Imagine this: It’s nighttime and a sleeping child suddenly has difficulty breathing due to asthma. The proper medication is dispensed through the vents in the child’s room and the asthma attack is averted. The next morning, the parents, who have slept through the night, see on their home “ambient intelligent” monitoring system that their child has had this episode and it was taken care of, without their knowing or having to go to the emergency room. Sound far-fetched? Cognitive computing will likely play a major role in the monitoring, diagnosis, and treatment of a child like the one in this scenario.

What is cognitive computing? Cognitive computing involves self-learning systems that use data mining techniques, pattern recognition, natural language and human senses processing, and system refinements based on real time acquisition of patient and other data. In other words, these systems mimic the way the human brain works and continue to learn. Scary?

Cognitive computing is currently being used at a number of leading oncology centers across the country, including Memorial Sloan Kettering in New York City and MD Anderson in Houston. Its use with unstructured data (data that is generally text-heavy and not organized in a predefined manner, with more than 80 percent of medical data being defined as such), best practice data, published clinical studies, and clinical trial data allow it to examine unlimited amounts of information in helping to make diagnosis and treatment decisions on these patients. It has become an important clinical support tool for clinicians in their decision making.

What has been illustrated above in the asthma case, however, is that cognitive computing can be more than just a clinical support tool. In this case, it is, essentially, a decision maker. So how do we get from being a clinical support tool (overseen by the clinical community) to a decision maker?

What is needed is the following:

  • Ability of cognitive computing to perform exceedingly well at pattern identification in individual patients. Medical sensors, which would perform the monitoring and diagnosis of the patient, need to not only fully understand the patient’s condition (including their disease and its treatments, social/contextual environment, patient needs, and prognosis) but also make correct decisions on which treatment(s) are appropriate, if any.
  • Clinical trials which demonstrate the clear clinical benefit of cognitive computing versus the standard of care (SOC, i.e. a clinician). This clinical benefit will need to demonstrate an improvement in patient outcomes, such as decreases in morbidity and mortality.
  • Cost effectiveness. Does cognitive computing in health care cost less than the current delivery of care model (i.e. SOC)? Clinical studies will also require that this be demonstrated—either in lower costs at the same quality of care, lower costs and improved quality of care, or increased costs and improved quality of care (at an acceptable incremental cost per incremental unit of quality of care—termed “incremental cost effectiveness ratio” or ICER).
  • A cultural change in how medicine is practiced. Physicians are the main decision makers in how patients are evaluated and treated. This may be the most difficult hurdle in trusting a machine to do the work of a human in life and death situations.
  • Willingness to pay for such technologies by payers and patients.

So while the future may be very promising for cognitive computing, there are some significant hurdles that still need to be overcome. However, if this can be accomplished, imagine your next primary care visit to Watson.

For those interested in hearing more about this, tune in to SiriusXM 111 on April 11 for the “Business of Healthcare” segment on cognitive computing. We have some great guests.