Health Care AI Requires Human Oversight
In a burgeoning landscape where artificial intelligence (AI) is becoming increasingly commonplace in health care settings, a crucial question looms large: How can we ensure that these high-tech systems operate effectively and ethically without breaking the bank?
Decaying AI Algorithms During Covid-19
Recent research conducted by Ravi Parikh, an oncologist at Emory University, revealed a troubling trend. An AI algorithm designed to prompt doctors to discuss critical treatment and end-of-life preferences with cancer patients suffered a significant performance decline during the Covid-19 pandemic. The algorithm, initially a reliable tool for physicians at the University of Pennsylvania Health System, experienced a 7% decrease in accuracy, potentially leading to missed opportunities to provide essential care when it was needed most.
Monitoring AI Performance: A Challenging Task
The study’s findings underscore a broader issue facing hospitals and health care providers: the necessity of consistent monitoring and maintenance of AI systems. Computer scientists and medical professionals alike acknowledge the importance of diligent oversight to prevent potential algorithmic failures that could compromise patient care. However, as the technology continues to evolve and proliferate, hospitals and institutions find themselves grappling with the practical challenges of ensuring AI reliability and effectiveness.
The Dilemma of AI in Health Care
As Nigam Shah, chief data scientist at Stanford Health Care, aptly summarizes, the promise of AI in improving access, capacity, and quality of care is undeniable. However, the escalating costs associated with implementing and managing these sophisticated systems raise critical concerns about their long-term viability. With hospitals struggling to allocate resources for rigorous AI validation and oversight, the sustainability of AI-driven health care remains a pressing issue that demands thoughtful consideration and strategic solutions.
