Trust and Sources of Health Information:  The Impact of the Internet and Its Implications for Health Care Providers: Findings From the First Health Information National Trends Survey FREE

The environment in which patients consume medical and health information has changed dramatically during the past decade. Rapid diffusion of Internet technology within the public sphere has placed an unprecedented amount of health information within reach of general consumers.^1- 4 Advances in information-tailoring tools and telemedicine have broadened the reach of health care specialists beyond the constraints of the traditional office.^5- 7 Increases in patient advocacy and consumerism have prompted pharmaceutical companies to launch direct-to-consumer advertising campaigns, while entertainment and the news media play host to an ever-increasing breadth of health and illness discussions.⁸ At the same time, concerns over a digital divide, due to the uneven diffusion of health technologies, merit close monitoring by the federal government.⁹

On the clinical side, the appointment in May 2004 of a health information “czar” by the US Secretary of Health and Human Services to coordinate development of a National Health Information Infrastructure crystallized a commitment by the United States to make patients’ medical records universally transportable through electronic means.^10- 11 As a result, data entry and review systems are becoming ubiquitous in hospital settings¹² and health care technologists are crafting applications in medical informatics to improve the quality of health care.^13- 14

Although there has been much speculation, it is unclear exactly how these changes are influencing the ways in which individuals obtain health or medical information and what patterns of trust are associated with use of different information channels.^15- 17 Traditionally, physicians have served a central gatekeeping role in providing health care information and services to their patients.¹⁸ With such offerings as prescription drugs¹⁹ and detailed technical information on prevention and treatment options online,²⁰ the communication dynamic between health care professionals and patients may be changing, potentially resulting in more shared decision making.^21- 24 Most physicians are already experiencing the effects of patients showing up to their offices armed with printouts from the World Wide Web and requesting certain procedures, tests, or medications.^25- 27

To track the dramatic changes that are occurring in the health information environment nationally, the National Cancer Institute (a component of the US National Institutes of Health) fielded the Health Information National Trends Survey (HINTS) for the first time, from October 28, 2002, to April 14, 2003. Herein, we report data from the baseline administration of this biennial survey to generate precise population estimates for the public’s use of, and trust in, different sources of health information.

Data for the study were obtained from the 2002-2003 administration of the HINTS. Details of this survey are published elsewhere.²⁸ Briefly, HINTS is a nationally representative telephone survey of persons 18 years or older, scheduled to be conducted every 2 years. Data from the survey will be used by the National Cancer Institute to monitor changes in the public’s medical information environment over time and to inform the National Cancer Institute’s communication services and research efforts.

Data for the baseline survey were collected between October 28, 2002, and April 14, 2003. A list-assisted design was used to identify household telephone numbers²⁹; telephone exchanges with high concentrations of Hispanic and black residents were oversampled to ensure adequate representation from the 2 largest minority groups in the United States. Response rates based on standards from the American Association for Public Opinion Research³⁰ were 55.0% at the household screening level (ie, the initial contact with the household used for sampling purposes) and 62.8% at the sampled person interview level (ie, completion of the interview by the sampled household member). Those rates are typical of response rates found in other national telephone health surveys.³¹ A total of 6369 persons were included in the final sample.

The survey covered general content areas related to the use of different health information channels, from the television to the Internet; health behaviors; knowledge of health risks; and adherence to guidelines for healthy living. We used the information channel items to generate population estimates for persons who reported going online to access the Internet, perform tasks on the World Wide Web, or use e-mail. Trust in health information sources was assessed based on individual questions about physicians, family or friends, newspapers, magazines, radio, television, and the Internet. Respondents were asked to rate their level of trust for each source as a lot, some, a little, or not at all.

Sociodemographic variables used in this study were age (18-34, 35-64, and ≥65 years), sex, race/ethnicity (Hispanic, non-Hispanic white, non-Hispanic black, and other), education level (less than high school, high school graduate, some college or technical school, and college graduate or more), and annual household income level (<$25 000, $25 000-$49 999, and ≥$50 000). Income bands were based on a trichotomous split of the data into approximately equal segments.

Data were weighted to produce overall and stratified estimates that would be nationally representative of the US population.³² Weights were derived initially from selection probabilities to compensate for planned oversampling procedures. The resulting weights were then calibrated using comparable population characteristics for sex, age, race, and education from data publicly available through the Current Population Survey. A set of 50 replicate weights was produced to allow for analyses that would produce an unbiased estimation of population variance.³³

Analyses were conducted using SAS statistical software, version 8.2 (SAS Institute Inc, Cary, NC) and other computer software (SUDAAN, version 8.0.1); the computer software (SUDAAN) was used to account for the complex survey design and to determine appropriate standard errors and 95% confidence intervals (CIs). Cross-tabulation procedures were used to generate population-based prevalence estimates for online activities across different sociodemographic strata and to investigate levels of trust in different information sources. Logistic regression models were analyzed using computer software (SUDAAN) to investigate the relationship between sociodemographic and health care access measures (independent variables) and level of trust for each of the health information sources (dependent variable). In these models, trust was recoded as a dichotomous variable (a lot or some vs a little or not at all).

To develop a profile of Internet use patterns across the sample, we identified those respondents who indicated they had ever “gone online to access the Internet or World Wide Web, or to send and receive e-mail.” From the HINTS data, we found that an estimated 63.0% (95% CI, 61.7%-64.3%) of adults (or roughly 132 million persons) reported ever using the Internet to access Web sites or to use e-mail, with a substantial majority of ever users (86.8%; 95% CI, 85.5%-88.1%) reporting personal access from home. Roughly two thirds (66.4%; 95% CI, 64.7%-68.1%) of those with home access went online through conventional telephone and modem connections, while almost a third reported going online through broadband connections such as a digital subscriber line (10.2%; 95% CI, 8.9%-11.5%) or cable (22.4%; 95% CI, 20.6%-24.2%). Connection rates through wireless devices (0.4%; 95% CI, 0.2%-0.6%) or some other means (0.5%; 95% CI, 0.2%-0.9%) were negligible.

We then took the online population and generated population estimates for different activities across levels of age, sex, race/ethnicity, education, and income (Table 1). Percentages represent proportions of the online population, those with Internet access, either at home or elsewhere. Overall estimates suggest that almost two thirds, or 63.7%, of adults online looked for some type of health or medical information either for themselves or for someone else through the Internet. Use of the Internet for health or medical information was generally more common among persons who were younger than 65 years, women, those who were white or other race (eg, Asian), and those who had higher levels of education and income.

Use of the Internet for purposes besides health information seeking was much less common, with only 3.9% having participated in an online support group, 7.0% using e-mail to communicate with a physician or physician’s office, 9.1% buying medicine or vitamins online, and 4.7% engaging in some other type of online activity (eg, looking up a physician’s address).

Data on trust of sources for information about cancer are listed in Table 2. Respondents expressed a high level of trust for information provided by physicians, especially in contrast to all the other sources. Trust in the Internet as a health information source was divided, with about one fourth expressing a lot of trust and one fourth expressing no trust. Radio was the least trusted health information source.

Table 3 contains findings from logistic regression analyses in which the dichotomous outcome of high vs low trust was regressed on sociodemographic variables. Several consistent patterns emerged from these analyses. Trust in health information sources was strongly age and sex dependent, with persons aged 18 to 34 and 35 to 64 years and women generally more trusting of most sources. The differences in trust by age are especially pronounced for the Internet: adults aged 18 to 34 years were more than 10 times as likely, and adults aged 35 to 64 years were more than 5 times as likely, as those 65 years or older to report a lot or some trust in the Internet. Level of education was also independently associated with trust of most health information sources: those with higher levels of education (ie, a high school education or greater) were more trusting of the Internet, magazines, and newspapers than persons with less than a high school education.

To examine the relationship between trust and actual behavior, we turned to those items in the HINTS baseline instrument that did the following: (a) asked all respondents where they would go first for information about cancer as a specific disease type and (b) asked those who had looked for cancer information where they actually went. A striking contrast between where people preferred going and where they actually went is illustrated in the Figure.

Figure, A, clearly demonstrates that health care providers and the Internet are the dominant sources people would prefer to use first when seeking information about a specific disease (cancer). When asked where they preferred going for cancer information, 49.5% (95% CI, 48.1%-50.8%) reported wanting to go to their physicians first. There were also strong age-related differences: persons 65 years and older were almost 10 times more likely to go to health care providers first before going to the Internet (75.6% vs 7.7%), whereas persons aged 18 to 34 or 35 to 64 years were almost equally split between health care providers and the Internet, with the Internet being their first choice (38.9% vs 46.6%). Figure, B, demonstrates what happened when health information seekers actually looked for information about cancer. Overall, 48.6% (95% CI, 46.1%-51.0%) reported going to the Internet as a source of first resort, while only 10.9% (95% CI, 9.5%-12.3%) reported going to health care providers first across all ages. Those in the 18- to 34-year age range were almost 9 times more likely to go to the Internet first before going to providers (61.1% vs 7.1%). An almost equal percentage of persons 65 years and older reported going to the Internet first compared with providers first (21.4% vs 20.9%).

The first population estimate of Internet use provided by a federal scientific survey placed penetration rate for Americans 3 years and older at 22.2% in 1997. Subsequent estimates documented a steady climb to 32.7% in 1998, 44.4% in 2000, and 53.9% in 2001.³⁴ Data from the first administration of the HINTS fixed the penetration rate for adults 18 years and older at 63.0% of the US adult population. That estimate is remarkably similar to data reported by the Pew Internet and American Life Project for the same period (also at 63%).³⁵ Subsequent reports by the Pew Research Center suggest that overall penetration has reached a plateau, and is hovering at the 63% mark for adults.³⁶ What has changed is the switch from dial-up to broadband access. About one third of the respondents in the HINTS sample accessed the Internet through high-speed connections in 2003. The most recent data from the Pew Research Center push that estimate up to more than half of the online user population. What is significant about the changeover is that the “always on” availability of broadband access encourages users to engage in new types of activities, including health activities, online.^37- 38

In the HINTS 2003 administration, health information seeking was by far the most commonly reported online activity, with 63.7% of the online population indicating that they had searched for health or medical information for themselves or for others at least once in the previous 12 months. The relatively high percentage of people in the HINTS interviews who reported looking for information for others is indicative of a secondary impact from the Internet, as observed previously by Fox and Rainie.⁴ The implication for physicians is that patients may come into the office with printouts that they did not personally collect; the implication for Web site designers is that health information seekers may be looking for easy-to-print materials to give others.

Those who searched online the most tended to be younger, women, and better educated, and tended to have higher incomes than those who searched the least. These findings parallel conclusions from a recent analysis³⁹ of patient preferences for type of primary care, as documented in the medical literature. From their systematic reviews, the researchers noted that those who are younger and who had higher incomes tended to be more active in their orientation toward primary care, while those who were older and had smaller incomes tended to be more passive. The effect may be generational, with those in the pre–World War II generation acculturated to be passive followers of physicians’ orders, and those in the postwar baby boom acculturated to be active consumers of health care. With the passing of the World War II generation and the “graying of the Baby Boomers,” the trend merits monitoring in future surveys.

Another trend that bears monitoring is the use of the Internet for personal health maintenance kinds of behaviors. The observation is what authors of the Pew Research Center’s Trends 2005 report³⁶ referred to as the new normal, or an incorporation of the Internet into daily mainstream activities,⁴⁰ especially as more and more Americans access the Internet via broadband.^37- 38 In this regard, the estimated percentage of online consumers who bought medicine or vitamins through the Internet was 9.1% (95% CI, 8.0%-10.2%). Although that number is relatively low at this stage, the absolute number is high and bears monitoring in future administrations of the survey. The American Medical Association and the US Food and Drug Administration have expressed concern over online pharmaceutical purchases as substandard medical care.⁴¹

Similarly, the estimated percentage of online Americans who report using e-mail to communicate with their physicians is also relatively low at 7.0% (95% CI, 6.0%-8.0%), a finding borne out in at least 1 other scientific sample.⁴² Again, it is worth monitoring that trend as increases in the use of patient-initiated e-mail will undoubtedly signal accompanying changes in office management practices.⁴³ Efficiency concerns with e-mail may be exacerbated by ongoing and increasing concerns over spam (unwanted junk mail) and phishing (use of bogus e-mail to obtain private information) in the online e-mail environment.⁴⁴

The online behavior with the lowest frequency was participating in online support groups, at 3.9% (95% CI, 2.9%-5.0%) of the online population. That percentage is consistent with data from the National Health Interview Survey regarding the small numbers in the general population who use support groups at any given time.⁴⁵ However, data on what constitutes a successful online support group are mixed, and ongoing research is warranted.⁴⁶

Analyses of the trust and reliance measures present an emerging picture of how the physician’s role may be perceived by the public in the changing health information environment. Respondents’ rating of trust in physicians in the HINTS data exceeded that of any other information channel for health or medical information, and the trend was strongest for respondents who were young, educated, and women. This is the same population that is online. Ongoing attention may be needed to adjust reimbursement policies for time spent with patients interpreting printouts,^47- 49 for accommodating shifts toward informed and shared decision making,^50- 52 for steering consumers to credible information sources,^53- 54 and for attending to the needs of those who fall through the cracks of the digital divide.^8,55- 56

The HINTS data were derived from a nationally representative sample of structured interviews conducted by telephone. The biggest limitations of the study come from the natural liabilities of a random digit–dial telephone survey, the limitations of a self-reported and largely retrospective survey method, the unknown validity of some of the measures at baseline, and diminishing response rates in telephone surveys.³¹ New methods are being explored within the National Cancer Institute to complement federal survey data collection with converging sources of evidence. Techniques such as ecological momentary assessment (or real-time data collection using portable technologies), Web-based surveys, online data capture, and network analysis are being explored as complementary sources of data.⁵⁷

In conclusion, evidence from the first administration of the HINTS interview gives credence to the observation that people are turning to the World Wide Web as an information source of first resort, while relying on health care providers as their most trusted arbiter of information quality. The finding heralds a change in the information role of physicians; physicians may no longer be seen as solitary caretakers but as trusted partners in helping patients sort through information derived from an expanding network of personal and mediated information channels. Although definitive conclusions from the national study are premature—the cross-sectional nature of the survey begs the question of causality—the biennial administration of a survey designed to study the public’s use of health information should serve as an important bellwether for dramatic changes in the national health information environment.

Correspondence: Bradford W. Hesse, PhD, National Cancer Institute, 6130 Executive Blvd, Mail Stop Code 7365, Bethesda, MD 20892-7365 (hesseb@mail.nih.gov).

Accepted for Publication: June 13, 2005.

Financial Disclosure: None.