Can AI thrive in health care’s HIPAA-shaped box

Photo Credit: Getty

Innovation continues to transform health care, enabling us to live longer and healthier lives. And now artificial intelligence (AI) is poised to supercharge that progress. . This includes everything from helping us diagnose health conditions like strokes far earlier, discover potential new drugs for Parkinson's, and more universally, save countless work hours by auto-filling forms. Yet current federal regulations limit the power of many existing innovations and prevent the discovery of countless new ones.

Between regular check-ups, emergency room and hospital visits, and other medical-adjacent trips, Americans spend a considerable amount of time engaged in health-care-related activities. Despite heavy government investment in health care (24 percent of the federal budget), health outcomes still lag behind those of other nations that spend less. However, the US stands tall as one of the world's foremost leaders in health care innovation.

Much of AI-driven health care innovation is due to machine learning, which refers to a class of artificial intelligence models, like linear regressions, that learn from datasets. These models work best when given access to large quantities of high-quality data. Unfortunately, while we have an abundance of health care data, much of it is off-limits due to federal regulation.

Health care is an incredibly data-intensive industry, with the average hospital producing more than twice the data held by the Library of Congress annually. Though much of this data is collected due to health care regulation, some of it is valuable. And much of it is governed by the Health Insurance Portability and Accountability Act (HIPAA). Title II of HIPAA establishes stringent, strongly enforced standards to protect individually identifiable information.

Under Title II, covered entities (providers, insurers, and clearinghouses) and their business associates must gain patient approval or qualify for an exemption to use "individually identifiable health information" containing information like demographics. Deidentified data presents a way around these restrictions but requires either the blanket deletion of helpful data or an expensive and lengthy data manipulation process. In addition to imposing direct limits on what data can be used and how, HIPAA has created an environment where people are wary of sharing any data.

As a result, viable data is harder to obtain, making it hard to improve existing AI-based health care innovations and create new ones. And regardless of how or if a company finds a way around the problem - be it through qualifying for an exemption or finding a way to use deidentified data - the process inevitably creates extra costs that make many innovations economically unfeasible or result in higher prices for consumers.

What's the solution to this problem? It doesn't have to be outright abolishing HIPAA. We could reduce the regulatory burden it imposes by expanding the carveouts for people attempting to innovate in ways that will improve health care. This would tackle both the regulation themselves and the culture of fear that surrounds them. By implementing these changes, we can make health care cheaper and better for all without compromising our privacy.