3 Ways to Manage AI Hype In Healthcare

Hospitals, medical practices, and health care facilities across the US are struggling. Their careers are busy and short. Their operating costs are rising. And demand for their services often exceeds capacity, limiting access to care.

Enter artificial intelligence. In the nearly two years since ChatGPT first brought AI into the spotlight, investors, tech companies, and healthcare organizations have invested heavily in AI, giving it plenty of press coverage, and they’re launching dozens of pilots, sometimes painting vague visions of AI saving healthcare.

However, the overall impact of AI on healthcare has been limited so far. Are we expecting more soon?

Expectations Compared to Reality

Across the broader (non-healthcare) economy, a growing chorus is decrying AI’s bearish status as the gap between expectations and reality widens. Although many companies are already using AI to generate emails, images and marketing materials, there is no “killer use” to justify the high cost of AI.

Compared to other industries, AI may have a more difficult time reforming healthcare, where the stakes are high, organizations are complex, and regulations are uncertain.

One, there are technical problems. Predictive algorithms are not generalizable to all settings. For example, hospitals using the sepsis algorithm “out of the box” (without training it on local data) had many false alarms and undiagnosed cases of sepsis. In addition, the resulting AI is too unreliable to perform high-quality tasks, such as triaging, diagnosing, and recommending treatment. The challenge is that “a system that produces an AI like GPT-4 is smarter than anyone you’ve met and dumber than anyone you’ve met,” explained the Research President of Microsoft Peter Lee. access to health care.”

Additionally, many doctors, nurses, and healthcare consumers are skeptical of AI, concerned that it will compromise privacy, exacerbate bias, and damage doctor-patient relationships. Based on their experiences with electronic health records – which have failed to meet expectations and contribute to fatigue – they find the fact that AI will have to improve health care no longer reliable.

Finally, implementing AI in the real world is complex, involves many stakeholders, requires significant resources, and is fraught with potential pitfalls. However, unlike early digital initiatives such as implementing electronic health records (over $34 billion in targeted payments) or holding back on virtual care (the COVID pandemic -19), donor organizations do not have significant incentives to adopt AI products, which increases costs and power. changing their job, often without directly increasing compensation.

Navigating Long-Term Change

None of this is to say that AI is useless or will be useless in health. Some organizations are already using AI solutions for positive benefits, such as preventing referrals and easing the documentation burden of doctors. As AI technology advances, it aims to improve various aspects of medical care, operations and research.

However, we must give up our expectations. History tells us that it will take years—not months—to build good AI products, integrate them into the workplace, and ultimately open up new, better ways to deliver care.

In this era of change, organizations that provide health care should take the following steps to maximize the benefits of AI now and in the future.

1. Safely Try and Test

They must follow the principles of evidence-based medicine, realizing that, while technology is exciting, healthcare is about people first, not products. Organizations such as the Coalition for Healthcare AI are developing standards for implementing AI healthcare systems and creating validation labs to test them. Healthcare providers should test solutions for critical problems and establish the governance and testing necessary to ensure they use AI-enabled tools safely and fairly.

2. Improve Maintenance Procedures

Health care is a complex adaptive system where multiple, dynamic components determine performance. Organizations implementing AI should take a holistic systems approach, looking beyond technology to include people, systems and design.

For one, before rushing into any health work, they must first ask if the work is worth doing in the first place. As Peter Drucker taught, “There is nothing so useless as doing with great success something that should not be done at all.”

Second, because bottlenecks—points where demand exceeds capacity—determine the speed at which the entire process can work, ​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​ had to identify and alleviate downstream problems before using AI to make processes more efficient. different is more effective. For example, self-scheduling will have little impact if doctors’ schedules are already full. And early identification of patients with sepsis won’t help if nurses and doctors can’t use the data.

3. Receive increased benefits

Organizations must refute the allegations clearly and establish the facts. AI will not magically fix everything related to health. And because large linguistic models cannot reason or understand well, many health problems may require new approaches that combine machine learning with traditional AI.

However, they can use today’s AI for modest benefits (eg, reducing some of the complexity and improving patient education and engagement) while preparing for future success. Importantly, they must not ignore non-AI opportunities to do better. And, above all, they need to reflect on who they are, what they do, and how they can do it better – with or without AI.