12 January 2018
Artificial Doctors in a Human Era
In Health IT Outcomes, Anton Dolgikh, Head of AI in the Healthcare and Life Sciences Practice at DataArt, discusses the diagnostic capabilities of modern AI, including the potential of AI-powered medical expert systems to curtail medical error, the third leading cause of death in the US.
“We intuitively believe that the chances of a doctor making an erroneous diagnosis and proposing an ineffective or harmful treatment plan are low. At least in our case. Unfortunately, the statistics about the accuracy of human doctors is disappointing. According to the 2016 research, the third leading cause of death in the US is medical error, directly following heart diseases and cancer (See Figure 1).
Human doctors are prone to error. In some cases, errors occur due to distractions, in others they stem from the difficulty of obtaining information such as the patient’s medical history. Whatever the reason, an artificial assistant could diminish the influence of these factors.
The most famous medical expert system is Mycin. It was designed to help diagnose blood infections caused by bacterimia. It demonstrated a level of accuracy 10.5 percent higher than that of human doctors: 65 percent of MYCIN’s antimicrobial therapy prescriptions were classified by experts as acceptable compared to 55.5 percent of prescriptions made by human doctors.”