By Irving Wladawsky-Berger
“Since they became publicly available in late 2022, generative artificial intelligence (genAI) tools such as ChatGPT have elicited enormous enthusiasm, as well as concern, in all sectors of the economy,” said “Will Generative Artificial Intelligence Deliver on Its Promise in Health Care?,” a recent article in the Journal of the American Medical Association (JAMA) by Robert Wachter, — professor and chairman of the Department of Medicine at the University of California, San Francisco (UCSF), and Erik Brynjolfsson, — professor and director of the Digital Economy Lab at Stanford University.
“But the potential impact of genAI in health care seems particularly noteworthy. In a field in which an estimated 30% of the $4.3 trillion spent each year in the US adds little to no value, in which many tens of thousands of people die yearly from preventable mistakes, and in which access to care is fragmented and inequities are commonplace, it is natural to be enthusiastic about the potential for genAI to improve quality, efficiency, equity, and patient experience.”
A few weeks ago I posted an entry in my blog on “AI and the Future of Healthcare Systems,” in which I discussed the huge challenges posed by the inherent complexity of the healthcare sector of the economy. Healthcare is a system of coupled systems, comprised of medical and pharmaceutical research; the delivery of healthcare to patients by a variety of practitioners, including hospitals, physicians, nurses, and pharmacists; and the insurance companies and governments that pay for healthcare. The convoluted interactions of these various domains, — each of which is quite large and complex in its own right, — is one of the main reasons why healthcare systems are so incredibly complex.
Given the central role of data in AI, electronic health records (EHRs) play a critical role in the development of AI systems capable of dealing with the complex real-world challenges in the healthcare industry. As explained by the Centers for Medicare & Medicaid Services (CMS): “An Electronic Health Record (EHR) is an electronic version of a patients medical history, that is maintained by the provider over time, and may include all of the key administrative clinical data relevant to that persons care under a particular provider, including demographics, progress notes, problems, medications, vital signs, past medical history, immunizations, laboratory data and radiology reports.”
Over the past 10-15 years, there’s been considerable progress in the adoption of EHR standards for sharing information among clinicians and organizations, helping providers make better decisions, improve patient care and reduce the incidence of medical errors. “While optimism regarding genAI is certainly warranted, so too is skepticism,” wrote Wachter and Brynjolfsson.
“Health care has several attributes that make the successful deployment of new technologies even more difficult than in other industries; these have challenged prior efforts to implement AI and electronic health records,” they added. “When considering whether genAI will deliver on its promise in health care, one way to shape the conversation is around 2 critical factors. First, is there something about genAI, compared with previous technologies, that will hasten iterative improvements in the technology? Second, is there something about the intersection of genAI and the current health care ecosystem that will accelerate the development of complementary skills and processes, or partly obviate the need for them?”
Let me summarize their answers to these questions.
The Productivity Paradox in Healthcare
The first critical factor is related to the so called productivity paradox that’s long puzzled economists, namely the repeated historical examples where a major transformative technology, — e.g., the steam engine, electricity, computers, the internet — failed to have a major impact on productivity in its early years. There’s generally been a significant time lag, — sometimes decades, — between the initial marketplace appearance of a transfirmative technology and its widespread deployment across companies and industries, when the now well accepted technologies and business models become the norm, leading to long-term economic and productivity growth.
Historically transformative technologies have great potential from the outset, but realizing that potential requires major complementary investments including business process redesign; innovative new products, applications and business models; the re-skilling of the workforce; and a fundamental rethinking of organizations. Moreover, the more transformative the technology, the longer it will take to reach the widespread deployment phase.
For example, while James Watt’s steam engine ushered the Industrial Revolution in the 1780s, its impact on the British economy was imperceptible until the 1830s because productivity growth was initially restricted to a few industries.
Similarly, US labor productivity grew at only 1.5% between 1973 and 1995. This period of slow productivity coincided with the rapid growth in the use of IT in business, giving rise to the Solow productivity paradox, a reference to Nobel Prize MIT economist Robert Solow’s 1987 quip: “You can see the computer age everywhere but in the productivity statistics.” But, starting in the mid 1990s, US labor productivity surged to over 2.5%, as fast growing Internet technologies and business process re-engineering helped to spread productivity-enhancing innovations across the economy.
“Over the past 15 years, health care’s dominant general technological transformation came with the implementation of EHRs,” wrote Wachter and Brynjolfsson. “EHR adoption was accelerated by federal incentive payments under the HITECH Act, passed during the Great Recession of 2008 and implemented beginning in 2010. In 2009, when HITECH became law, fewer than 1 in 10 US hospitals had an EHR; a decade later, fewer than 1 in 10 did not.”
“While EHRs have cut the rate of medication errors and delivered numerous other benefits, the evidence that they have improved productivity is mixed, particularly when factoring in the EHR-associated increase in clinicians’ documentation burden. The latest unanticipated consequence is the explosion in electronic messages coming from the patient portal to the physician’s EHR inbox. Clinicians often cite the EHR as a key factor in their dissatisfaction with work and high levels of burnout.”
The authors explain that in addition to its inherent complexity, the digital transformation of healthcare is significantly more challenging than has been the case in other industries. These challenges include:
Healthcare is highly regulated. HIPPA, for example, prohibits healthcare providers and healthcare-related businesses from disclosing protected information to anyone other than a patient without their consent. While protecting the privacy of individuals, these regulations restrict the data sharing that’s essential to achieving many of the benefits of generative AI.
Market concentration. The market for EHRs is highly concentrated among a handful of companies that support the vast majority of healthcare providers, making it difficult for smaller companies to develop innovative AI tools and applications.
Massive amount of new medical and technical information makes it difficult to integrate these advances into patient care recommendations. It’s estimated that nearly 2 million articles are being published in scientific journals each year.
Stakes in healthcare are too high to tolerate flaws that could harm patients, making it harder to introduce new digital technologies and applications compared to most other industries.
Will GenAI Overcome the Productivity Paradox in Health Care?
Despite these serious challenges, Wachter and Brynjolfsson wrote that they see several reason why genAI will lead to healthcare productivity and quality gains more quickly than those achieved in the past. These include:
GenAI-related tools are remarkably easy to use compared to most other technologies. “The unprecedented adoption curve of ChatGPT and subsequent versions of language-based generative AI (100 million users in the first 2 months) is partly due to the fact that no special training is required.”
GenAI can be deployed via the existing software and internet infrastructure, which helps to accelerate adoption. “Contrast this with the implementation of EHRs, a transition that required a large investment in hardware and a wholesale change in the way most health care work, both clinical and back-office, was organized.”
Application programming interfaces (APIs) make it easier to integrate genAI third party applications with the EHRs of the major vendors. “Trying to stay ahead of this potential competitive threat, major EHR vendors are also rapidly integrating genAI into their own software offerings.”
GenAI technologies are able to deliver productivity and quality gains more quickly than prior technologies. Historically one of the key factors “in overcoming the productivity paradox was the speed with which the technology underwent iterative improvement cycles.”
“In fact, we expect that genAI will notch its early wins in health care delivery systems not so much by handling patient-facing tasks (such as making diagnoses and recommending treatments) but rather in addressing areas of waste and administrative friction, whether in creating a physician note, scheduling a patient appointment, or sending a bill or a prior authorization request to an insurance company,” wrote Wachter and Brynjolfsson in conclusion. “Experience gained in these areas will likely pave the way for broader implementation in areas that more directly affect patient outcomes and experience.”
“[W]e are optimistic that the 2 key factors that have historically been critical in overcoming the productivity paradox — the ability of the digital tools to rapidly improve and the capacity of organizations to implement complementary innovations that allow IT tools to reach their potential — are more advanced than in the past. Because of this, we believe that genAI will deliver meaningful improvements in health care more rapidly than was the case with previous technologies.”
“Does that mean that health care will be completely transformed by genAI in the next few years? That seems unlikely, although certain use cases, such as digital scribes and some forms of back-office automation, could make a big difference relatively quickly. But it does mean that what might have been a decades-long path for genAI to overcome the productivity paradox in health care may now be traversed in 5 to 10 years, and for some digitally advanced organizations, even sooner.”
