Artificial Intelligence

There’s a ground swell of change coming to healthcare, and it will have a direct impact on every provider and every patient. The wave that’s coming is being talked about throughout the industry, and was the focal point of this year’s HIMSS conference. It’s called big data, and it represents the convergence of data from providers, health systems, researchers and patients.

Maximizing the value of this overwhelming amount of data requires capabilities beyond traditional analytics. Enter the latest evolution of artificial intelligence (AI), or smart machines, that are designed to mimic human intelligence. The promise of AI is to achieve the holy grail of health management by reducing costs and increasing quality of care. But is it possible? In 2013, the University of Indiana did a study using 500 randomly selected patients. Using AI and simulations, they were able to demonstrate a 30% to 35% increase in patient outcomes, and reduce costs by 50%. While this is not empirical evidence, the promise has proven to be such a strong elixir that dozens of companies have cropped up offering AI-based solutions in every facet of the healthcare continuum.

Clinical Decision Support with AI

According to Bruce Hoyle, MD, FACPH, of the Advanced Vein Center in Orange, CA, research has told us that risk factors for venous disease include family and personal history. “Based on genetics, we know if both parents suffer from venous disease, the odds of their offspring experiencing it are up to 90%. If one parent has venous disease, that probability drops to 40%. If neither parent has it, the probability drops to 20% and other factors come into play, like lifestyle and occupation.”

Dr. Hoyle predicts we’ll soon be able to identify his future patients before they are aware of it themselves, and reaching out to those future patients to encourage behavior modification to reduce risk is part of what AI can assist with. But accomplishing that requires the population of his community to have shared patient records. Those records will need to be accessible to systems like IBM’s Watson, or a number of others that are capable of processing structured and unstructured data-like transcriptions, and of outputting actionable information.

AI will soon be everywhere in healthcare. In fact, clinical decision support (CDS) is required for Meaningful Use qualification. While CDS does not require AI, it can be significantly enhanced with it. AI is being used in financial analysis by hospitals, insurance companies, and at the state and federal levels. With the coming of the “Internet of Things”, micro-sensors in the home and those that are incorporated into medical devices will be generating data that will become part of your patient’s health record – and eventually, the collective of data from around the world. You’ll have more information for making decisions than you’ve ever had before. In fact, some are concerned we’ll have too much information, and that’s where artificial intelligence can both assist and hurt you.

Along with vast amounts of information to utilize when assessing a patient, at your side will be a computer that will also be evaluating that same information.

The computer will be intelligent enough to tell you the predicted outcome of your proposed method of treatment. For example, when evaluating a patient presenting with DVT, AI will be capable of determining the statistical probability of a potentially fatal DVT-associated massive pulmonary embolism, in addition to the probable outcome of your recommendations. In theory, this enhancement to your decision-making process should enable you to render more targeted and effective care. What’s more, as you document your treatment plans and monitor the results, your computer will be learning along with you, and will be improving its own decision-making skills.

If you’re thinking there may come a time when bureaucrats and others may deem your presence redundant, you’re not alone. It’s conceivable that financial pressures, resource limitations and geographic constraints may compel healthcare decision makers—few of which have any clinical experience—to trade your presence for machines. Those same machines would also be capable of denying treatment if the statistics aren’t favorable.

While replacing you is a more complex argument, using this patient-specific data and intelligent processing to dictate courses of treatment, or denial of treatment, is a more plausible reality. Dr. Brian Madden, head of Peak Health Medical Group in Santa Monica, CA, and medical director for the palliative program at Providence St. John’s Medical Center, considers such theorizing as conspiracy nonsense. He believes, “the ethical and moral implications of that type of decision making are significant, and we’re nowhere near those crossroads. It’s one thing to use population-based statistics to make decisions about what is paid for, but getting as granular as you suggest would be highly controversial.” While I agree with Dr. Madden that the point may be farfetched, if the food is there and the table is set, I can’t help but wonder how long before a famished government or corporation will resist the urge to dig in. One thing we can predict with certainty is that artificial intelligence will add exponentially to the quality and accuracy of actuarial tables, and this information can easily be combined with genetic and patient-specific details. How this new information will be used will indeed be interesting to watch.

Will the doctor-in-a-box ultimately replace you?

With 12 years of working closely with physicians on IT and EMR implementations under my belt, one point I have learned quite clearly is that it’s your presence that makes a difference in what and how a patient presents. As physicians, you still hold a trump card. The debate you’ll face will be centered not on what the patient tells you, but how they tell it to you. It’s this nuance that computers cannot yet detect, and this is your protected turf.

As smart as computers may get, until patients are comfortable enough to interact with computers as they do with you, and until machines are able to detect and interpret their subtle inflections and mannerisms, computers should remain data aggregators and decision support devices, not the final arbiters of patient care. It will be your responsibility to enforce the value of personal interaction in the care-giving cycle to keep bureaucratic decision-makers sober. Now, isn’t that just what you all need: Another thing on your plate and another battle to wage!