Millions of people are turning to the Internet for health-related information, and the information they find often directly affects the decisions they make about their health care. A recent study estimated that 73 million Americans, or 62% of Americans with access to the Internet at the time, have used the Internet to search for medical information or other information relevant to their health care [1]. Another study estimated that 12.34 million health-related Web searches are conducted worldwide every day [2]. People often turn to the Internet for health and medical information because of its convenience and for the opportunities it provides for communication [3]. Many people use the Internet to supplement the information given to them by their physician [1,4], to become more involved in their health care decisions [5], or to become more educated about their condition or the condition of a friend or relative.

The medical information that people find on the Internet may affect their health care decisions, health status, and mental status. Two recent studies reported on Americans who have searched the Internet for medical information, termed "health seekers." Ninety-two percent of health seekers reported that the health information they obtained during their last Internet search was useful and relevant [4], and 68% reported that the medical information they found on the Internet played an influential role in the health care decisions they made for themselves or their loved ones [1]. In 2002, 62% of health seekers said the information they found online improved how they cared for themselves, up from 48% in 2000 [1]. People may be able to meet some of their psychological needs related to their health care by using Internet-based tools to answer their questions [3]. Interacting with others online may also lead to improved health status and decreased health care utilization, according to a recent study that examined e-mail discussion groups [5].

Research suggests that the Internet has revolutionized the way patients access health care information, learn more about their conditions, and make health care decisions. Little data exists, however, about the ways that patients actually use the Web [6,7], or about patients' information needs when they use the Internet, such as the "reasons behind online information-seeking" and the "behavior of health users" on the Internet [6]. To explore patients' information needs when they consult the Internet, this study analyzes what people search for when they use a patient-education Web site operated by the Department of Orthopaedics and Sports Medicine at the University of Washington Medical Center. Additionally, it determines who is performing these searches, and it compares questions submitted by Web site visitors from around the world.

Recent studies asking similar questions have analyzed the topics of e-mails sent to doctors [8,9] and the data from surveys asking about patients' Web searches for medical information [10]. Our study builds on this previous work by investigating the actual searches performed by one specific set of Internet users—those who visit the University of Washington Department of Orthopaedics and Sports Medicine Web site. The University of Washington Department of Orthopaedics and Sports Medicine, as well as other organizations providing health information Web sites, can use these data about these users to develop their understandings of the information needs of patients in general, and to improve their Web sites to provide effective educational materials for their users. In turn, patients who are knowledgeable about their health can participate more actively with their physicians in determining a health care plan that is best suited for them.

Content analysis was performed on free-text queries submitted to the Web site of the University of Washington Department of Orthopaedics and Sports Medicine [11], which features multimedia information on arthritis, orthopaedics, and sports-medicine topics. The Web site, which has been operating since 1995, provides articles and videos on over 200 topics to an average of 4000 visitors a day worldwide. The goal of the Web site is "to offer patient-education materials that support users' self-directed learning and help them answer their questions" [12]. In addition to browsing the Web site, visitors can query it in 3 ways: (1) by conducting a simple keyword search, (2) by e-mailing the Web site manager, or (3) by searching the Web site with the Ask a Question function (see Figure 1 and Figure 2 for screenshots of the search box page and the search results page). This search function was built specifically for this Web site, after informal analysis of searches performed in the keyword box revealed a high number of long queries resembling questions. This study analyzes the queries users submitted to the Ask a Question function, because that function encourages the users to describe what they are looking for in more detail than the small keyword search box does.

When using the Ask a Question function, users type free-text questions into a search box that is large compared to typical keyword search boxes. A simple computer program built for the Ask a Question function parses these free-text questions to make them compatible with the program's search algorithm. The users receive search results consisting of similar questions to which the Web site already offers answers. They can then choose which question/answer is the most relevant or similar to their information need. Implemented in November 2001, the Ask a Question system handles more than 1000 questions each month.

These results indicate that when people turn to the University of Washington Orthopaedics and Sports Medicine Web site with health questions, their most frequent motives are to find information about a condition, treatment, or symptoms. In our data set, conditions, treatments, and symptoms were the most frequent topics whether the purpose of the query was to seek information, seek advice, or share information. These results are similar to those of other studies suggesting that that the most frequent reasons people use the Internet for medical information are to seek information and to seek advice about topics such as conditions, treatments, and symptoms [8,10].

One similar study, conducted by Eysenbach et al, analyzed 209 unsolicited e-mails sent to physicians through a university hospital Web site, and found that 34% of the e-mails contained requests for general information about a condition or disease, while 75% of the e-mails contained specific questions, most frequently pertaining to treatment options, specialist referrals, alternative treatments, and whether a condition was curable [8]. The main topics asked by Web site users in our study were similar to those asked in Eysenbach's study. Conditions and treatments were the 2 most frequent topics in our study, while physicians and prognosis were present though less frequent (see Table 3). However, our findings differed from Eysenbach's with respect to the "purpose" of the requests: of the 793 search queries analyzed in our study, 73% were asking for information, and 23% were seeking advice (see Table 2). This variation is likely due to the different types of questions analyzed in the 2 studies. In Eysenbach's study, the objects of analysis were e-mails that patients sent to the physicians responsible for the Web site. Our study, on the other hand, analyzes questions that were submitted as a search method on a Web site, not questions that were sent in e-mail messages. Thus, we are analyzing what patients seek when they are anonymously searching a health information Web site, rather than what they seek when they make the effort to formulate a question and e-mail it to the Web site owners.

Another study with similar results as ours (but different methods) was conducted by O'Connor et al [10]. O'Connor's study investigated the information that patients of a gastroenterology clinic sought when using the Internet. By analyzing questionnaires that patients filled out when they visited the clinic, the researchers found that of the 462 patients who reported having access to the Internet, 51% had used the Internet within the last year to search for medical information. Of the patients who had utilized the Internet for medical information, 31% sought general disease information, 23% sought information about treatment options, 18% inquired about medications, 14% sought information about diet and nutrition, and 10% inquired about alternative medicine. Our results show similar trends: 32% of the queries in our study asked about a condition or disease, 21% asked about treatment, 5% asked about medications, and 1% asked about diet (see Table 3). However, because the data in O'Connor's study were derived from responses to questionnaires, their results describe what patients perceive and report about their Internet use, which may differ from their actual behavior [2,7,19]. In contrast to the study by O'Connor et al, our study examines the log files of a health information Web site to analyze the actual questions that Web site visitors ask, rather than relying on patients to report their behavior while at the doctor's office.

The results of our study suggest that people searching the Internet for medical information that pertains to their health (or the health of a friend or relative) may have different information needs than users searching for medical information that does not apply to their health (or the health of a friend or relative). The inquiries submitted by askers who identified themselves as the patient in question (or a friend or relative of the patient in question) did not follow the overall trend of seeking information. As shown in Table 10, when the relationship between the asker and patient could be deduced, the most frequent purpose of the inquiry was to seek advice, whereas when the relationship between asker and patient could not be determined, the most frequent purpose was to seek information.

These results also suggest that Web site visitors from around the world have similar goals when turning to the Internet for health-related information. The 3 most frequent types of inquiries submitted from the United States were not statistically different from those submitted from all other countries combined ( P= .92) (see Table 14). Thus, among all the Web site visitors from around the world, the most frequent searches performed were to seek information about a condition, treatment, or symptoms. These results indicate that the global audience of the University of Washington Orthopaedics and Sports Medicine Web site has similar information needs, regardless of geography.

The results of this study will inform the future development, design, and content of the University of Washington Orthopaedics and Sports Medicine Web site. The Web site managers aim to improve the site's usefulness, and this data about the users' information behavior and needs will help the managers improve it specifically for the Web site's audience. It appears that the focus of the Web site is already in line with the main needs of its users, but this study has identified several areas in which user interest outstrips the Web site's coverage. First, although there are many articles about various conditions, the Web site can be improved in the areas of treatment and symptom information for orthopaedic conditions, which were the second and third most frequent topics for all types of questions. Second, "treatment" in general was the second most frequent topic asked about by users, but the Web site focuses on surgical interventions and offers few articles on nonsurgical treatments. Thirdly, the Web site does not yet provide any articles that help patients learn more about specific symptoms. Of the queries that were seeking advice, symptoms were the most frequent topic. This suggests that the articles focusing on symptoms might need to be written differently than articles about conditions or treatments (which were the most frequent topics when users sought information). Another way to respond to the findings of this study might be to redesign the Web site's user interface to reflect the topics for which the users search most often. The current design features 10 main labels: 5 body parts (eg, "shoulder & elbow") and 5 orthopaedic specialties (eg, "rheumatology"). Based on this study's findings, the Web site designers are considering changing the user interface to feature labels that are more in line with the topics for which its users search, such as "conditions," "treatments," and "symptoms." These are just a few of the ways that the Web site can be adapted to better meet the needs of its users, based on this data about what patients are really interested in learning.

This study has several possible limitations. First, the development of the coding system and the coding of the queries were performed by the same person. Krippendorf [14] suggests that it is less than ideal for the developer of a code to also be the administrator; however, he concedes that it is acceptable when resources are limited. We determined the inter-rater and intra-rater reliability measures to diminish and be aware of the possible effects of this limitation.

A second limitation is that the inter-rater reliability was determined by coding queries according to their broad topic category (refer to Table 1) rather than by specific category. Once moderate agreement was achieved with respect to the broad topic categories (eg, "condition, symptoms, and/or an injury"), the lead author went a step further to determine each query's specific topic (eg, "injury"). However, no inter-rater or intra-rater reliability measures were determined for this step.

A third limitation of this study pertains to the specific set of Internet users considered. Because our data consisted of questions submitted to a health education Web site focused on arthritis, orthopaedics, and sports-medicine information, most individuals submitting the questions were probably interested in information relating to these topics. Therefore, it may be difficult to use these results to make claims about how people in general use the Internet to search for any type of medical information. Moreover, this study focused on user information needs based on search queries, which excludes people who browse the Web site instead of searching it.

A fourth limitation is a common drawback of log-file analysis. When examining the IP addresses of the computers used to submit questions, we cannot know if the people who submitted the questions were actually from the countries in which their computers were operating. For example, questions from computers not based in the United States could have been American travelers or Americans who choose to live abroad, just as questions from United States-based computers could be from exchange students, travelers, or other people not from the United States. Thus, the distinction made between questions from United States-based askers and those from other countries may not have reflected the actual identity of the askers.