Google inadequate for diagnoses

A study published last week in the British Medical Journal and reported in Reuters Health and other general interest media suggested that physicians might turn to Google as a diagnostic aid. In the simulation run by the journal authors, Google selected the correct diagnosis 58% of the time.

While the idea of using Google as an ersatz clinical decision-support system is clever, a 58% accuracy rate is unacceptable -- in either a human clinician or a software program. Google, of course, was not designed for this purpose. It uses a Boolean search engine -- with and/or logic -- to scan a massive, unverified database, the Internet.

Older-generation diagnosis-decision software systems have much higher accuracy rates. In addition, the latest generation of diagnosis reminder systems, such as Isabel, consistently suggest the proper diagnosis 90% of the time. These new programs use advanced natural-language processing algorithms -- a newer, more powerful search technique -- to scan a specific database of medical journals and texts. This produces more accurate, higher-quality search results.

The aim of diagnosis support software is to support, not replace, the "learned intermediary" by providing highly reliable medical information in an easy-to-use format. Google is great for some things, but it would be a mistake for physicians to rely on it in critical cases.

Joseph Britto, M.D.

Chief executive officer

Isabel Healthcare

Reston, Va.

Comment on today's news and share your opinions with other readers. Submit your letter to Your Views at hitsdaily@crain.com. Submissions must include name, title, affiliation, city and state. Health IT Strategist reserves the right to edit all submissions for news value, style, grammar, length and appropriateness. For more information, contact Webmaster/Copy Editor Patrick Shrader at pshrader@crain.com.