Searching for cancer-related information online: Unintended retrieval of complementary and alternative medicine information

Summary

Purpose:

The Web is an important source of health information for consumers. Use of complementary and alternative medicine (CAM) is also increasing. Therefore, we studied the likelihood that consumers will incidentally encounter CAM information while searching the Web and the factors that influence retrieval of CAM information.

Methods:

We evaluated results retrieved by 10 cancer-related searches on six common search engines.

Results:

Of 1121 search results, 16.2% displayed CAM information. Sponsored (i.e., paid) results were more likely to display CAM information than non-sponsored results (38% versus 7.5%, p < 0.001). In Overture and Google, sponsored results accounted for 51% and 39% of results on the first page. These search engines also retrieved more CAM web pages. Search engines distinguished sponsored and non-sponsored results, but disclosure statements describing the differences were confusing. Cancer type used as the search keyword did not influence the number of CAM web pages retrieved. However, synonyms of cancer differed in their retrieval of CAM web pages (p < 0.001). Consistent with prior studies of Web search engine overlap, we found that 28% of CAM results were retrieved by two or more search engines.

Conclusions:

Clinicians should help consumers recognize sponsored results and encourage search engines to clearly explain sponsored results.

Introduction

Sixty-two percent of Web users in the USA have searched for health information online [1]. Consumers usually use general-purpose search engines without a defined search plan [2]. Although users express concern regarding the quality and accuracy of online health information, few users recall where they obtained the information to answer their health-related question [2]. In spite of growing concern about quality and accuracy of online health information, there is mounting evidence that online information affects consumer behavior [3], [4], [5]. Therefore, we studied the likelihood that consumers will incidentally encounter information regarding complementary and alternative medicine (CAM) while searching for cancer information on the Web, as well as the factors that influence the retrieval of CAM information.

In contrast to conventional medicine, CAM is controlled primarily by consumers. CAM therapies are generally not regulated by government agencies and do not require a prescription. Therefore, consumers have direct access to CAM therapies, such as dietary supplements, without the help of a licensed clinician. In addition, many patients who use CAM do not share this information with their physician [6].

CAM use is high, particularly among cancer patients. Richardson et al. found that over 80% of cancer patients surveyed at a large cancer institute had used CAM therapies [7]. Further, the type of cancer may affect CAM usage. Over 72% of breast cancer patients interviewed 2–3 years after initial surgical treatment used some form of CAM [8]. Morris et al. report that “breast cancer patients are far more likely to be consistent users compared with other tumor sites” [9].

Some patients specifically search for CAM information online. Indeed, 48% of searchers, report looking for information about CAM or experimental treatments. Those treated for a serious illness in the past year were more likely to have searched for such information [1]. Others encounter CAM information when searching for information about disease states. How such unintentional retrieval of CAM information affects patient choices is not known.

As an unregulated domain, CAM information on the Web may be particularly susceptible to poor quality. In prior work, we found consumers searching for ‘arthritis’, ‘fibromyalgia’ and ‘diabetes’ were likely to encounter commercially oriented CAM websites whose information was not supported by the conventional literature. About 70% of these CAM websites were selling products or services [10]. In another study, Ernst et al. reviewed 13 cancer-related CAM websites retrieved via CAM-specific searches, and found 38.5% of websites contained information that may harm a patient if the advice were followed [11].

Women use CAM more [12] and look online for information in greater numbers than men [3]. Therefore, searches on cancers that affect women may return more CAM web pages than searches on cancers that affect men. Specifically, we expected ‘breast cancer’ searches to retrieve more CAM web pages than searches for ‘prostate cancer’. Synonyms of cancer may also affect the retrieval of CAM web pages, as consumers likely search using lay terms rather than scientific ones.

Many search engines accept payments for websites to be prominently listed in their results. In fact, this industry is projected to be worth US$ 4 billion by 2005 [13]. Such ‘sponsored results’ are different from traditional advertising, such as banner ads, because they are generated in response to specific searches. Sponsored results are prominently displayed at the top of the results page, sometimes along with regular search results. Perhaps most concerning of all, consumers are unaware of the existence of sponsored results and do not differentiate them from non-sponsored results which are generally retrieved using an algorithm that does not factor payment for placement. In fact, 41% of links clicked by consumers are sponsored [14]. The Federal Trade Commission has recommended that search engines clearly identify sponsored results [15]. Since most CAM information is commercially driven, we hypothesized that sponsored results contain more CAM information than non-sponsored results.

Little is known about the overlap of health-related search results between search engines. High overlap suggests consumers are likely to see the same group of websites regardless of which search engine they choose. Low overlap suggests that consumers may find a diversity of information through the use of more than one search engine. In 1998, Lawrence and Giles discovered that no search engine indexed more than one-third of the Web, and using six different search engines retrieved 3.5 times more web pages than one search engine alone [16]. However, since 1998, Google introduced large databases created using crawlers with a ranking algorithm based on link-popularity. The success of Google encouraged other search engines to use similar strategies. Further, overlap between search engines in a restricted domain (cancer CAM) may be greater than 1998 estimates which were domain independent. Therefore, we sought to determine the overlap for CAM search results between search engines, and to identify those sites that appear most frequently on the first page of search results.

To summarize, we hypothesized that:

• Consumers are likely to incidentally encounter CAM web pages while searching for cancer information on the Web.

• Sponsored results are more likely to contain CAM information compared to non-sponsored search results.

• The search keyword will influence the number of CAM web pages retrieved.

• There is limited overlap between result sets retrieved by different search engines.