Machine Learning and Patient Expectations: Potential Point of Disconnect

Integration of machine learning in medicine has been linked with emphatic phrases such as “instant,” “personalized,” “real time,” “superhuman data analysis capacity.” A recent article in the American Journal of Medicine presented a futuristic view of machine learning in medicine and its potential to expedite the diagnostic and therapeutic pathway for a patients at various stages.1 This should make us think. Does the story really end after administering the right treatment or therapeutic intervention? In another article, about half a century ago, WB Schwartz prophetically predicted the integration of computers in medicine quite accurately just short of the term “machine learning.”2 The above examples and many others in the literature inadequately consider certain non-modifiable factors involved in the process: time scale dissociation of machine learning, the natural time course of a disease, and patient expectations.