AI in Health Care Could Improve Accuracy, Efficiency, and Outcomes: Amrita Basu, PhD

Over the next 5 years, experts should focus on improving the speed, accuracy, and reliability of artificial intelligence (AI) in aggregating patient data, while also validating its potential to meaningfully advance cancer care, Amrita Basu, PhD, of the University of California, San Francisco (UCSF), told The American Journal of Managed Care^® at last month's San Antonio Breast Cancer Symposium.

Watch part 1 to learn more about the most impactful AI tools currently being used in breast cancer care and how they can be integrated into clinical workflows without adding burden or disrupting patient care.

This transcript has been lightly edited for clarity; captions were auto-generated.

Transcript

What concerns do patients and clinicians have regarding the use of AI tools in breast cancer care?

For clinicians, the safety, the hallucinations, or what we call false signals, these could be really detrimental. I think we're not at prime time yet to launch all of these things without humans really in the loop, again, really checking, verifying. How they do those checks and those verifications is something that we still are working out, and that's something I think the field has to really, really intentionally spend concerted effort doing.

For the patient, it's obvious that with these tools, such as ChatGPT, being public, is my health data going to be at risk?

The institutions themselves, for example, I'm at UCSF [University of California, San Francisco], we have a whole AI governance board. These institutions and health systems have a layer now that is kind of overseeing these efforts. In order for us to really become a safe environment for clinicians and patients, we have to be working with our leaders and these boards.

Looking ahead, what would an ideal roadmap for AI use in breast cancer care over the next 5 years look like?

I think, again, the text part of everything has exploded. The fact that we can potentially pull all these things in together, right? I work on patient-reported outcomes. Our patients are sending us surveys every week. I have that. I've got the notes, I've got the case study, I've got the mammograms, I've got the MRIs. So, it's all this, looking at all these things in a longitudinal fashion, and looking at the trajectory.

Being able to do that faster and better is really where I think we're moving towards in the next 5 years. Can we reliably use all this information and really kind of validate the promise of AI in cancer and in breast cancer? Is that going to be possible? I think that's the question, and that's what we will be trying to answer in 5 years.