There is already evidence that AI systems can raise diagnostic accuracy and disease tracking, improve prediction of patient outcomes, and suggest better treatments. It can also raise the efficiency of hospitals and doctor’s offices by taking over tasks such as drug transcription and patient monitoring, as well as streamlining administration. It may already be speeding up the time it takes for recent drugs to enter clinical trials. Modern tools, including generative artificial intelligence, can expand these capabilities. However, as ours Technology Quarterly this week shows that while artificial intelligence has been used in healthcare for many years, integration is sluggish and results are often mediocre.
There are good and bad reasons for this. A good reason is that healthcare requires high evidence barriers when introducing recent tools to protect patient safety. Bad reasons include data, regulation, and incentives. Overcoming them could bring lessons for AI in other areas.
Artificial intelligence systems learn by processing huge amounts of data, something healthcare providers have plenty of. However, health data is very fragmented; strict rules control its exploit. Governments realize that patients want their medical privacy protected. But patients also want better and more personalized care. Each year, approximately 800,000 Americans suffer from impoverished medical decision-making.
Improving the accuracy and reducing bias of AI tools requires training them on enormous datasets that reflect the full diversity of patients. It would be helpful to find secure ways to allow health data to flow more freely. However, this can also benefit patients: they should be given the right to access their own records in a portable digital format. Consumer health companies are already using data from wearable devices, with varying degrees of success. Portable patient records would enable patients to make more exploit of their data and take greater responsibility for their health.
Another problem is how to manage and regulate these innovations. In many countries, AI governance in health, as in other areas, is struggling to keep pace with the rapid pace of innovation. Regulators may be sluggish to approve recent AI tools or may lack capacity and expertise. Governments must equip regulators with the tools to evaluate recent AI tools. They also need to fill regulatory gaps in the surveillance of adverse events and the continuous monitoring of algorithms to ensure their accuracy, safety, effectiveness and transparency.
This will be hard. One solution would be for countries to cooperate, learn from each other and create minimum global standards. A less intricate international regulatory system would also facilitate create a market where petite businesses can innovate. Poorer countries with less developed health infrastructure have much to gain from the introduction of recent tools, such as a portable, artificial intelligence-powered ultrasound device for obstetrics. Since the alternative to AI tools is often no treatment, they may even be able to leapfrog the established healthcare systems in affluent countries – although lack of data, connectivity and computing power will stand in the way.
The final problem concerns institutions and incentives. Artificial intelligence promises to lower medical costs by assisting or replacing workers, increasing productivity, reducing errors, and flattening or reducing expenses, all while improving care. This is urgently needed. By 2030, there could be a global shortage of 10 million healthcare workers, representing approximately 15% of today’s workforce. The administration accounted for about 30% of excess health care costs in America compared to other countries in 2022.
However, saving money through innovation is hard. Health care systems are set up to exploit them to improve care, not to cut costs. Modern technologies may account for up to half of the annual growth in healthcare spending. Layering recent systems will raise cost and complexity. However, redesigning processes to effectively exploit AI will likely be met with resistance from patients and physicians. While AI may be able to triage patients over the phone or provide routine results, patients may request an in-person visit.
Worse still, many healthcare systems, such as America’s, are set up to reward volume of work. They have no reason to implement technologies that reduce the number of visits, tests and procedures. Even public health systems may lack incentives to adopt technologies that reduce costs rather than improve outcomes, perhaps because saving money may result in a smaller budget next year. Unless governments change these incentives so that AI combines better treatments with recent efficiencies, innovation will drive up costs. As a result, governments and health authorities will need to fund programs dedicated to testing and implementing recent AI technologies. Countries including America, the UK and Canada are leading the way.
AI, medical doctor
Much of the burden of supporting AI in healthcare falls on governments and regulators. However, companies also have a role to play. Insurers have already used AI tools to unfairly deny care; companies mis-sold or overestimated the capabilities of AI in health; the algorithms made mistakes. Companies have a responsibility to ensure that their products are safe and sound, reliable and responsible and that humans, regardless of their faults, remain in control.
These obstacles are formidable, but the potential benefits of AI in healthcare are so enormous that the case for overcoming them should be obvious. And if AI can be applied to medicine, it could be a recipe for adoption of the technology in other fields.
Published: May 30, 2024, 6:00 PM EST
