The search engine as a diagnostic tool in complex immunological and allergic case reports: is google useful?

Background:

Web-based search-engines have become an important source of knowledge and communication. Google is the most popular search-engine (64% of all web-searches in March 2007) whilst Yahoo accounted for 21%. Recently Google's value to guide doctors to a correct diagnosis in case records of the New England Journal of Medicine was reported.

Objective:

To evaluate the utility of searching clinical information with Google in order to obtain a correct diagnosis of complex immunological and allergic (CIA) case reports.

Study Design:

Comparative cross-sectional study.

Methods:

Firty-five CIA case reports were randomly selected by an independent investigator from peer-viewed medical journals. Clinical data was presented separately to three observers blinded to final diagnosis. Observer A is a Consultant in Internal medicine and Allergy with an expert knowledge of these fields and basic computing skills. Observer B is a Registrar in Internal medicine and Allergy. Observer C is a research nurse. Both observers B and C had a more familiar knowledge of the regular use of computer search engines. An internet-based search using Google was conducted. In order to perform this, the observers individually studied each text and independently selected five search terms, of their own choosing, from each case record to enter into the standard Google search engine. The observers then recorded for each case the single most prominent diagnosis that was evident from within the first three results pages of the conducted Google search. Since Google does not necessarily include diagnoses within the search results page itself, observers were permitted to select the diagnosis that best fitted the case record from information after opening each direct results link only. The independent investigator then compared the diagnoses obtained by each observer with the definitive diagnoses as published in the Journals. The main outcome measure of this study was the percentage of correctly obtained diagnoses achieved by each observer.

Results:

Observer A identified the definitive diagnosis in 30/45 cases (66%, 95%CI 52-79). Observer B in 39/45 (86%, 95%CI 76-95) and Observer C in 29/45 (64%, 95%CI 50-77). Most diagnostic inaccuracies for both observers were those related to primary immunodeficiency or pediatric cases.

Conclusion:

This Google-based search was useful to achieve an appropriate diagnosis in CIA cases. Computer and Internet-based search skills could influence the results.