Cancer Internet Search Activity on a Major Search Engine, United States 2001-2003
Background: To locate online health information, Internet users typically use a search engine, such as Yahoo! or Google. We studied Yahoo! search activity related to the 23 most common cancers in the United States.
Objective: The objective was to test three potential correlates of Yahoo! cancer search activity—estimated cancer incidence, estimated cancer mortality, and the volume of cancer news coverage—and to study the periodicity of and peaks in Yahoo! cancer search activity.
Methods: Yahoo! cancer search activity was obtained from a proprietary database called the Yahoo! Buzz Index. The American Cancer Society's estimates of cancer incidence and mortality were used. News reports associated with specific cancer types were identified using the LexisNexis “US News” database, which includes more than 400 national and regional newspapers and a variety of newswire services.
Results: The Yahoo! search activity associated with specific cancers correlated with their estimated incidence (Spearman rank correlation, ρ = 0.50, P = .015), estimated mortality (ρ = 0.66, P = .001), and volume of related news coverage (ρ = 0.88, P < .001). Yahoo! cancer search activity tended to be higher on weekdays and during national cancer awareness months but lower during summer months; cancer news coverage also tended to follow these trends. Sharp increases in Yahoo! search activity scores from one day to the next appeared to be associated with increases in relevant news coverage.
Conclusions: Media coverage appears to play a powerful role in prompting online searches for cancer information. Internet search activity offers an innovative tool for passive surveillance of health information–seeking behavior.
J Med Internet Res 2005;7(3):e36)
Health care providers [- ] and their patients [ - ] regularly search for health information online. Internet users generally begin looking for health information using a search engine [ - ], an Internet tool that searches for Web pages containing terms specified by users [ ]. In February 2004, Google and Yahoo! were the most popular search engines in the United States, with 38% and 32% of market share, respectively [ ].
To date, few studies of Internet search activity related to health topics have been published. Most analyzed the proportion of health and non-health searches on various search engines and found that health searches constituted a small percentage of total searches [- ]. This finding is not surprising given how infrequently Internet users search for health information compared with how often they look for news reports, product information, and other topics [ ]. In any case, a small percentage of total Internet searches translates into thousands of health searches each day [ ], and an estimated 95 million Americans have used the Internet at least once to search for health information [ ].
Cancer appears to be a health topic of some interest to Internet users. Eysenbach and Köhler  found that searches for cancer information accounted for 5% of health-related searches on the search engine MetaCrawler. Phillipov and Phillips [ ] found that “skin cancer” was one of only 17 health-related search terms among the most popular 300 Internet keywords identified by Wordtracker, a private research company. Bader and Theofanos [ ] studied cancer searches conducted using the search engine AskJeeves during a 3-month period and found the most commonly searched cancers were digestive/gastrointestinal/bowel, breast, and skin. This study also compared the incidence of selected cancers with their associated search activity. While this relationship was not statistically tested, the authors observed that some rarer cancers constituted a higher proportion of total searches than their proportion of total cancer incidence. In addition, Bader and Theofanos, as well as Phillipov and Phillips, noted that media coverage appeared to prompt Internet searches for health topics, but they did not systematically investigate or test the relationship.
The study reported here builds on this prior work by analyzing cancer-related searches conducted in the United States from 2001 to 2003 using the search engine Yahoo! Specifically, we investigated three potential correlates of Yahoo! cancer search activity—estimated cancer incidence, estimated cancer mortality, and the volume of cancer news coverage. Cancers that afflicted more individuals, claimed more lives, and generated more news coverage were expected to be associated with more Internet search activity than other cancers, given the interest generated by relevance and publicity. In addition, we assessed the periodicity of Yahoo! cancer search activity and examined sharp increases in Yahoo! search activity related to specific cancer types.
This analysis included three types of 2001–2003 US data: Yahoo! cancer search activity, cancer burden (estimated incidence and mortality), and cancer news coverage. The study protocol was reviewed by the Institutional Review Board of the National Center for Chronic Disease Prevention and Health Promotion and was designated as “research not involving human subjects.”
Yahoo! Cancer Search Activity
During 2001 (the beginning of the study period), Yahoo! was the most popular US search engine, with a 49% market share; however, in 2003 (the end of the study period), Google surpassed Yahoo! as the leading US search engine . Yahoo! remains a widely used search engine; more than 45 million US Internet users conducted Yahoo! searches in February 2004 [ ].
Yahoo! maintains a database of search activity called the Yahoo! Buzz Index . This index provides a search activity score for individual search terms—the words or characters that users type into the Yahoo! search box. Each point of a Yahoo! Buzz Index score equals 0.001% of users searching Yahoo! during the time period of interest (day, week, or month). For example, if 250 out of a total of 1 million users searched for “breast cancer” on January 1, 2001, the Yahoo! “breast cancer” search activity score on this day would be 25 (250/1 million × 100000). For a search term to register a search activity score, it must generate at least 50 searches during the time period for which the score is calculated. Yahoo! search activity generated by search terms can be segmented by country, US state, or selected US cities. This study used daily US Yahoo! search activity data from January 1, 2001 (the earliest date for which Yahoo! search activity data are available) to December 31, 2003. We limited our analysis to Yahoo! searches because at the time of this study no other Internet search engine offered a dataset like the Yahoo! Buzz Index, which provides search activity scores adjusted for the size of the population searching for online information.
Yahoo! employs professional “surfers” or content indexers who manually classify Web pages into one of more than 2000 content categories, such as “movies,” “footwear,” “astrology,” or “cancer or neoplasms.” The Yahoo! Buzz Index classifies search terms in the same content category as the first Web page link that a user “clicks” or activates after conducting a search. For instance, if a user entered the search term “colon” and then clicked on a cancer website, “colon” would be classified as a “cancer or neoplasms” search term. If the user clicked on a grammar website, however, “colon” in that instance would be classified as an “education” search term. When a user does not click on a Web page link or when a user clicks on a Web page link that has not been classified, the Yahoo! Buzz Index categorizes the search term using a variety of algorithms that analyze recent content viewed by the user.
To identify commonly used Internet search terms related to specific cancers, we reviewed the search terms classified in the “cancer or neoplasms” category of the Yahoo! Buzz Index, which generated at least 50 searches in any month from January 2001 to December 2003. This search strategy identified 76 unique search terms, of which 23 were included in the analysis (). The remaining 53 terms were discarded because they did not relate to a specific cancer or included non-English words. Discarded terms included drug names or treatment modalities, such as “chemotherapy” (n = 19); the names of organizations or events, such as “Race for the Cure” (n = 16); general cancer or anatomy terms, such as “oncology” (n = 11); names of celebrities, such as “Gilda Radner” (n = 5); and the carcinogen “DES” (n = 1). Also, “leucemia” (n = 1), the Spanish word for “leukemia,” was discarded because the Yahoo! Buzz Index does not consistently track foreign words, as it excludes search terms that contain non-English characters. For instance, the Yahoo! Buzz Index would not capture a search term with an accent mark, such as “cáncer colorectal” (Spanish for “colorectal cancer”).
The estimated incidence and mortality for 23 cancers during the study period were obtained from annual burden reports published by the American Cancer Society [- ]. All cancers with at least 7500 new cases in 2001, 2002, or 2003 were included in the analysis (n = 23) whether or not they were associated with Yahoo! search activity.
Cancer News Coverage
News reports associated with specific cancer types were identified using the LexisNexis “US News” database, which includes more than 400 national and regional newspapers, such as the Wall Street Journal and the Baltimore Sun, and a variety of newswire services, such as the Associated Press and UPI (United Press International). News reports published from January 1, 2001, to December 31, 2003, related to specific cancer types were found by locating reports with the identified Yahoo! search activity terms (eg, “breast cancer”) in their headlines. In the case of cancers located in the esophagus and oral cavity, for which no Yahoo! search activity terms were associated, the terms “esophageal cancer” and “oral cancer” were used as the headline search terms.
Descriptive statistics were calculated for the Yahoo! search activity score, estimated incidence, estimated morality, and news coverage volume associated with the cancers included in the study. Spearman rank correlations were used to establish the consistency of these variables across the study period, and the data were aggregated. Next, the relationships between Yahoo! search activity and the potential correlates of interest were tested using Spearman rank correlations.
The relationship between Yahoo! search activity and news coverage was further analyzed for the five cancers with the highest daily Yahoo! search activity. The number of news reports published each day was transformed into a categorical variable with four levels (0 news reports, 1–2 news reports, 3–4 news reports, and 5 or more news reports), and one-way analysis of variance (ANOVA) was used to compare mean daily Yahoo! search activity scores at increasing levels of news coverage. To detect possible periodicity effects, Yahoo! search activity data were visually inspected. Three possible periodicity effects were noted: a rise during weekdays (Monday–Friday) compared with weekends; a rise during national cancer awareness months compared with other months; and a decline during summer months (June–August) compared with other months. These possible effects were tested using t tests. Finally, the Yahoo! search activity associated with several cancers was marked by sharp increases of 100% or more from one day to the next. For these cancers, we identified the three highest peaks in 2003 Yahoo! search activity and investigated corresponding news events.
We found Internet search terms generating Yahoo! search activity associated with 21 of the 23 cancers included in the study (). Of these, 19 cancers were associated with only one Yahoo! search term each. The 2 remaining cancers were associated with two search terms each: cancers of the brain (“brain tumor” and “brain cancer”) and multiple myeloma (“multiple myeloma” and “myeloma”). In these cases, the daily Yahoo! search activity scores associated with each term were summed into a composite score for these cancers.
The highest mean daily Yahoo! search activity scores were generated by breast cancer (mean = 14.37), lung cancer (mean = 9.08), and leukemia (mean = 7.15). Cancers with the highest US 2001–2003 incidences were breast (n = 611300), prostate (n = 608000), and lung (n = 510800). For cancer mortality, lung (n = 469500), colorectal (n = 170400), and breast (n = 120800) cancer were the leading causes of death. Breast cancer (n = 5840), leukemia (n = 2143), and prostate cancer (n = 1822) were associated with the most US news reports from 2001 to 2003. Some cancers, such as leukemia, ovarian, and testicular, appeared to be associated with more Internet search activity than their burden would dictate.
Cancers were ranked by mean daily Yahoo! search activity score, estimated incidence, estimated mortality, and number of related news reports for each year in the study period (2001 to 2003). To explore the consistency of ranks over the study period within each variable, Spearman rank correlations were determined for each pair of years (2001 and 2002, 2002 and 2003, 2001 and 2003). We found statistically significant correlations (P < .001) between all year pairs tested (data not shown). Because the ranks associated with these variables were highly consistent from 2001 to 2003, we aggregated the data across the study period.
Correlates of Yahoo! Cancer Search Activity
We tested the relationships between variables by determining Spearman rank correlations between each pair. Statistically significant correlations were found between all variable pairs ().
The relationship between Yahoo! search activity and its most statistically significant correlate—news coverage—was further analyzed for the five cancers with the highest daily Yahoo! search activity (breast, lung, leukemia, colorectal, and prostate). For these cancers, the number of news reports published each day was transformed into a categorical variable with four levels. The mean daily Yahoo! search activity at each level was compared using ANOVA, and all tests were statistically significant (P ≤ .001). For all five cancers, the mean daily search activity score was higher at each increasing level of news coverage ().
Periodicity of Yahoo! Cancer Search Activity and News Coverage
Three possible periodicity effects were tested: a rise during weekdays (Monday–Friday) compared with weekends; a rise during national cancer awareness months compared with other months; and a decline during summer months (June–August) compared with other months. To test for these potential effects, we used the five cancers with the highest daily mean Yahoo! search activity scores (breast, lung, leukemia, colorectal, and prostate) (). For all five cancers tested, both mean daily Yahoo! search activity scores and mean daily news reports were higher Monday–Friday than they were Saturday–Sunday (P < .001). Higher mean daily Yahoo! search activity scores were found for breast cancer (P < .001), lung cancer (P < .001), and colorectal cancer (P < .001) during their respective national awareness months. The number of mean daily news reports related to breast cancer (P < .001), colorectal cancer (P < .001), and prostate cancer (P = .007) rose during their respective national awareness months. Mean daily Yahoo! search activity scores for breast cancer (P < .001), lung cancer (P < .001), and leukemia (P < .001) were lower during the summer months than during the rest of the year. While mean daily news reports about breast cancer also decreased during the summer (P < .001), mean daily news reports about prostate cancer rose (P = .01).
Peaks in Yahoo! Cancer Search Activity and News Coverage
On several occasions, Yahoo! search activity scores associated with breast cancer, colon cancer, and prostate cancer were marked by sharp increases of 100% or more from one day to the next. We investigated news events that corresponded with the highest three spikes in 2003 Yahoo! search activity for these cancers. These peaks in “breast cancer” and “colon cancer” search activity all occurred during their respective national awareness months and appeared to be related to news coverage promoting the awareness months. The highest peak in “prostate cancer” search activity (22.34) occurred on July 17 after news reports of a study  exploring the association between sexual behavior and prostate cancer risk ( ). These news reports generally focused on the possible protective benefit of masturbation. This study was not covered widely by the US news media, but it generated substantial news coverage in Australia and filtered onto the Internet via chat rooms, message boards, and medical news Web pages. While there was no corresponding spike in “masturbation” search activity, there was a 117% increase in the search activity score (from 61.88 on July 16 to 133.08 on July 17) for “masterbation,” a common misspelling. The second highest spike in “prostate cancer” search activity (14.59) occurred on October 21 after news broke that Academy-Award-winning actor Robert DeNiro had been diagnosed with prostate cancer. This story was widely covered by the US news media, and a 277% increase in “Robert DeNiro” search activity was observed on the same date (from 15.87 on October 20 to 59.90 on October 21). The third highest peak in “prostate cancer” search activity (12.41) occurred on December 29, when a study linking obesity with increased prostate cancer risk [ ] was covered by several US news outlets. No corresponding rise in searches for the terms “obesity,” “overweight,” or “weight loss” was observed.
This study suggests that media coverage plays a powerful role in prompting online cancer information seeking. News coverage correlated significantly with Yahoo! search activity (P < .001). Also, Yahoo! search activity was found to rise as news coverage increased, and sharp rises in search activity from one day to the next appeared to be associated with increases in relevant news coverage. This study also suggests that the Internet can rapidly disseminate health news: the highest spike in 2003 US “prostate cancer” Yahoo! search activity seemed to be generated largely by news coverage in Australia that rapidly filtered onto the Internet via chat rooms, message boards, and medical news Web pages. Thus, it possible that a news story does not necessarily have to be covered by the US news media in order to generate US Internet search activity.
News coverage volume also correlated with estimated cancer incidence (P = .015) and mortality (P < .001). This is interesting because past studies [- ] on this topic have not generated consistent findings, with most [ - ] finding no relationship between disease burden and news coverage volume. However, none of the past studies focused on cancer, and none used our method for identifying news reports. While the news coverage of specific cancers generally matched their burden, we noted that some cancers, such as leukemia, ovarian cancer, and testicular cancer, were associated with more Internet search activity than their burden would suggest. A similar observation was reported by Bader and Theofanos [ ], who suggested that this discrepancy may result from more searches being required to locate online information about less common cancers. The high correlation between cancer-specific news coverage and associated online search activity in the present study suggests another explanation: some cancers received a disproportionate share of news coverage relative to their incidence and mortality, and online search activity, often prompted by news coverage, reflects this imbalance.
We detected several periodicity effects in US Yahoo! cancer search activity, which tended to be higher on weekdays and during national cancer awareness months but lower during the summer months. It should be noted that these observations are not artifacts of the size of the online population during these periods because Yahoo! search activity scores are based on the percentage, not the number, of total users. One explanation for these results is that the volume of cancer news coverage tended to follow these trends. It is also possible that users tend to search for online cancer information from school or work settings. As a result, Yahoo! cancer search activity would be expected to drop during weekends when people are at home and over the summer months when many students are out of school and many workers go on vacation.
Although Yahoo! is a leading US Internet search engine, the extent to which the findings of this study can be generalized to other search engines is not known. Also, we were unable to discern the motivations of Yahoo! users searching for cancer information. For instance, news coverage of a breast cancer drug might be associated with an increase in “breast cancer” search activity. While the Yahoo! Buzz Index would detect this rise, it cannot tell how many searchers were breast cancer patients or family members and how many were investors interested in buying stock in the company developing the drug.
Internet search activity offers an innovative tool for passive surveillance of health information–seeking behavior. While our work focused on cancer, Internet search activity may be useful in gauging health information seeking related to other diseases. For example, the volume of Internet searches related to symptoms or conditions might be used to predict disease outbreaks (eg, influenza) or to assess mental health following a disaster. Researchers at the Centre for Global eHealth Innovation have begun to experiment with analyses of this type [, ]. The Yahoo! Buzz Index is unique among Internet search datasets because it provides search activity scores adjusted for the size of the population searching for online information, which has steadily grown each year [ ]. Perhaps in the future, other Internet search engines will offer databases similar to the Yahoo! Buzz Index, and research could be conducted using a combination of search engines.
The Centers for Disease Control and Prevention funded all aspects of this study, including design; collection, management, analysis, and interpretation of the data; and preparation, review, and approval of the manuscript. Yahoo! Inc provided technical support as part of the purchase of a Yahoo! Buzz Index subscription. Use of trade names is for identification only and does not imply endorsement by the US Department of Health and Human Services.
Conflicts of Interest
One of the authors, KP Mallon, was employed by Yahoo! Inc at the time of this study and owns Yahoo! corporate stock.
- Sigouin C, Jadad AR. Awareness of sources of peer-reviewed research evidence on the internet. JAMA 2002 Jun 5;287(21):2867-2869. [Medline]
- Jadad AR, Sigouin C, Cocking L, Booker L, Whelan T, Browman G. Internet use among physicians, nurses, and their patients. JAMA 2001 Sep 26;286(12):1451-1452. [Medline] [CrossRef]
- Casebeer L, Bennett N, Kristofco R, Carillo A, Centor R. Physician Internet medical information seeking and online continuing education use patterns. J Contin Educ Health Prof 2002;21(1):33-42. [CrossRef]
- Baker L, Wagner TH, Singer S, Bundorf MK. Use of the Internet and e-mail for health care information: results from a national survey. JAMA 2003 May 14;289(18):2400-2406 [FREE Full text] [CrossRef] [Medline]
- Diaz JA, Griffith RA, Ng JJ, Reinert SE, Friedmann PD, Moulton AW. Patients' use of the Internet for medical information. J Gen Intern Med 2002 Mar;17(3):180-185. [Medline] [CrossRef]
- Murray E, Lo B, Pollack L, Donelan K, Catania J, White M, et al. The impact of health information on the internet on the physician-patient relationship: patient perceptions. Arch Intern Med 2003 Jul 28;163(14):1727-1734. [CrossRef] [Medline]
- Fox S, Fallows D. Internet health resources: health searches and email have become more commonplace, but there is room for improvement in searches and overall Internet access. Washington, DC: Pew Internet & American Life Project; 2003. URL: http://www.pewinternet.org/pdfs/PIP_Health_Report_July_2003.pdf
- Eysenbach G, Köhler C. How do consumers search for and appraise health information on the world wide web? Qualitative study using focus groups, usability tests, and in-depth interviews. BMJ 2002 Mar 9;324(7337):573-577 [FREE Full text] [Medline]
- Fox S, Rainie L. Vital decisions: how Internet users decide what information to trust when they or their loved ones are sick. Washington, DC: Pew Internet & American Life Project; 2002. URL: http://www.pewinternet.org/pdfs/PIP_Vital_Decisions_May2002.pdf
- Frisby G, Bessell TL, Borland R, Anderson JN. Smoking cessation and the Internet: a qualitative method examining online consumer behavior. J Med Internet Res 2002 Nov 22;4(2):E8 [FREE Full text] [Medline] [CrossRef]
- O'connor JB, Johanson JF. Use of the Web for medical information by a gastroenterology clinic population. J Amer Med Assoc 2000;284(15):1962-1964. [CrossRef]
- Peterson G, Aslani P, Williams KA. How do consumers search for and appraise information on medicines on the Internet? A qualitative study using focus groups. J Med Internet Res 2003 Dec 19;5(4):e33 [FREE Full text] [Medline] [CrossRef]
- Wilkas LR. Scientific inquiry: an introduction to search engines. J Soc Pediatr Nurs 2001;6(3):149-151. [Medline]
- Lynch S. Google tops search sites as users flock to features. New York Post 2004 Feb 24:32.
- Eysenbach G, Kohler Ch. What is the prevalence of health-related searches on the World Wide Web? Qualitative and quantitative analysis of search engine queries on the internet. AMIA Annu Symp Proc 2003:225-229. [Medline]
- Eysenbach G, Köhler C. Health-related searches on the Internet. JAMA 2004 Jun 23;291(24):2946. [Medline] [CrossRef]
- Spink A, Yang Y, Jansen J, Nykanen P, Lorence DP, Ozmutlu S, et al. A study of medical and health queries to web search engines. Health Info Libr J 2004 Mar;21(1):44-51. [Medline] [CrossRef]
- Phillipov G, Phillips PJ. Frequency of health-related search terms on the Internet. JAMA 2003 Nov 5;290(17):2258-2259. [CrossRef] [Medline]
- Madden M. America's online pursuits: the changing picture of who's online and what they do. Washington DC: Pew Internet & American Life Project; 2003. URL: http://www.pewinternet.org/pdfs/PIP_Online_Pursuits_Final.PDF
- Fox S. Health information online: eight in ten Internet users have looked for health information online, with increased interest in diet, fitness, drugs, health insurance, experimental treatments, and particular doctors and hospitals. Washington DC: Pew Internet & American Life Project; 2005. URL: http://www.pewinternet.org/pdfs/PIP_Healthtopics_May05.pdf
- Bader JL, Theofanos MF. Searching for cancer information on the internet: analyzing natural language search queries. J Med Internet Res 2003 Dec 11;5(4):e31 [FREE Full text] [Medline] [CrossRef]
- Elgin B. The search war is about to get bloody. Business Week 2003 Jul 28;3843:72.
- . Home page. URL: https://buzz.yahoo.com/client/ [accessed 2005 Jun 27] [WebCite Cache]
- . Cancer Facts & Figures 2001. Atlanta, GA: American Cancer Society Inc; 2001.
- . Cancer Facts & Figures 2002. Atlanta, GA: American Cancer Society Inc; 2002.
- . Cancer Facts & Figures 2003. Atlanta, GA: American Cancer Society Inc; 2003.
- Giles GG, Severi G, English DR, Mccredie MRE, Borland R, Boyle P, et al. Sexual factors and prostate cancer. BJU Int 2003 Aug;92(3):211-216. [Medline]
- Irani J, Lefebvre O, Murat F, Dahmani L, Doré B. Obesity in relation to prostate cancer risk: comparison with a population having benign prostatic hyperplasia. BJU Int 2003 Apr;91(6):482-484. [Medline]
- Adelman RC, Verbrugge LM. Death makes news: the social impact of disease on newspaper coverage. J Health Soc Behav 2000 Sep;41(3):347-367. [Medline]
- Van Der Wardt EM, Taal E, Rasker JJ, Wiegman O. Media coverage of chronic diseases in the Netherlands. Semin Arthritis Rheum 1999 Apr;28(5):333-341. [Medline] [CrossRef]
- Frost K, Frank E, Maibach E. Relative risk in the news media: a quantification of misrepresentation. Am J Public Health 1997 May;87(5):842-845. [Medline]
- Kristiansen CM. Newspaper coverage of diseases and actual mortality statistics. Eur J Soc Psychol 1983;13:193-194.
- Combs B, Slovic P. Newspaper coverage of causes of death. Journalism Quart 1979:837-843.
- Eysenbach G. SARS and population health technology. J Med Internet Res 2003 Jun 30;5(2):e14 [FREE Full text] [Medline] [CrossRef]
- Eysenbach G. Infodemiology: toward methods and measures for understanding and quantifying the distribution, determinants, and impact of information. Proc AMIA Annu Fall Symp 2005 [submitted].
Edited by G. Eysenbach; submitted 29.04.05; peer-reviewed by J Jansen; comments to author 18.05.05; revised version received 07.06.05; accepted 12.06.05; published 01.07.05
© Crystale Purvis Cooper, Kenneth P Mallon, Steven Leadbetter, Lori A Pollack, Lucy A Peipins. Originally published in the Journal of Medical Internet Research (http://www.jmir.org), 1.7.2005. Except where otherwise noted, articles published in the Journal of Medical Internet Research are distributed under the terms of the Creative Commons Attribution License (http://www.creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited, including full bibliographic details and the URL (see "please cite as" above), and this statement is included.