by Angela Ferrante, UVM Larner College of Medicine
Clinical visits are set to change as AI quietly transforms health care. Ambient recording technology, which captures the details of patient-clinician conversations, has the potential to streamline care, reduce clinician workloads, and improve patient outcomes—particularly in rural areas.
However, this shift also presents several challenges. A recent paper co-authored by Dr. Robert Gramling of the Larner College of Medicine, examines the potential benefits and risks of using AI for clinical recordings.
In their recent publication, Gramling and colleagues from Dartmouth University address three key concerns related to AI’s role in the health care setting: burden, fairness, and commoditization.
While AI can ease the documentation burden on clinicians, it may also lead to increased patient loads, which could affect the quality of clinician-patient interactions. As demand for health care services grows faster than the supply of clinicians, time saved through automated documentation might quickly be used up by an influx of patient visits.
Although AI may improve short-term productivity and accessibility, it risks reducing the human element in care. Such a situation could lead to “automation bias,” where clinicians rely too heavily on AI-generated outputs in high-pressure situations.
To counter this, the team suggests using explainable AI (XAI) to provide clear insights that enhance shared decision-making.
Algorithmic bias is another critical issue. The authors emphasize the need for diverse data collection, patient involvement, and regular bias monitoring to ensure fair AI use in health care. They point out that speech data contains important nuances beyond words, such as accent, tone, and inflection, which affect meaning.
To address bias in AI and encourage fairness, the authors recommend three strategies: increased focus on protecting patient information, identifying and correcting biases in training data, and adopting an “ecological” approach that considers the complex nature of conversations and their contexts.
By implementing these strategies, Gramling believes scientists and doctors can create AI systems that are both technically sound and culturally sensitive.
Despite these challenges, Gramling is optimistic about AI’s potential to improve patient care and clinician workflows. He emphasizes the importance of including diverse perspectives—such as those of patients and clinicians—in the design and implementation of AI technology.
“Improving communication in health care is essential for 21st-century medicine,” he states. “Open recordings offer the chance to understand what actually happens in clinical conversations, helping patients feel heard and understood.”