Artificial intelligence
Generative AI doesn't always save time and money
Readers, you are not going to believe it, but many generative AI tools being rapidly developed for health care are still pretty rough and not quite ready for patients.
A series of recent research papers by academic hospitals has revealed significant limitations of large language models (LLMs) in medical settings, undercutting common industry talking points that they will save time and money, and soon liberate clinicians from the drudgery of documentation, STAT's Casey Ross reports.
"Doing this research we start to understand the failure modes," said Danielle Bitterman, a radiation oncologist at Mass General Brigham who co-authored one paper told Casey. "If we roll this out too soon, we will see early errors and people will get scared of it."
Read more here
Hospitals
HCA doubles down on AI for clinician notes
HCA Healthcare, the largest for-profit hospital chain in the United States, is planning to expand the use of an artificial intelligence tool to document doctor-patient interactions in its emergency rooms, STAT's Brittany Trang reports.
The plans for a broader roll out of the AI tool across its network comes nearly a year after the 184-hospital health system started working with medical documentation company Augmedix to pilot its ambient scribe technology at a handful of ER departments within the HCA network.
HCA's plans are part of growing efforts by health care providers across the country to lean on AI to reduce clinician burnout by shifting the burden of repetitive administrative tasks like medical note-taking to technology. The move comes at a time when federal agencies are still writing rules to make sure that the use of AI-based tools in clinical settings do not harm patients.
Read more here
Research Telehealth experiment for PTSD gets funding boost
Here's an update that answers the question: How do we scale promising digital health research?
A few years ago, I wrote about a remarkable study in which researchers at the University of Washington used telemedicine to beam specialized mental health treatment to people with bipolar disorder and post-traumatic stress disorder who lived in communities with limited access to mental health care. The comparative effectiveness study funded by the Patient-Centered Outcomes Research Institute randomized patients to either a referral group or a collaborative care group. The referral group was connected over video to a psychiatrist and psychologist for a standard course of treatment. Meanwhile in the collaborative care model, a remote psychiatrist evaluated a patient and then supported treatment delivered by care managers in the local primary care setting. Collaborative care has been around for decades, but it's mostly used in the context of anxiety and depression.
The results were positive and significant in both treatment groups, but the collaborative care group used less of the costly and hard to come by psychiatric care. In my story at the time, I pondered how this innovative, but somewhat complex, care model might be scaled so more people could benefit.
Well, PCORI just announced a $2.5 million grant to the University of Washington's AIMS Center and collaborative care company Concert Health to fund implementation of the research at 150 primary care practices in medically underserved areas. The funding will support training staff on how to adapt collaborative care to the more serious mental health conditions, as well as to help Concert develop its technology for the new model.
No comments