using state-of-the-art language models detected error presence but could not reliably classify subtypes. These findings show that large language models may provide fluent but clinically unsafe recommendations when reasoning is flawed. The taxonomy provides a generalizable framework for evaluating and improving reasoning fidelity before clinical deployment.

“The irony’s not lost — going from the physician to the caregiver to the patient,” Michael Zinner mused in June.

He was speaking by phone from his Florida home about the unusual and heart-rending way his professional life and personal life had intertwined through pancreatic cancer — his specialty as a surgeon.

There were the patients he operated on — seemingly too many to count in his years as a top surgeon, including 21 as chief of surgery at Brigham and Women’s Hospital.

Continue to STAT+ to read the full story…

© Baptist Health South Florida