A new white paper explores how chief information officers are making decisions about AI, a rapidly evolving technology with a potential for tremendous impact on the business of healthcare.
While the vast majority say their AI strategic plans are still in development, most are turning against the prospect of building algorithms in-house and are turning to their electronic health records or third-party vendors to supply solutions.
AI operations company Qventus surveyed chief information officers and chief medical information officers at medium and large health systems and completed one-on-one interviews to develop the insights laid out in the company's new white paper.
The white paper explores key issues facing health systems around the adoption of AI, including choosing whether to build AI in-house or buy a solution, how to choose a vendor and how to govern AI.
Qventus CEO Mudit Garg said the company’s trusted status with its customers and in the healthcare ecosystem allowed CIOs and CMIOs to speak candidly about their thoughts on AI and where their organizations stand on strategy and oversight.
Some of the experts surveyed for the report are Qventus customers, and the majority of CIOs and CMIOs quoted in the report are not named.
The white paper stresses that the goals of deploying AI in health systems aligns closely with the systems’ business objectives. Garg said AI could bring relief to a system that is facing declining reimbursements, higher operating costs and staff burnout.
The report concedes that health systems may struggle to find a pocket of cash to invest in AI. Garg suggested health systems find an AI use case with a strong return on investment, then continually reinvest the savings in AI to create space in the budget for the technology.
Two-thirds of CIOs surveyed said their AI strategies align well with overall business strategy. CIOs reported that their chief metrics for determining ROI for AI are improved margins (26%), cost reductions (24%) and staff productivity and clinician satisfaction (each 16%).
Most survey respondents said their organizations are still developing their AI strategies. Just 14% of those surveyed said that their strategies were comprehensive and well-defined, and 20% admitted their AI strategies were limited or fragmented.
Joseph Sanford, M.D., chief clinical informatics officer at University of Arkansas for Medical Sciences, said that in some cases AI is becoming a must-have to stay competitive.
“We are seeing these technologies move out of the toy box and into meaningful clinical application where there is a strong ROI,” Sanford said in an interview. “In some cases, we're seeing them move beyond bleeding edge kind of state-of-the-art technologies, where you have to make a significant investment to see the ROI, into almost becoming standard features where you are incomplete as a health system if you do not have some technology addressing this kind of a need.”
CIOs reported that they lack the internal expertise to build models, and many are turning to their electronic health record vendors to provide them with AI technologies.
The report says: “Many CIOs opt to move forward with their EHR, it’s what’s familiar and it’s already deeply embedded in workflows, but stagnation can quickly creep up, as they wait for the right technology solutions that are promised on the roadmap. In some cases when those solutions are eventually deployed, they don’t produce nearly the same results that a company specializing in AI could provide.”
Garg explained that flipping a switch to turn on an algorithm from an EHR does not guarantee outcomes. He contends AI technologies are more successful when the implementers of the technology understand the health system’s workflow and unique clinical data.
When survey respondents were asked to rank their primary considerations when choosing between EHR-based AI development and third-party vendors, the top considerations were functionality and features, integration capabilities, and cost.
CIOs stressed that having a clinical champion of any new AI product is critical to help with staff change management. One CIO mentioned that third-party vendors should have a business partner to help ensure clinicians and staff understand how to use the technology.
“We recognize we don’t have all the IT experts internally, but we can work with a vendor more in a partnership model, which means lower licensing costs and overall costs in terms of using their resources,” a CMIO of a large academic system said, as quoted in the report.
Health systems are opting to choose a few strategic AI partners and are more likely to continue buying new products from them than searching elsewhere.
When choosing an AI vendor, CIOs and CMIOs said cost comparisons, the size of a return on investment and speed to unlock ranked above customization, change management and time to implementation.
Health systems are likely to take a mixed approach, depending on the algorithm and the use case, Garg said. Sanford explained that health systems should focus on understanding the underlying problems at hand to decide how and if to apply AI.
“You spend 90% of your time trying to define problems really well so that you can choose the correct tool from your many options and apply them at exactly the right point in the operational workflow for best effect,” Sanford said. “AI is just another tool … but where you have well-defined processes and workflows that are amenable to the systems you get ideally something like an exponential return.”