| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||||||||||
Correspondence: Address correspondence to Lisa Smith Wagner, PhD, Department of Psychology, University of San Francisco, 2130 Fulton Street, San Francisco, CA 94117-1080. E-mail: wagnerl{at}usfca.edu
 |
Abstract |
|---|
 TOP Abstract MethodsResultsDiscussionReferences |
|---|
Key Words: Health information • Older adults • stereotype • Medical informatics • Community health • Internet
The supply of health information has also exploded in the past decade, primarily because of the Internet. Numerous Internet sites have sprung up to provide information—both disease- or condition-specific information and more global health and wellness information. The Internet has great promise, as it has the potential to provide information to underserved populations: people with disabilities, people in remote areas, and people with socially stigmatized conditions, such as urinary incontinence (Nahm & Resnick, 2001; Roan, 2000). In addition to computerized sources of information, health handbooks, newsletters, and advice nurses have also become more prevalent. As Glied (1999) noted, the first edition of Dr. Spock's book sold 15 million copies in 1957, and the sixth edition sold 40 million copies in 1992.
Older adults are more likely than other age groups to benefit from this increase in health information. First, older adults may be more motivated to get health information because they are at greater risk than younger adults for illnesses, both chronic and acute. Second, because of retirement or a reduced workload, older adults may have more time than younger adults to examine health information. Finally, older adults may have only limited access to other health care sources; this difficulty in reaching a provider may be the result of limited mobility, remote location, or loss of social networks (Nahm & Resnick, 2001; Roan, 2000).
Although health information may be particularly pertinent for older adults, there is the perception that older adults will not use it. This perception is especially true for computerized information (Marwick, 1999). In part, this belief is due to stereotypes that older adults are resistant to change and unwilling to use new innovations (McCann & Giles, 2002; Nuessel, 1982). Support for the perception is mixed. On one hand, older adults are less likely to have previous computer experience and to be computer literate (Fox, 2001; Vastag, 2001). In addition, cross-sectional data show a negative relationship between age and information use (Ende, Kazis, Ash, & Moskowitz, 1989; Fox & Rainie, 2000; Philipp et al., 1990). On the other hand, and contrary to the stereotype, older adults are willing to use new innovations: their computer usage is increasing (SeniorNet, 2001), and some computerized disease-management systems have reported success with older adults (Robitaille, 2001).
Given older adult stereotypes and the perception that older adults do not use health information technologies, one concern is that practitioners might shy away from providing older patients with health information, particularly computerized information or Internet sites. Although the stereotype that older adults are unwilling to use new innovations or technologies may be far from the truth, one concern is that the stereotype may create expectations that older adults are less responsive to health information interventions, resulting in fewer interventions designed for this group (Marwick, 1999). Yet, the negative correlations found in the cross-sectional data do not mean that interventions should focus on younger individuals, nor do they imply that a health information intervention would be more effective among younger adults than older adults.
Using data from an assessment of a communitywide informational intervention, this study has three objectives. First, in the absence of an intervention, we assess whether age is negatively associated with the use of computers, medical reference books, and telephone advice nurses for health information. Second, we evaluate whether a health information intervention had a differential effect for people of different ages. Specifically, we attempt to answer the question of whether older adults are less likely to respond to a health information intervention than younger adults. Third, we compare older adults' use of books, telephone advice nurses, and computers for health information.
|
Methods |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
The main part of the intervention involved sending every Boise household a Healthwise Handbook, which is a self-care reference guide. The guide is over 300 pages in length and is designed to complement patient–physician interaction by encouraging appropriate care. The book covers topics on wellness, prevention, symptom identification, self-care, and when to seek professional care. According to Healthwise Inc., over 132,000 Handbooks were distributed to Boise households.
Second, a toll-free telephone advice nurse was established. As of October 1998, Healthwise Inc. reported that the advice nurse received over 25,000 calls from more than 21,000 households. Third, as a way to improve access to computerized health information, a health information database was created. The computer database contained similar information to the books but provided more detail and greater searching capability. Residents could access the database through the Internet as long as they had a Boise zip code. Information kiosks, which included computers connected to information databases, were set up in public settings throughout the Boise area. Although there were a limited number of kiosks throughout the city at any given time, companies were also encouraged to put the database on their local-area networks.
Healthwise advertised the availability of these resources on billboards, on the radio, and through print advertisements. In addition, they sponsored workshops for residents and health professionals to promote self-care behaviors and to improve health-related communication skills.
The HCP improved access of self-care and health information for the entire community. Making the information easier to access led to significant increases in the use of self-care books, advice nurses, and computers for health information. The effect of the HCP on information use is of secondary interest here, and it has been reported in detail elsewhere (Wagner & Hibbard, 2001; Wagner, Hu, & Hibbard, 2001).
Data and Sample
Through an independent evaluation, a random sample of Boise households was surveyed. Eugene, Oregon, and Billings, Montana, were chosen as control cities on the basis of their geographical proximity to Boise and on similarities in metropolitan characteristics and medical systems.
Before the start of the HCP, a list of householder names for all three cities was purchased from a marketing firm. A random sample of 7,500 (2,500 per city) names was selected. Surveys were mailed in January 1996. Approximately 14% (1,048) were returned as undeliverable; one was returned with an indication that the household member was deceased. A total of 3,067 (47% of those delivered) surveys were completed.
As a follow-up survey, the 3,067 households who responded to the baseline survey were resurveyed 24 months later. Given the length of time between surveys, and concerns about attrition, a second list of householder names was purchased from the same marketing agency. At the same time as the follow-up survey, an additional 3,600 (1,200 per site) households, not previously selected, were sent the same questionnaire. Thus, 6,667 surveys were sent out in January 1998. At follow-up, 2,090 were returned as undeliverable, 12 were marked as deceased, and 2,842 (62% of those delivered) were completed. Of those who completed the baseline survey, 39% completed the follow-up survey. People with graduate school education, people with excellent health status, retirees, people living in rural areas, and older adults were significantly more likely to complete the follow-up survey. People with children in the household and those with Internet access were significantly less likely to complete the follow-up survey. Although the majority of people who completed the baseline survey did not complete the follow-up survey, attrition rates were similar for the experimental and control sites, and characteristics associated with attrition did not vary across the sites. In total, 5,909 (54% of those delivered) surveys were completed.
Variables
The survey asked respondents to indicate whether they used a medical reference book, telephone advice nurse, or a computer for health information in the past few months. People responded by stating yes or no to each of these questions. No specific reference was made to Healthwise or the HCP, making the questions appropriate for the control sites.
For independent variables, a dummy variable was included to differentiate the intervention city (Boise) from the control sites. We also added a dummy variable for survey year (0 = 1996, 1 = 1998). The interaction between site and year (Site x Year) was included to identify the intervention effect. As with any quasi-experimental design, the primary assumption is that the regression coefficient on the Site x Year interaction would be zero in the absence of any intervention effect.
Other independent variables for the household included income and a dummy variable if the residence was in a rural area. We also adjusted for the respondents' sociodemographic factors, including education, insurance status, race, gender, marital status, employment status, and whether they had any children at home. When we assessed the use of computers for health information, we also adjusted for computer ownership and access to the Internet.
It is likely that poor health and chronic illness motivate people to seek health information. Because health status is associated with age, it could confound the relationship between age and information usage. Therefore we controlled for health in two ways. First, we included a general rating of overall health: excellent, very good, good, fair, or poor. Because we had small samples, we pooled the fair and poor responses into one category. Second, people were asked to check whether a doctor had ever told them that they had high blood pressure, high cholesterol, arthritis, chronic back pain, cancer, heart disease, diabetes, depression, asthma, or chronic bronchitis. People were asked to mark all that applied. From this information, we made a dummy variable indicating any chronic illness.
The primary focus of this paper is the relationship between age and the use of health information. Age was collected as a continuous variable. In the analyses, we tried age in three different forms. First, we used chronological age as a variable. As a second option, we turned age into a set of 4 dummy variables (18–29, 30–44, 45–64, and 65+). Third, we broke age into 13 dummy variables (18–24, 25–29, 30–34,..., 80+). Age as a linear variable is more statistically efficient than age as a set of dummy variables, yet the dummy variables more easily allow for nonlinear effects. Using the 13 dummy variables provided the most flexibility, yet most age categories had small samples, making them unreliable in the statistical analysis. Moreover, the effect of age was consistent in the models with 13 dummy variables and those with 4, so we present the models with the 4.
Analysis
The first aim of this study was to assess whether there was a negative association between age and information use. This was tested with cross-tabulations between information use and age. To control for other variables, we tested this hypothesis by including age as a set of dummy variables in the regression model. The multivariate analysis was done with logistic regression. The multivariate analyses controlled for income, residence location (rural vs. urban), education, insurance status, race, gender, marital status, employment status, whether there were children in the home, and health status. For the logistic models, we ran goodness-of-fit tests, including Pregibon's (1981) link test and Hosmer and Lemeshow's (1989) goodness of fit; the hypothesis of a good fit was never rejected. To simplify the interpretation, we use the logistic regression models to compute conditional probabilities, which were then plotted. These conditional probabilities were calculated by age categories. Calculations of the conditional probabilities were based on a person who had postgraduate education, had an income >$50,000 per year, was White, had health insurance, was female, had excellent health, did not have a chronic illness, was married, had no children, and was living in an urban area. It should be noted that changes to these base variables do not affect the relative difference of using health information for the different age groups.
The second aim of this study was to test whether the health information intervention (i.e., the HCP) had a differential effect by age. This aim was tested by interacting the Site x Year interaction with the age dummy variables. This is a three-way interaction, so we also included all two-way interactions (i.e., Site x Age and Year x Age). This three-way interaction tested whether older adults were less likely than younger adults to use self-care books, telephone advice nurses, or computers for health information as a result of the HCP intervention. To facilitate interpretation, we used the logistic regression models to compute conditional probabilities. These conditional probabilities were calculated by intervention status and by age categories. When three-way interactions were significant, the conditional probability calculations also included these interaction effects. When three-way interactions were not significant, the conditional probability calculations were based on the main effects and did not include these three-way interactions.
The third aim was to compare the use of books, advice nurses, and computers by older adults. We examined the frequency with which older adults used these different sources of information in 1998, including both intervention and control sites. This aim was primarily descriptive in purpose.
Although most respondents completed most of the survey questions, there were some missing data. Missing data are problematic, as most statistical packages drop cases with missing data (i.e., listwise deletion), reducing statistical efficiency and introducing potential biases (King, Honaker, Joseph, & Scheve, 2001; Schafer, 1997). Missingness was not particular to one variable, and no one variable had more than 15% incomplete data. However, without missing data being taken into account, 2,713 (45%) cases would have been dropped. We used AMELIA (Honaker, Joseph, King, Scheve, & Singh, 1999) to address this problem. AMELIA uses EM imputation to impute the missing data. After AMELIA finished, we had five data sets wherein nonmissing values were constant across the five data sets and the missing data values varied. Bivariate and multivariate analyses were then run as usual on each of the five "complete" data sets. Then, the estimates were combined across the five data sets to yield consistent and efficient results. This technique complicates data analysis, but it is preferred to other methods for handling missing data (Schafer, 1997). For a sensitivity analysis, we ran the analyses with listwise deletion. The results were very similar, suggesting that the pattern of missingness may be random (Little, 1992).
|
Results |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
|
|
Aim 2: Are Older Adults Differentially Affected by the Informational Intervention?
This aim assessed how the different age groups responded to the HCP intervention. The intervention data show that older adults were no less likely (and sometimes more likely) to respond to the HCP intervention than the younger adults.
The HCP had a differential effect among older adults compared with younger adults on use of self-care books (see Figure 2). The three-way interaction (Age x Time x Site) was statistically significant (p <.05). Compared with persons 18–29 years of age, those over the age of 65 had a 17-percentage point greater increase in using a self-care book. It is probably more accurate to state that people 18–29 years of age were less likely than others to use a self-care book as a result of the HCP. Nevertheless, the overall conclusion remains the same: People over the age of 65 responded to the HCP in a manner consistent with how adults over the age of 30 responded.
|
|
|
|
|
Discussion |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
When we do not consider the intervention effect, we find a negative association between age and information use—older adults were less likely than other age groups to report using books, telephone advice nurses, and computers to access health information. These data confirm past research (Ende et al., 1989; Fox & Rainie, 2000; Philipp et al., 1990). Perhaps this is why there is the stereotype that older adults are less likely to use new technologies, but clearly this stereotype should not translate to the expectation that older adults are less likely to benefit from a targeted health informational intervention.
One hypothesized reason why older adults are less likely to seek additional sources of information is that they are more reliant on physicians for information. Beisecker (1988) found that although all ages desire health information, older adults put more trust in their physician's decision regarding their medical care. This research also found that older adults desired less involvement in medical decisions and had less interest in challenging the physician's decision. Similarly, Turk-Charles, Meyerowitz, and Gatz (1997) found that older adults were less likely to seek information from health professionals.
Another interesting finding from our study is that in all age groups, books were the most frequently used information source. Among older adults, the use of an advice nurse varied little, depending on whether the person had a computer or not (approximately 20%). However, computer use for health information varied tremendously by whether one owned a computer and had Internet access. For people who did not own a computer and have Internet access, computer health information use was very low (<5%). For older adults with a computer and Internet access, computer use for health information was the next most frequently used information source (see Table 2). As other research has found (Fox, 2001; Gustafson et al., 1998; Robinson, Flowers, Alperson, & Norris, 1999), access to computers remains a major barrier for older adults seeking computerized health information.
There were some limitations with this study that must be acknowledged. First, the data are from three cities in the Pacific Northwest. The results may not easily generalize to other parts of the country. In particular, almost everyone in our sample self-identified as White non-Hispanic. Communities with a greater ethnic diversity may respond differently to health information interventions. In particular, the growing use of computers and the Internet for health information highlights the "digital divide." Research has shown that low-income African Americans are less likely to have computers and Internet access at home compared with Caucasians (Brodie et al., 2000). Access to computers was important for older adults' health information use in this study. Computer access may be an even more significant barrier for older adults of different ethnic groups. Rates of using computers in this study might also be higher than would be observed in more diverse locales. Second, the study used a quasi-experimental design in which participants were not randomly assigned to receive the intervention. The benefit of this design is that it has more external validity. Information was made available to a broad community, mimicking how information is disseminated in real life. However, as with any quasi-experimental design, there are greater questions about the study's internal validity. Most importantly, an unobserved characteristic may be responsible for the effects that we attribute to age or to the intervention effect. Although we controlled for many variables in the logistic regression, this was not a true experiment with random assignment to condition.
Given the potential tremendous benefit that older adults can gain from health information and the fact that older adults do respond to interventions designed to increase their use of information, more interventions should target older adults and more research should examine the factors that impede older adults' use of information.
|
Footnotes |
|---|
1 Department of Psychology, University of San Francisco, San Francisco, CA.
2 Health Economics Resource Center, VA Palo Alto Health Care System, and Department of Health Research and Policy, Stanford University School of Medicine, Menlo Park, CA.
Lawrence G. Branch,, PhD, Decision Editor:
Received for publication May 29, 2002. Accepted for publication September 6, 2002.
|
References |
|---|
|
TOPAbstractMethodsResultsDiscussionReferences |
|---|
This article has been cited by other articles:
 |
|
 |
|
  Y. Yueh-Feng Lu and M. G. Austrom Distress Responses and Self-Care Behaviors in Dementia Family Caregivers With High and Low Depressed Mood Journal of the American Psychiatric Nurses Association, August 1, 2005; 11(4): 231 - 240. [Abstract] [PDF]
|
|
| ||||||||||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
