AI is a big area of interest, but 79% of the leaders reported concerns about the possibility of data bias exacerbating disparities in health outcomes. DiLullo gave an example of a hospital in Tennessee that may use personal biomarkers in that demographic area to lead generative AI to make conclusions about diagnoses for cancer or Alzheimer's disease.

He explained that if you're building on a data set of a reasonably homogenous population, you can create biases against another population. An example of that could be West Memphis, which is largely an African-American population and may have completely different biomarkers.


"The truth is the bias is going to be there because you can't have it be perfect," he said. "Our view is to make it a diagnostic tool that informs and guides a radiologist, and they should always make the determination because they're the person that's trained."

Something that nearly all — 99% — of the leaders agree on is that there are opportunities for data-driven insights to improve care. The challenge is bringing all the data together from disparate sources — 61% citing a desire to improve interoperability between different platforms or healthcare settings.

One area of interest is population health analytics. Philips is helping hospitals use data science to start looking for population trends so they can shift their priorities of care to where it's needed the most.

"It's not mainstream yet, but it's the power of common interoperable data that we're pulling from our devices or other devices to create that ecosystem of patient-level data," said DiLullo.

Back to HCB News

Business Affairs Homepage