health records Patients want immediate access to imaging. But doctors say they need more time.
As federal data sharing mandates take effect, at least one medical specialty faces mounting anxiety as they weigh patients' call for quick, convenient health record access against their own need to carefully examine test results. At this year's Radiological Society of North America meeting in Chicago, Katie Palmer heard experts debate the merits of the rule and its impact on patients and clinicians.
Previously, radiology reports would be "embargoed" until clinicians had time to interpret the results, often rife with medical jargon; skipping that step could lead patients to become even more anxious if they don't know how to interpret them, or seek inappropriate care, some suggested. But patient advocates have long argued that immediate access is a step toward critical, potentially life-saving medical agency.
"There is danger in the immediate release of reports, even among educated individuals who are using the resources like Google or ChatGPT," says University of Virginia's Arun Krishnaraj. But it's clear that patients prefer immediate access, says CT Lin, chief medical information officer at UCHealth. Read more on radiologists' efforts to navigate those challenges here .
Artificial intelligence
Balancing AI's potential with concern for patients
Elsewhere at that meeting, Katie observed a pronounced tension between enthusiasm for artificial intelligence and skepticism. On the demonstration floor, behemoths like Philips and Microsoft hawked AI products they claimed could change the practice of radiology, but implementation in the industry is still limited.
"This is the third time we've given this program, and each time I think this will be the year where we start to talk about the role in the clinics as they dispersed out into community practices," Harvard Medical School radiology professor Constance Lehman said during a session on AI in breast imaging. "But in widespread routine clinical implementation, how we're using AI is still early." Read more of Katie's dispatches from the conference here.
research
Health systems should know which patients have died
A research letter in JAMA Internal Medicine highlights a troubling pattern in hospital record-keeping: Electronic health records don't always reflect patients' death. In an analysis of 41 clinics at one academic health system covering roughly 11,700 patients, almost 700 were marked deceased in a California state public health file but listed as alive in their electronic medical records — meaning that about 19 percent of deceased patients overall were deemed alive in their records. What's more, dead patients received more than 200 telephone calls and 300 portal messages after their death, all unrelated to their death, authors wrote.
"Not knowing who is dead hinders efficient health management, billing, advanced illness interventions, and measurement," authors wrote. "It impedes the health system's ability to learn from adverse outcomes, to implement quality improvement, and to provide support for families."
ChatGPT doesn't know much about drugs
OpenAI's generative AI tool ChatGPT isn't great at tackling drug-related questions, including about potential side effects or drug interactions — and health care workers should be cautious when using it, finds a new study presented at the American Society of Health-System Pharmacists midyear meeting this week.
For the study, Long Island University pharmacy practice associate professor Sara Grossman asked a free version of ChatGPT several questions that patients had submitted to Long Island University's drug information service. Only 10 of 39 were deemed satisfactory answers; others either didn't directly address the question, were inaccurate or incomplete.
No comments