Googling for Diagnoses
Googling—the use of search-engines such as Google—is a new diagnostic tool in the modern clinician’s armamentarium.
A 2006 study published in The British Medical Journal examined how often searching with Google leads physicians to a correct diagnosis.1 The investigators were blinded to the correct diagnoses and compared the results of diagnostic cases published in the 2005 case records of The New England Journal of Medicine (NEJM) with diagnoses made using Google searches. Their results indicated that Google searches revealed the correct diagnosis in 15 of 26 cases, or 58% of the time. (See Table 1, p. 34.)
The limitations of the study are important to recognize when considering how well a Web-searched diagnosis can compare with one made without Web searching.
In some cases, although Google supplied information that approximated the diagnoses made in NEJM, the researchers thought the diagnoses were not specific enough.
For instance, one case for which Google suggested a diagnosis of extrinsic allergic alveolitis was described in the NEJM case record as extrinsic allergic alveolitis caused by Mycobacterium avium, also known as “hot tub lung.” This exemplifies the point that search engines don’t make diagnoses—people do. Expert knowledge is required to choose the best search words and select among the results.
Experts agree that using search engines for a diagnosis is likely to be more effective for conditions with unique symptoms and most useful in a certain class of diagnostic challenges in which clinical experts are at the helm of diagnostic deliberations.1,2
click for large version
click for large version
The choice of search engines is important and will increasingly affect accurate diagnoses as the technology develops.
“Hospitalists may utilize [Google] less; they are often working in environments [hospitals] that have ready access to more thoroughly vetted sources such as UpToDate, Medline, or MDConsult,” says Joshua Lee, MD, a hospitalist and medical director of Information Services at the University of California at San Diego. “Within Google there are variations on a theme, such as GoogleScholar, which accesses many of the same academic journals as Medline. Given that they are accessing the same journals, many would feel that searching with GoogleScholar is just as accurate as [searching with] Medline—at least as far as the source material is concerned.”
Search engines offer an impressive volume of new scientific material that is indexed daily. But, current search engines cannot identify differences of meaning in symptoms, drug names, anatomic features, or diagnoses.
“The key drawback to using open Google is that you cannot guarantee that the information comes from peer-reviewed journals,” Dr. Lee says. “When a search with ‘regular’ Google brings back untested medical information that is not peer reviewed, it gets dicey, and thus less reliable.”
Current Web technologies may merely be a first step in creating a much more powerful infrastructure for Web-based diagnostics. The so-called semantic Web will allow computers to process meaning in documentation.3 Two key elements of this revolution are the resource description framework (RDF) and Web ontology language (OWL) technologies. These next-generation search tools are being beta tested and will be freely available within the next decade.
There are a number of caveats to using and trusting information from search-engine investigations. Google stands up much better than other search engines in simple diagnostic searches. Also, diagnostic problems are not solved on a linear scale of difficulty.2 Research and development of the benefits of the semantic Web in clinical practice will help maximize this tool for diagnostics. TH
Andrea Sattinger is a frequent contributor to The Hospitalist.
References
- Tang H, Ng JHK. Googling for a diagnosis—use of Google as a diagnostic aid: internet based study. BMJ. 2006 Dec 2;333(7579):1143-1145.
- Gardner M. Diagnosis using search engines. BMJ. 2006 Dec 2;333(7579):1131.
- Berners-Lee T, Hendler J, Lassila O. The semantic web. Sci Am. 2001 May 1;34-43.
