Healthcare providers are increasingly integrating artificial intelligence (AI) technologies into their services. However, the deployment of AI in healthcare systems is not devoid of challenges. A key aspect often overlooked by healthcare entities during the design phase of tech pilots is the monitoring of metrics, as highlighted by Bill Fera, Principal and Head of AI at Deloitte, in a recent interview.
Fera pointed out that many health organizations fail to identify which AI pilots to scale and which ones to abandon because of inadequate or absent metric tracking. “There’s a lot of languishing in pilots that are inherently not going to create value,” Fera said. He suggested that organizations should focus on use cases that can bring significant returns and establish appropriate metrics that align with these use cases.
Sympathizing with Fera’s views, David Vawdrey, Chief Data and Informatics Officer at Geisinger, emphasized the importance of designing a robust plan for evaluating tech pilots’ success. According to Vawdrey, a health system must initially understand the problem it is striving to solve before introducing an AI tool.
For Geisinger, patient care and safety are the prime outcomes to evaluate. The organization monitors the efficacy of algorithms used for procedures such as cancer screenings or flu complications based on the number of prevented hospitalizations, saved lives, and reduced spending. Vawdrey cautioned against the industry’s tendency to hastily introduce technology without a solid action plan, thereby risking ineffectiveness.
Thus, to devise an effective evaluation plan, health systems must identify their problem, know the outcomes they value most, visualize what success looks like, and comprehend the figures that will confirm if the tool is functional or not. If the tool fails to meet expectations, the health system must determine whether the issue arose from a strategic or execution problem, Vawdrey concluded.
For more details on this matter, read the full piece on MedCity News.