Since ChatGPT roared onto the scene, the rising concern over AI’s impact on the healthcare sector has been met with oscillating amounts of enthusiasm and scepticism. Speculation has been rampant and hyperbolic, oftentimes ignoring both historical innovations at a similar scale and the current state of the healthcare industry. While the literature cycles, both the healthcare industry and the engineers working to develop AI have been creating models and strategies to make use of the new tech and solve the problems of the complex, regulated, and important healthcare industry.
The Rise and Role of ChatGPT in Healthcare
Language models, neural nets, and the other underlying AI technologies have existed for decades in various levels of potency. Nevertheless, chatGPT and other models represent a monumental advancement in the domain of AI – achieved both through new, more effective architectures and a critical mass of training data & model parameters. Since its inception, ChatGPT has transformed numerous sectors with healthcare witnessing significant strides. Its ability to comprehend, engage, and provide quick resolutions has been a boon for things like script management, note generation, patient triaging, and providing resources in real time. There are downsides. Even at face value, issues have appeared around occasional “hallucinations” of fake information. More subtly, the training data used to create these models have well-documented diversity problems. Models that work less well for marginalised groups, will leave the disadvantaged even further behind. Any mistake or diversity issue that results in harming a person’s health is unacceptable. Future solutions must be built with safety checks in place to avoid these types of problems.
A Glimpse into Historical Innovations
AI is not the first revolutionary technology that has been released. It is not even the first in our lifetimes. Consider the invention of the computer. Initially, the computer was perceived as a potential threat to jobs, especially those related to data entry, calculations, and manual record-keeping. And to be fair, there are jobs that existed at the time that no longer do. However, we did not see an impact in employment rates. It didn’t snatch away jobs but rather transformed them. Accountants, for instance, could now focus on analysis rather than mere number-crunching. Notably, people were not replaced. Instead, an individual’s productivity and accuracy skyrocketed as they worked in tandem with a new, useful tool. Similarly, the rise of the internet was seen by many as a disruptive force that would eliminate several professions. In contrast, it broadened horizons, creating fields that we hadn’t even conceived of before—think of digital marketers, web developers, and e-commerce specialists. Communication and access to information were dramatically improved.
In both these instances, the initial fears that arose around these new technologies sound similar to the issues raised with artificial intelligence. However, rather than introducing a multitude of quality issues, replacing humans, and creating hosts of new concerns, technology did what it does best: eradicated the monotonous, strenuous tasks, and enabled humans to channel their energies into more meaningful, value-added roles.
Healthcare’s Current Quandary
Now, consider the current state of healthcare in the US. Today, the sector is grappling with two overarching dilemmas: burnout and employee shortage. Though the outbreak of COVID-19 has exacerbated these issues significantly, the root causes are unrelated. Medical professionals find themselves inundated with administrative tasks, sapping valuable time and energy that could otherwise be directed toward patient care. Additionally, both the US population and the proportion of that population over the age of 60, are growing faster than the number of service providers, leading to a skyrocketing demand for healthcare services. Predictions suggest that the future could witness even more pronounced worker shortages, further straining an already beleaguered system.
AI: A Solution?
In this backdrop, AI emerges not as a super-powered villain to take over people’s livelihoods and mislead, but rather as an important tool to step into a troubled industry and provide relief. AI’s capabilities to handle non-standard, yet still repetitive tasks—be it data parsing, preliminary diagnoses, or summarising patient history records—can alleviate the immense pressure on healthcare workers. By automating these tasks, medical professionals can redirect their focus on direct patient care for the more complex cases, research, and other critical areas.
Moreover, AI’s potential isn’t just restricted to easing current burdens but also addressing impending challenges. With the spectre of worker shortage looming, AI can fill critical gaps, ensuring that healthcare delivery doesn’t falter. The current model for this is referred to in the tech world as the “agent” model and is already employed to great effect within healthcare today. Rather than delivering care in isolation, some primary care providers manage teams of PA’s, nurse practitioners, and others to help to handle lower-impact cases, things outside the norm are effectively escalated to the physician in charge. Although this is less prevalent today, social workers also manage teams of community health workers, and through this model are able to solve the base needs of a much larger population than they can reach by themselves. This, coincidentally, mirrors a highly effective strategy for AI utilisation employed outside of healthcare. Those at the forefront of AI development often complete work using a team of specialised “AI agents” responsible for various tasks, or managing interactions with various sources. The manager in this model similarly provides feedback and deals with any necessary escalations, but leaves the more redundant work to the AI.
Like the computer and the internet that have come before it, AI is neither terrifying nor irrelevant. It will change the world fundamentally, but when it is finished things will look about as similar as the 80’s look to today. At Pear Suite, we see this as a good thing. Yes, work still needs to be done to ensure that the hallucinations won’t harm outcomes, and a better dataset needs to be built, but we are striving for a world where everyone has access to the social health resources they need to live a healthy and happy life. This necessitates effective social care navigation at scale, which in turn necessitates automation and artificial intelligence. Not as a replacement for the essential, human aspect that a good care navigator can provide, but as an augmentation. A teammate that can handle the boring, tedious, and painful bits that suck the life out of our days for everyone.
Nick Lockett
Nick Lockett is co-founder and CTO at Pear Suite, a digital health company working to empower community health workers and better address the social determinants of health.
