How Can We Leverage Healthcare AI While Ensuring Data Privacy and Security?

Editor’s note: Jakub Hlávka authored Security, Privacy and Information-Sharing Aspects of Healthcare Artificial Intelligence, published in 2020 by Academic Press. The book is available here.

Healthcare is awash with data, and digital security and privacy concerns are justified given the sensitivity and personal nature of these data. Novel artificial intelligence (AI)-based technologies aiming to integrate and analyze data from multiple sources and for different audiences continue to face multiple privacy and security challenges, ranging from practical (making sure that data can be extracted, decrypted and analyzed) to regulatory (ensuring that privacy of patients and healthcare providers are protected, and data are used for legal activities).

Moreover, some stakeholders have taken advantage of the digital ‘Wild West’: it is not uncommon for emerging health technology companies to emphasize data privacy and security in their marketing materials, only to be found later to use sensitive data in ways their consumers did not authorize or anticipate. For example, by sharing information with law enforcement.

The problem is compounded because archaic data infrastructure or overly restrictive internal policies often prevent data from being used to benefit patients: hospitals often are not able to release records electronically when requested by a patient, and different healthcare providers use incompatible systems. The use of outdated technologies, such as fax, and behavior like ‘data blocking’ prevent the use of advanced tools in the analysis of data.

Similar to other digital solutions, AI will not resolve existing challenges, but once regulators and innovators address them, it may provide suitable tools to deliver high-quality care at a lower cost.

The Future of AI is Promising

There are a number of promising developments and trends in healthcare AI: the emergence of solutions in both secure and private artificial intelligence (to defend against threats to data integrity and privacy, respectively) using encryption and other tools, data virtualization approaches (data do not leave their secure storage), and others. Yet, many new solutions add to the complexity of existing challenges and are simultaneously going to exacerbate some of the risks associated with electronic information exchange while increasing the clinical utility of patient-level information and the productivity of healthcare professionals.

The situation is more complex because policymaking often trails technological advances with significant delays. Regulations that are updated are usually general and do not provide specific guidance needed for the development of technical standards and data interoperability. As a result, innovators often face the grave dilemma: push the boundaries – and face the risk of litigation – or remain conservative and perpetuate existing issues in data sharing or even data blocking, which significantly degrades the experience of patients, providers and payers in the delivery of care.

Emerging Challenges and Opportunities in Healthcare

Healthcare AI creates new opportunities for care delivery. Primarily, it is associated with greater data utility, ranging from drug development to reimbursement. It may also increase physician productivity and can even help attribute privacy violations to specific actors. However, the use of AI also results in additional challenges, including greater susceptibility to data privacy and integrity attacks, a more limited level of control over data ownership, a less intuitive understanding of data privacy, and privacy externalities.

Despite these complex challenges to data privacy and security, innovation in healthcare AI will continue, conditional on clearer regulatory standards and buy-in from key stakeholders, particularly providers and payers. Perhaps not surprisingly, the United States is trailing other countries in healthcare digitalization due to its fragmented healthcare system, and as a result will take longer to broadly adopt new AI-based solutions. This, however, creates a unique opportunity for private and public stakeholders to come together in addressing some of the most important challenges outlined in the chapter: data ownership, data silos, interoperability standards and excessive litigation, all of which make the adoption of AI-based solutions in American healthcare challenging.