Working Paper: NBER ID: w30639
Authors: Ariel Dora Stern
Abstract: For those who follow health and technology news, it is difficult to go more than a few days without reading about a compelling new application of Artificial Intelligence (AI) to health care. AI has myriad applications in medicine and its adjacent industries, with AI-driven tools already in use in basic science, translational medicine, and numerous corners of health care delivery, including administrative work, diagnosis, and treatment. In diagnosis and treatment, a large and growing number of AI tools meet the statutory definition of a medical device or that of an in-vitro diagnostic. Those that do are subject to regulation by local authorities, resulting in both practical and strategic implications for manufacturers, along with a more complex set of innovation incentives. This chapter presents background on medical device regulation—especially as it relates to software products—and quantitatively describes the emergence of AI among FDA-regulated products. The empirical section of this chapter explores characteristics of AI-supported/driven medical devices (“AI devices”) in the United States. It presents data on their origins (by firm type and country), their safety profiles (as measured by associated adverse events and recalls), and concludes with a discussion of the implications of regulation for innovation incentives in medical AI.
Keywords: Medical AI; Regulation; Innovation Incentives
JEL Codes: I11; I18; K2; K32; O31; O32; O33
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| regulatory classification of AI tools as medical devices (L64) | development and market entry (O36) |
| stricter regulatory requirements for Class III devices (L64) | deter firms from pursuing AI innovations (O31) |
| regulatory oversight (G18) | safety of AI-driven technologies (J28) |
| regulatory environment (G38) | adoption of AI tools in healthcare (I11) |