Chronic conditions are among the leading causes of death and disability in the United States (1) and are responsible for a disproportionately large share of health care use and cost (2). New technologies frequently target people with chronic conditions with the hope of increasing system efficiencies and improving patient quality of life. One particularly promising area of innovation has been consumer-oriented health information on the Internet.

Certain attributes of the Internet make it particularly appealing for patients with chronic illnesses. The cost of distributing information on the Web is extremely low, and people in rural areas and those with disabilities can access the same information as people in urban areas and those with no disabilities. Also, compared with printed documents, Internet information can be easily updated to reflect new scientific findings. This has become particularly useful for patients looking for cutting-edge treatments and new clinical trials for chronic illnesses such as human immunodeficiency virus (HIV) and cancer. While the first few generations of Internet sites offering health information consisted primarily of digitized copies of printed materials, developers were quick to exploit the Internet's interactive capabilities. Consumers can now use the Internet to search for risk-assessment tools, interactive health advice, and the latest medical news.

Researchers have begun to examine the potential effects of the Internet on health issues. The low cost of distributing information on the Internet has prompted some researchers to test whether the Internet could be used as a disease-management tool. In two studies, high-quality, disease-specific information was distributed to randomly selected participants, and participants in control groups received information in the traditional form of a health magazine or book (3,4). Study results indicated that the Internet was a better conduit for providing health information. It is not clear, however, which characteristics of the Internet drive its effect on patient education. One study compared a group that received tailored information and personalized feedback via the Internet with a group that received Internet information only and found few differences between the two groups (5).

In addition to disease management, two other areas of Internet health information research have attracted attention. First, studies by Jadad et al (6), Eysenbach et al (7), and Berland et al (8), among others, have shown that the Internet is saturated with both good and bad health information and that consumers are not good judges of quality. Second, studies have described how people are using the Internet for health information. Although the estimates of how many people use the Internet for health have been heavily debated (9-12), questions about how a person's health affects their use of the Internet have not been investigated in as much depth. A recurring finding is that people with depression are more likely than people with other health conditions to use computers to find health information. This finding was noted as part of a community-wide intervention that provided participants with self-care books, a telephone advice line, and computerized health information (13). Compared with participants without depression, participants with depression reported a higher probability of using all three media. More recently, Haviland and colleagues analyzed data from the 2001 Healthcare Market Guide survey and found that people who reported a psychiatric condition (including depression) were more likely to use the Internet to access disease/wellness information than people with no chronic health problem (14). Similarly, a recent national survey by the Pew Internet and American Life Project found that depression, anxiety, or mental-health issues were among the 10 most frequent health-related search topics on the Internet (15). But beyond depression, there is little research from which to identify other emerging themes.

People with chronic conditions have unique needs; this paper investigates Internet use among respondents who reported having hypertension, diabetes, cancer, heart problems, and/or depression. We conducted a nationally representative survey to assess Internet use for health advice and information by individuals with these chronic conditions. Our analysis focused on individuals with at least one of five common chronic illnesses: diabetes, hypertension, cancer, heart problems, and depression. This study addressed three issues. First, we assessed the extent to which people with any of the five chronic illnesses used the Internet for health information. Second, we compared perceptions of the Internet among participants who have one of the five chronic illnesses. Third, we examined respondents' self-reported effects of the information. The sample used in this analysis represents a sub-population of a previously published paper (9).

Table 1 shows the summary statistics for the sample. People who reported depression only were the largest group (27.9% of the sample). People who had hypertension only constituted the second-largest group (25.5%); people who had two chronic conditions made up the third-largest group (22.8%). The smallest group was people who had cancer only (3.0%).

Of the 641 respondents with more than one chronic condition, 201 (31.4%) reported having hypertension, 171 (26.7%) reported having depression, 116 (18.1%) reported having diabetes, 88 (13.7%) reported having heart problems, and 66 (10.3%) reported having cancer (data not shown). Combinations of two and three chronic conditions comprised 93.0% of this group; 39 (6.1%) people reported four chronic conditions, and only six (1.0%) reported having all five conditions.

People with chronic conditions vary in their use of the Internet for health information and advice. After adjusting for education, sex, and age, Internet use for health information in the past year clustered at about 33.8% to 42.9% for hypertension only, cancer only, and heart problems only. Rates were slightly higher (47.6%) for people with any two of the five chronic conditions and were 51.0% to 52.0% for people with diabetes only, depression only, and three or more chronic conditions.

People who have depression only and people who have multiple chronic conditions were among the most frequent users of the Internet for health information — for overall use and for communicating with health professionals. Use of the Internet was also common for people with diabetes only, although these people were less likely to communicate with family and no more or less likely to communicate with health professionals or other patients than the other chronic condition groups. These data confirm past reports that people who have depression are more likely to seek health information than people who have other chronic conditions (13,18). The higher rates for depression only may reflect that the disorder still carries some stigma, leading individuals to seek information outside traditional routes. The higher rates for depression could also reflect that depression care often has greater limits on mental health care insurance benefits and higher out-of-pocket expenses. It would be useful to know whether this association is driven by stigma or by costs. If the association is driven by stigma, this could identify opportunities for using the Internet to reach people with stigmatized and potentially communicable chronic conditions (e.g., HIV).

Although a large percentage of people with depression searched the Internet only for health information, the depression-only group reported one of the lowest rates for having been affected by Internet use. People who had diabetes, cancer, or heart problems were more likely to agree that Internet information improved their understanding of their condition than were people who had hypertension or depression. More research is needed to determine which type of information is received by people who have depression and whether they find it helpful.

We interpret Internet use among people with chronic conditions as a glass half empty or half full. We see smaller effects on treatments and providers and larger effects on self-management, eating, and exercise. Fewer than one in eight people agreed that the Internet led them to seek care from different health professionals for their conditions, and fewer than one in four said that the Internet affected their choice of treatment. These numbers can be viewed as substantial or meager, depending upon perspective. If these numbers are accurate, the effect of the Internet on improved understanding is larger than other computerized patient education interventions, such as the one described in the study by Rostom et al on decision support for hormone replacement therapy (19) or in the study by Consoli et al on hypertension (20). Caution must be used in interpreting these responses because the data are self-reported, and we do not have information about respondents' knowledge before they used the Internet and cannot compare these data to a control group. We also have no way of verifying if the information they obtained was factually correct.

Attitudes toward health information on the Internet were generally favorable. Slightly more than one third of the people with one or more of the five chronic conditions agreed that it takes too long to find information on the Internet, indicating that search time is an important determinant in using the Internet. When people do find information, they then have to identify whether it is high quality and accurate. Approximately one in five people agreed that they cannot trust Internet health information, but it is unclear how they determine whether they can trust the information. Other studies have discussed problems with the quality and coverage of health information on the Internet (6,8,21), and research has found that people are not particularly good judges for identifying accurate health information and often forget which sites they searched (7). Efforts to help people identify high-quality information more quickly could result in more people using the Internet for health information.

A limitation of this study was that all the data were self-reported. Some respondents might have avoided labeling themselves as chronically ill, especially for depression, which is stigmatized. Additionally, the survey questions required that people reconstruct memories of how they used the Internet in the past year. This process can be cognitively difficult, especially when a question asks respondents to remember how they used the Internet and then to estimate its net effect.

The KN panel has been used in other research studies (22,23). This method of sampling departs from traditional random-digit dialing. Both methods start with a sampling frame that consists of U.S. households with telephone access. Both methods have strengths and weaknesses. The strength of the panel approach is that people are asked to participate in the panel, and a subset is sampled for a particular survey. We have information on those who were sent the survey and did not respond. The weakness is that some people may dislike being on a panel and opt out when first asked or ask to be removed from the panel over time. KN and independent researchers have studied these issues and reports are available online (http://www.knowledgenetworks.com/ganp/reviewer-info.html). Random-digit dialing is performed each time a survey is fielded, so it is not susceptible to panel attrition. At the onset of the call, however, the respondent is told about the intent of the survey. People then choose whether to complete the survey, and, typically, little if any information on the non-respondents is collected. Many national surveys, including the Behavioral Risk Factor Surveillance Survey and 2000 Census, report median response rates below 70%. The latest study on the Internet and health conducted by Pew Internet and American Life Project reported a response rate of 32.8% (15). As mentioned earlier, we compared our sample to other national surveys. Although the results were similar on all the variables we compared (17), we cannot rule out the existence of potential biases on other variables, such as Internet use.

This study focuses on common chronic conditions. Perhaps we would see higher rates of Internet use among people who have rare diseases. There is a substantial amount of health information available on the Internet (24), and people with rare illnesses can obtain peer support on the Internet in ways that would not be possible off-line. Further research could evaluate these matters.

A common perception is that the provision of health information via the Internet is a "field of dreams" — that is, if we build it, they will come. In the past decade, public and private investments have poured into Internet sites. Although the Internet can offer several clear advantages over traditional information sources, such as very low distribution costs, we found that few people who have the five common chronic conditions studied use it routinely. When they do, however, they report notable gains in knowledge and small changes in behavior.