If AI couldn’t clear your desk Wednesday so you could attend this compelling Healthleaders webinar featuring three industry experts — yet you’re curious to know what’s up on the healthcare AI front — here are my key takeaways:
- “AI helps clinicians operate at the top of their license and employees to operate at the top of their skill set,” espoused Jared Antczak, Chief Digital Officer of Sanford Health. ROI is important, but not the only consideration. AI can assist, augment or automate tasks to improve quality of care, population health, patient experience and employee experience, while reducing costs.
- “The real key for any AI initiative is not AI for AI sake,” stated Tim Boomershine, VP of Data Science at Waystar. Explaining the importance of problem definition before solution, he summarized key AI usage goals of cost reduction, patient satisfaction and improved clinical outcomes.
- “AI is not a strategy in and of itself, it’s a tool to leverage technology,” agreed Antczak, espousing the importance of defining the challenge before employing a solution such as AI. He recommends reframing the industry mindset from “sick care” to “healthcare,” utilizing AI to help prevent disease further upstream. AI applications at Sanford Health include algorithms identifying higher risk for colon cancer and kidney disease, predicting and forecasting nursing staffing needs and saving hundreds of work hours by removing the “sea of administrative overhead.”
- “The biggest challenge is getting AI in the hands of the right people at the right time,” claimed Albert Karam, Vice President of Data Strategy Analytics, Parkland Center for Clinical Innovation. Lamenting the patient information disconnect among providers due to various EMRs, he anticipates beneficial nationwide interoperability, which AI will assist in making a reality.
- “Trust is a big factor,” affirmed Antczak. It’s critical that AI technology is perceived as a benefit, not a threat, especially in the risk-averse environment of healthcare. Hence, he brings all the right people to the table at project inception, including clinicians and consumers. Robust data governance, including ongoing model training and testing, also assures trustworthy accuracy.
- “Most people don’t realize it takes a significant amount of time to create a new AI model,” asserts Karam, outlining the stages of model creation, silent mode and piloting, totaling a 1-to-1.5-year timeframe. Boomershine added that the successful launch of an AI model involves a significant amount of time, effort and dollars, involving IT, administrative and clinical minds.
Most importantly, webinar experts chimed in on the critical importance of keeping humans in the loop. “Augmentation takes mundane, repetitive tasks out of the system and lets experts deal with the things that challenge them,” maintained Boomershine. “Alert fatigue is a real thing. AI helps support, but the doctor or nurse is the ultimate decisionmaker,” explained Karam. “AI is about co-piloting vs. auto-piloting, complementing clinicians to help them operate at the top of their license,” avowed Antczak. In the end, the experts agree AI is here to stay but successful implementation will not come from pushing an “AI mindset.” Rather, it’s up to the AI innovators to make AI a welcome, familiar addition to current user workflows.