Big Blue and the future of medicine

This evening, I watched as the computer took on the jeopardy champions from the treadmill in the hospital gym. I cant help thinking that the computer would make a fantastic physician. In fact, it will likely not be long before similar systems start aiding, then replacing doctors, especially when it comes to formulating plans for diagnosis and treatment.

The jeopardy playing computer is an experiment in programing machines to mine a large data set to answer questions based on a random input. The system does this using Bayesian reasoning algorithms (i am assuming here) to determine %'s of each correct answer and uses a threshold to determine if that answer is correct. In theory, this is the exact same process that a physician uses to make a diagnosis (assuming they are using evidence based medicine (EBM), doctors should technically be assigning a probability that they are correct although I have yet to meet one that consistently does this. IBM needs to figure out how to program an Ego). If the data set entered into the jeopardy system is roughly the same size and complexity as the set of all medical data (I cant even guess this one) then that system should be capable of making expert medical diagnosis. If medical data is more complex, then it will be just a matter of time.

All this is depressing, because a computer should in theory be much faster and more efficient at parsing numerical data (epidemiology, trial data, testing sensitivities) then a human would ever be, and so computers should make better doctors. But what about the human touch? Certainly the person to person interaction adds value to the work of human physicians that computers will probably never match. However computers can add value by developing capabilities that go above what humans are capable of, such as faster learning, or enhanced sensory ability. Each part of the decision making process represents a technical challenge that can be chipped away at by engineers. This is why computers are evolving much faster than humans. Robotic surgery systems are clumsy and slow, but they are getting better. This is of particular concern for primary care physicians because in the modern healthcare field, the doctor's job is decision making and communicating the decisions with patients. (Secretaries and transcriptionists do the paperwork while nurses and techs provide the direct patient care). Watching the computer on jeopardy leads me to believe that in the not so distant future, machines will be better at making medical decisions than the physicians that humanity has relied on since the dawn of civilization. The evidence base and obsessive outcomes research that we had hoped would improve our profession may end up making us obsolete at the hands of machines that can use the research more effectively.