Health-Related Searches on the Internet
- Gunther Eysenbach, MD, MPH
- Christian Köhler, MD
- KEYWORDS:
- health education
- internet
To the Editor: There is ongoing uncertainty about how often consumers use the Internet to search for health-related information. For instance, although a 2000 survey found that 55% of individuals with Internet access have used the Web to find health or medical information,1 Phillipov and Phillips2 reported that 17 health-related search terms they found in the first 300 search terms of the Wordtracker Top 500 keyword report comprised fewer than 1% of all queries. It is possible, however, that people may use many different health-related keywords that are so rare that they never appear on the Top 500 Keywords list. To address this, we analyzed a random sample of search terms entered into a search engine over a longer period of time.
Methods
We used the Metaspy Exposed Web site (http://www.metaspy.com/), which lists 10 unfiltered search strings people are currently entering on Metacrawler, and which refreshes every 15 seconds with new terms. We developed a computer program that visited this page every 30 minutes between February 2001 and April 2002, extracted the search terms, and added them to a database. From a sample of 298 512 query strings in the database, a computer program randomly selected 2985 (1%) search queries that were independently classified by 2 physicians as either "non–health-related" or "health-related." Discordant ratings were resolved by discussion.
Results
Of the 2985 search expressions, 2827 (94.7%) were concordantly classified as non–health-related, 108 (3.6%) were coded by both raters as health-related, and 50 (1.7%) received a discordant classification (κ = 0.80 for interobserver variability, indicating substantial agreement). Thus, our estimate of health-related search terms as a proportion of all search queries conducted is in the range of 3.6% to 5.3%. Examples of search expressions that were coded discordantly included "stop thumb sucking" and "calcium crystals." After discussion, 135 (4.5%) of all search terms were classified as health-related.3 We also coded health-related search queries by broad categories (partly corresponding to clinical specialties) and found that the searches for which people entered the name of a medical organization or looked for specific health care services, as well as searches for medicines and diet, were the most frequent queries (Table 1).
Table. Non−Health-Related Queries and Health-Related Queries, by Category
Comment
Our data suggest that health information is not among the most commonly sought topics on the Internet. Nonetheless, given the millions of searches performed each day, neither could they be described as uncommon. Studies suggest that more than 95% of consumers use a search engine rather than going directly to a Web site when confronted with specific health-related questions.4-5 While our data were harvested from a single search engine, we doubt that the pattern of searches in other search engines would be different.
Acknowledgment: Jim Lai programmed the scripts and David Mason provided technical input.
