The rise in automated intelligence tools such as IBM’s Watson is one of the more exciting developments in healthcare in recent years.  Alas, I wrote recently about a study explaining how our standards tend to be much higher when it comes to automated intelligence.

It found that we’re much more forgiving of mistakes when they’re made by humans than when they’re made by a machine, with simple errors often all it takes to permanently discard the merits of the automated system.

A recent study from the Intermountain Medical Center highlights just how harmful this bias could be.  The study looked at the potential for improvements to patient care when pneumonia based decision support tools were used in advanced clinical settings.

The research revealed that use of the tool saved 12 lives across the year of its application in the few hospitals in which it was used.  The researchers believe it could help to make a significant dent in the 50,000 or so Americans that die from pneumonia each year.

The decision support tool took advantage of the personal medical data from each patient together with other risk factors to provide a real time alart to physicians.

Once pneumonia was confirmed in the patient, the tool would then automatically calculate the severity as well as provide recommendations for managing the condition, based upon the recommended guidelines.

“Because of the complexity of pneumonia, physicians can’t easily make consistent decisions that follow current treatment recommendations,” the authors say. “The result is, doctors have to rely on their unaided judgment on how to best treat pneumonia patients.”

The authors go on to highlight how valuable the tool is at extending the capabilities of the human doctors on duty at any given facility.

“Unlike the human mind, computers aren’t limited in their scope,” they say. “Doctors, as with all people, are only able to process three to five factors at any given time despite their experience. Our screening tool uses 40 different factors in its evaluation of whether or not a patient has pneumonia and whether they should be admitted to the hospital for treatment.”

The researchers hope to release their tool more widely in the coming year, but as the initial study mentioned in this post highlights, there remains a significant psychological bias to overcome before these tools are an accepted part of the healthcare landscape.

Hopefully the initial adoption of the support systems will help to educate patients and healthcare professionals alike as to their robust ability to add value to the process, and begin to shift mindsets towards something rather more fair.