AI in Healthcare is a Looming Inevitability

Processes and regulations still need to play catch-up, however.

Artificial intelligence (AI) promises to greatly reduce the cost of healthcare. To explain, let’s go back in time. Alan Turing in 1927 published a seminal paper – “On Computable Numbers, with an Application to the Entscheidungsproblem.” 

In the paper he described the feasibility for a machine to make computations. During World War II, British intelligence had recruited him into the secret operations engaged in codebreaking at Bletchley Park. He built a cryptologic machine that could break the Enigma code being used by Germany to coordinate its battlefield and naval operations. (There was no encryption of wireless traffic at the time, so it was necessary to encode messages, because everyone could hear them.)

Soon, building these special machines was moved to building 2600 at the National Cash Register plant in Dayton, Ohio, operated by what later would be called the National Security Agency (NSA).

Computers were born, and the war was won. Even though the codebreaking was operating during the war, the existence of Turing’s machine remained secret until the early 1980s. During that time, no one suspected that the United States and the United Kingdom could decipher encrypted ciphers used by governments around the world.

In the 1950s, Turing published another paper –  “Computing Machinery and Intelligence” – containing the famous “Turing Test” for AI. If the responses from a computer could not be distinguished from responses coming from another human, then the machine would pass the test. In the past 70 years, AI has become more human-like. Some argue that the Turing Test has not been passed, but it has.

AI has won chess and Go tournaments, and now can pass an MBA exam. It can make complicated decisions. It recently created a complete Al Joe Rogan show, including the voice, indistinguishable from the real personality’s voice. It is better at summarizing information than humans. It writes and responds better than 99.9 percent of humans. In the workforce, AI is replacing humans at a relentless rate.

AI is not unionized; it does not sleep; it works 24 hours per day; it has a capacity to process millions of operations per second. A single giant data center could service the entire United States, and more.

AI in Healthcare

It is only fanciful to believe that AI will not be applied in healthcare, and eventually replace doctors, or at least take over much of what they do. Doctors are not much different from other experts who are being replaced.

But what might be the economic effects? Let’s do a quick estimation. According to Physicians Thrive data, there are 950,000 patient care physicians in the U.S. Eighty-six percent deliver care across 48 practice areas. What does this cost? If we average physician salaries of the 10 highest-paying cities (Milwaukee is the highest), combined with the 10 lowest-paying (San Antonio is the lowest), it comes to $380,000 per year. Some physicians obviously make far more, perhaps many far less.

We must assume that AI is more appropriate for some practice areas of medicine than others. Looking at the 48 practice areas, we can isolate those where AI could not possibly provide service. For example, in surgery, it is necessary to physically touch the patient, something AI cannot do. In preventative medicine or psychiatry, however, such specialized touching is not as important, and if there are tests, many could be done by non-MDs. This “complex touching” is one way to find a fit for AI in the medical field.

If we isolate and screen out those practice areas that do not require such complex touching, it will reduce the number of physicians needed by 31 percent.

What would be the economic effect? Here, of the $195 billion paid in salaries (2022 data), there could be a cut of $60 billion – not an insubstantial figure.

AI can help with diagnosis, treatment planning, and monitoring of patients. It can analyze medical images and patient data to help identify potential health issues more quickly and accurately than humans, even if they are well-trained. AI can identify patterns and trends in patient data that may be missed by humans alone, and it can do this more or less immediately.

Of course, healthcare is a complex and highly regulated field. Any AI will be required to meet strict regulatory requirements to ensure patient safety and protect privacy. Many of these requirements have yet to be drafted.

There are many standardized medical procedures and treatments, but patient care involves a complex interplay of medical knowledge, clinical experience, patient history, and individual patient preferences. The consensus is that it is unlikely that AI will fully replace doctors, even for standardized medical procedures. My view is that this assessment is too cautious, or narrow-minded.

AI already is proving itself. In underserved populations, or the developing world, AI-powered telemedicine platforms are being developed that will allow patients to consult remotely. AI can provide automated medical triage and diagnosis for common health issues, such as fever, cough, or sore throat; yet this is the very beginning. AI’s capabilities are much greater, but the medical community faces a learning curve in adjusting to this new silent partner.

In these instances, particularly where Medicare is scarce, AI promises to drive down the cost of doctors to zero. This may be better accepted in doctor-scarce countries than in doctor-rich countries.

But the bottom line is that AI is smarter and cheaper than humans. In addition, many persons prefer computers and robots to humans. The economics are compelling, and AI has the potential to eliminate many of the problems with coding.

It is only a question of time.