The Rise of AI. Well, Almost

AI is still in the early stages of development and application, but modest growth is evident, according to a new HealthLeaders Intelligence Report.

While people and processes are critical in applying analytics to healthcare industry data, the growing use of artifical intelligence (AI) solutions has the potential to change the equation. For now, AI is in the early stages of development and application, and its impact is still playing out.

According to healthcare leaders in the November/December 2019 HealthLeaders Intelligence Report, Investing For the Future: Analytics, AI, and ROI, AI use has been steadily growing, with 22% of respondents saying that their organization uses a software platform that provides an AI capability, up eight percentage points from our previous analytics survey in September 2017.

This modest growth is expected to continue going forward—for example, 31% of respondents say they don't currently have this capability but plan to within the next three years.

On the other hand, AI is not for everybody, and 25% say they don't plan to have this capability. However, this response is 10 percentage points lower than in our previous survey, indicating that fewer respondents are ruling out the use of AI capability at their organizations.

Respondents report that the types of data they use with AI are comparable to analytics use in general—the top three data types on which their organizations currently use or plan to use AI capabilities are clinical data (81%), financial data (72%), and patient data (59%).

AI use across the enterprise

According to Todd Stewart, MD, vice president of clinical integrated solutions and clinical informatics at Mercy Technology Services, the IT division of St. Louis–based Mercy, an integrated health system that includes more than 40 hospitals, 900 physician practices and outpatient facilities, and more than 45,000 employees, Mercy's AI path originated in operations, although the company is shifting its AI strategies and applications so they're not driven by departments, which further fragments information in silos. Instead, AI at Mercy is viewed as a tool to address organizational challenges across the enterprise.

"If you go five to 10 years back at Mercy and look at our pioneering work, advanced analytics came through our supply chain experience and some early partnerships in machine learning," says Stewart. "That work evolved into standardization of medical supplies and then standardization of process, but much more from an operations standpoint. But it became very clear early on that there were significant financial and clinical impacts for that work as well, par-ticularly as Mercy gained insight through analyzing which medical devices were producing the best patient outcomes at the lowest cost. The efforts resulted in saving $33 million in only the first three years while maintaining a high level of quality care."

Note that AI and machine learning are being used across the full scope of the healthcare industry and are not limited to enabling breakthrough, headline-capturing innovations like curing cancer or enabling precision medicine through the use of genomics.

For example, a recent machine learning initiative at Mercy focuses on the fundamental challenge of patients not showing up for their appointments.

"The project was initiated by one of our regional chief operating officers who was trying to improve the problem of patients not showing up for their appointments," says Stewart. "We have a shortage of providers, like everybody else in the industry, and we wanted to make sure that we're using our provider resources as efficiently as possible."

"The initial task was to determine whether we could predict with a reasonably high probability that a given patient was not going to show up for their appointment. Could we, in certain operational circumstances, then do extra things to try to contact that patient to try to ensure that they would show up or get them to reschedule? The project goal was to generate a list of probable no-shows and provide that to the operational people at the local level," he says.

"The MTS data science team analyzed the existing data across many different dimensions and did some complex modeling. And, in the end, they did a really good job at creating a model that runs daily and feeds this information on a regular basis to the local leaders, allowing them to take steps to better ensure that the patients that are scheduled will actually be there."

According to survey respondents, scheduling efficiency improvement is of growing importance. It ranks fourth on the list (30%) of responses for most promising areas of analytics development, and its response is 13 percentage points higher compared with the previous survey, where it was ranked sixth.

To download the full November/December 2019 HealthLeaders Intelligence Report, click here.