Results of a factorial survey investigating the health information seeking behaviors of older adults

Authors


Abstract

This paper presents the results of research of an experimental research method, the factorial survey, investigating the health information seeking preferences of older adults (60 years and older). Volunteer subjects from rural, urban, and suburban areas in Ohio participated in this study; 450 subjects responded to health information seeking vignettes (the key feature of factorial surveys) and provided demographic data. Using regression analysis and ANOVA, findings from this study address preferences for the format of materials, information on particular health topics, and, preferences for receiving assistance from information providers such as librarians, nurses, pharmacists, and, doctors.

Introduction

Health information seeking behaviors, including social, economic, educational, and age-related aspects of those seeking health information, can be studied at many levels. These behaviors can be studied at the level of the individual, the information provider, and the information provider network of relationships, services, and resources. We conducted a 2-part research project to address the health information seeking behaviors of older adults. We examined this issue experimentally with a factorial survey of older adults and with focus groups of library and community health care professionals, and, health literacy specialists. Through use of a factorial survey, we came to a better understanding of individual health information seeking preferences for format, provider, assistance, and content as well as the potential relationships between those preferences and knowledge, skills, and access to health information. The results of the factorial survey are reported in this paper. While this project uses an exemplar of older adults, the research method has the potential to be used to study other groups in order to better understand preferences affected by knowledge, access and skills related to accessing and using health information. The hoped-for result is that health information programs and services, informed by these research findings, will be more effective, more efficient, more outcomes-oriented, and more responsive.

There are many efforts across the U.S.A. focusing on eliminating health care disparities. For example, one of the goals of a broadly embraced framework, Healthy People 2010 (U.S. Department of Health and Human Services, 2000), is to identify and address health disparities in segments of the American population. Contributing factors to disparities include access to and use of health information. The research presented in this paper, which addresses the health information seeking decisions of older adults, identifies essential health information factors that may contribute to disparate health care as expressed by the preferences of subjects. These factors include literacy skills, health challenges, preferences for information format and provider.

Approach to the Research

Within the next twenty years, the United States will be faced with increased pressure on its health care delivery mechanisms as the dramatic aging of our population becomes more evident and as current economic stresses have an effect on health care choices. Presently, two paradigms underlie efforts to maintain the health of older adults. The first, a medical paradigm, is based on “crisis management,” when an individual's health is critically impaired. The second, a wellness paradigm, is based on prevention and empowerment of the individual (Moore and Raybourne 2001). The research presented in this paper uses the wellness paradigm by addressing and identifying those factors that are associated with the health literacy and information-seeking needs and behaviors of older adults.

In the U.S., there are millions of people who simply do not seek or use health information services. The poor, the educationally disadvantaged and older adults are among the people who often are not health literate and are not very likely to seek or appropriately use health information (Institute for Healthcare Advancement, 2004). Health literacy is generally referred to as the ability to access, understand, and use health related information to make decisions (Medical Library Task Force, 2004). Health literacy goes beyond the ability to decode health-related text and includes all of the social and cultural practices that are associated with any one group of people. It involves the process of understanding, accessing and using health-related tools and services by individuals so that they can make informed choices, influence events and exert greater control over their own lives (Seldon, et al., 2000; Shohet, 2002).

Linking adult literacy skills with an understanding of health and wellness is of growing interest as a component of managing personal health (Schloman, 2004). Findings show that older adults and especially those with chronic physical or mental health conditions scored at the lowest two levels of a literacy scale defined by the National Assessment of Adult Literacy (U.S. Department of Education, 2004). Significantly, individuals with low health literacy have higher medical costs due to a variety of factors, including but not limited to, more medication and treatment errors, more hospitalizations, longer hospital stays, more doctor visits, and lack of necessary skills to obtain needed services (Gazmarian, et al., 1999).

Digital resources for health information needs are widely available, if one has access to the Internet. However, a research provides evidence of a “substantial digital divide among seniors based on income, education, age, and gender” (Kaiser 2005). Strikingly, 64% of seniors on Medicare have an annual income under $20,000 and it is this income group that is only 15% likely to have gone online for any reason, compared to a 65% likelihood that those with incomes of $50,000 or more have gone online. There is an equally striking divide for use of the Internet for health information; only 6% of seniors with incomes under $20,000 have gone online for health information, compared to 43% of those with incomes of $50,000 or more.

Many poor, educationally disadvantaged, and older adults may not seek or use health information from the more traditional health settings or the Internet (Kaiser 2005; Huttlinger, et al. 2003; Parker 2000) but they may utilize the services of other organizations in the community, such as churches, libraries, community centers, and educational programs where there are opportunities for participating in “health information events.” Fisher, Durrance and Hinton (2004) call these places “information grounds” and they are places where people can come together to perform a given task, but from which emerges a social atmosphere that fosters the spontaneous and serendipitous sharing of information. We use the factorial method to determine whether gathering places such as a church, clinic, or library are acceptable locations for older adults to seek health information and guidance on how to use this information.

The literature explores health information seeking by examining objects, behaviors, and, systems. In terms of behaviors, studies of the health information seeking behaviors of various populations include Pena-Purcell, 2008; Yoo, et al., 2008; Benotsch, Kalichman, & Weinhardt 2004; Ellis-Danquah, 2004; Warner & Procaccino 2004; Huber, Hughes, & Peek 2003. There is also considerable interest in studying interactions with digital health information (Detlefsen, 2004; Fulda & Kwasik 2004; Huntington, Homewood, & Nicholas 2004). Because of the complexity of the health information seeking process and the systems which frame the process, several researchers and national organizations have outlined the need for a thorough and rigorous approach to research of health information seeking behaviors and needs (Lee, et al. 2004; Nutbeam 2000).

Questions, Design and Method

Our project team comes from several disciplines and practice areas, including library and information science, nursing, adult literacy, and, gerontology. This interdisciplinary approach resulted in a research project that addresses the following key concepts: literacy, information resources, information providers, and, health topics. These key concepts were then addressed in the research design and research questions.

The primary research question is: What are the variables that affect health information seeking behaviors of older adults? The specific research questions addressed by the factorial survey are: 1) How do knowledge, access and skills affect the older adults' decision-making about selecting and reading health information and seeking additional health information?; and,2) How do the demographic characteristics of the subject relate to the subjects' decision-making about selecting and reading health information and seeking assistance with health information?

We use the factorial survey method to examine the contextual factors that explain the health information seeking preferences of older adults in urban, suburban, and rural areas of the northeastern quadrant of Ohio.

Subjects, Data Collection and Sampling

The 450 participants in this study were drawn from a population of older adults in an area of the state where the population is economically vulnerable and less educated than the rest of the state and the United States and is at risk for overall poor health. The participants were drawn from eight (8) counties that have a slightly higher average of adults age 65 or older (14.2%), compared to 12.4% for the overall U.S average (U.S. Census Bureau, 2002). Approximately 16.9% of adults over the age of 25 in the counties targeted in this study have a Bachelor's degree compared to 24.4% for the United States (U.S. Census Bureau, 2002). The study area consists of a population with 17.4% of inhabitants over the age of 25 having less than a high school education, compared to 15.4% for the United States (U.S. Census Bureau, 2002). We chose these counties because many individuals in these areas have substantial risk and practical needs for help related to health, finances and education.

All subjects participating in this study were 60 years of age, agreed to participate, and lived in the county in which they were solicited. If we found that a subject could not read or see well enough to participate, the survey was read to him or her. The research team administered the instruments at senior centers and a public library in each of the 8 selected counties and collected four-hundred-fifty (450) usable surveys for a 94% response rate. As an encouragement to participate, all subjects who completed a questionnaire were given a gift certificate of $10.00 for CVS (drugstore chain). Data were collected over a 6 month period during 2007 and 2008. Half of the subjects were drawn from 1 public library randomly selected from each of the 8 counties and half were drawn from 1 senior nutrition site selected from each county. Senior nutrition sites are sponsored programs offered by the respective Area Agency on Aging across communities where congregate meals and socialization are offered to seniors over the age 60. These services were enacted with the 1972 Older Americans' Act.

Factorial Survey–Background

The factorial survey is a method not previously used in library and information science. The factorial survey design combines the use of vignettes with sample survey procedures. It allows for a large number of realistic vignettes, with multiple independent variables that have multiple levels and allows for up to 3 dependent variables that can be included in a single design (Ludwick & Zeller, 2001; Ludwick et al 2004). The factorial survey vignettes are formulated mathematically, based on a random table of numbers, (Ludwick, et al, 2004), so that each respondent receives a set of unique vignettes to judge in order to causally test the factors that significantly impact the respondents' decisions. The factorial survey combines the benefits of a factorial experimental research design with the benefits of a sample survey. The design was established in sociology by Rossi & Nock (1982) and has been used in the social and health sciences to study a variety of professional and lay judgments.

Variables and Instrument

The study questionnaire consists of two parts: two randomly generated vignettes; and, background questions describing the respondents' demographic characteristics. The vignette design allowed us to causally examine factors that are closely interrelated in real life (mulitocollinear) and, thus, often are only examined descriptively. This complex factorial design is possible because the levels for each independent variable are randomly generated. This means that the correlations of the independent variables will approach zero, thus the multiple regression slopes will not suffer from the interpretational problems encountered when the independent variables are collinear.

The 10 independent variables (IV), the levels for operationalization, the values of the IVs, and the rationale for variable inclusion are listed in Table 1. These variables are conceptualized as relating to access (e. g. location and distance) and knowledge, (e.g. health problems and vocabulary). The 3 dependent variables are attitudinal and using a rating scale (0 to 5, where 0 = definitely no and 5= definitely yes) asked the subjects their judgment of the likelihood to take action based on the information provided in the vignettes:

  • A)What is your likelihood of using the information (i.e. reading material, watching video or using the computer)
  • B)sWhat is your likelihood of participating in the meeting?
  • C)What is your likelihood of talking to your family or friends before going to the meeting?

Following is an example of one of 1200 randomly generated vignettes – ALL CAPS indicates IV values placement. Literacy level was determined by vocabulary and sentence structure.

Example Vignette: A GROCERY STORE which is in WALKING DISTANCE from your house is offering information on how to improve your health like healthy eating. The free information is offered in a PAMPHLET TO TAKE HOME. In addition there are free scheduled GROUP MEETINGS given that last 15 MINUTES. They are led by a REGISTERED NURSE. These are scheduled during the DAY during the WEEK.

Table 1. Independent Variables.
original image

The second part of the survey included demographic characteristics of the respondent, such as the respondent's age, health problems, city or township of residence, marital status, education, primary care source, health insurance coverage, transportation, frequency of library use, Internet access, and computer and Internet skill level. All of these demographic questions have been asked in the previous work of team members, or, are commonly found in similar research and the reading level for all of these questions has been determined to be third grade level.

Study Analyses and Results

The central analysis for model building is use of the vignette variables as predictors on the judgments of the participant's likelihood to take action as previously noted. Additionally, subject characteristics were used to predict the outcomes. The unit of analysis for the regression is the vignette and the number of vignettes for analysis in this study were 900 (450 subjects X 2 vignettes each). Regression analysis was used to provide the effect sizes, statistical significances, and variances explained for each of the 3 models that explored the effect of the independent variables on the 3dependent variables. Regression analysis is the technique recommended by Rossi and Nock (1982) and is the most common form of analysis for this method (Wallander 2009).

Dummy coding was used as the independent variables were noncontinuous. As an example, let us examine the variable of type of health problem, which was one of the independent vignette variables This variable had 6 categories (healthy eating, exercise, stopping smoking, diabetes, high blood pressure, cancer and stroke). When the 5 dummies are used as predictors of the outcome variables, the regression equation emerged as follows: Y = a + b1×1 + b2×2 + b3×3 + b4×4 + b5×5 (where a is the Y intercept, bi are the slopes, and Xi are who offered the program).

Results – Descriptive Data

The average age of the subjects was 72 (SD 8); the oldest participant was 93. In terms of demographic data, our subject pool can be characterized as follows:

  • 68% Women44% Married

  • 11.5% <High School 51.2% High School 27.1% Some College or Degree

  • 45% - Don't use a computer on a regular basis or at all

  • 50% - Don't use Internet on a regular basis or at all

  • 25% - Don't use library on a regular basis or at all

In terms of health-related characteristics, only 15% reported no health related problems. Most reported carrying out some activity to keep healthy with 74% reporting eating healthy and 62% reporting exercising regularly. Regarding specific health problems, it was found that

  • 11% smoke 18% have diabetes

  • 7% have had a stroke 55% have high blood pressure 13% have or had cancer

When compared to the state population data, this subject pool was actually better educated than expected. We did expect that the library sites would lend themselves to subjects with higher educational attainment than the general population. However, even the subjects at the senior centers were relatively well educated.

In terms of computer and Internet use, our subject pool is below the national average as reported by the Pew Internet and American Life project (Pew Internet & American Life). Coupled with the higher than expected educational attainment which is typically a predictor of computer use, we can only surmise at this point that this particular region of the country may have a higher percentage of non-computer and non-Internet users in this age group.

Results – Regression Analyses

The central analysis in the study focused on the vignette variables. The regression analysis of each of the dependent variables provided information on the amount of variance explained, effect sizes, and significance of variables predicting each of the three dependent variables.

Model 1 examined the impact of the independent variables on the dependent variable, “What is your likelihood of using the information (i.e. reading material, watching video or using the computer)”. With all the vignette variables from were entered into the analysis the resulting regression model was significant (R=.357; R2=.127; R2adj=.102; F (25, 849) =4.956, p<0.000). Specifically, the independent vignette variables had a.357 correlation with likelihood of judging that the information presented by print, video or computer would be used, and accounted for 10% of the variance.

Secondly, we found that health issue, source and location of the information available were significant predictors in the regression model that examined likelihood of using the information. Specifically, all of the health issues (healthy eating, exercise, diabetes, high blood pressure, cancer and stroke) were significant in predicting use when compared to the base category of smoking. Both pamphlets and video for home use were significant predictors of likelihood of use when compared to the base category of computer use at the location. Finally, church was significantly less likely to impact the use of information when compared to the base category of a school. (See Table 2 where the significant unstandardized partial regression slopes (B), standard errors (SE), and significance levels are presented.)

Model 2 examined the impact of the independent variables on the dependent variable, “What is your likelihood of attending a meeting” to get information. With all the vignette variables from were entered into the analysis the resulting regression model was significant (R=.297; R2=.088; R2adj=.062; F (25, 850) =3.298, p<0.000). Specifically, the independent vignette variables had a.297 correlation with likelihood of judging that the subjects would attend a meeting and accounted for 6% of the explained variance.

In a like manner to Model 1 we found that health issue again was a significant predictor in this regression model. All of the health issues (healthy eating, exercise, diabetes, high blood pressure, cancer and stroke) were significant in predicting the likelihood to attend a meeting when compared to the base category of smoking. Further, a meeting that was 30 minutes in length was significantly more likely to result in judgment of likelihood to attend a meeting than a meeting that lasted 60 minutes. A meeting led by a registered nurse was significantly less likely to be judged likely to attend as compared to the base category of doctor. ((See Table 2 where the significant unstandardized partial regression slopes (B), standard errors (SE), and significance levels are presented.)

Model 3 examined the impact of the independent variables on the dependent variable, “What is your likelihood of talking to your family or friends before going to the meeting?” With all the vignette variables entered into the analysis the resulting regression model was significant (R=.221; R2=.049; R2adj=.021; F (25, 849) =1.738, p<0.014). Specifically, the independent vignette variables had a.221 correlation with likelihood of judging that the subjects would talk to family or friends before attending a meeting and accounted for 2% of the explained variance.

In a like manner to the previous models we found that health issue again was a significant predictor in this regression model, but there was one difference. All of the health issues (healthy eating, exercise, high blood pressure, cancer and stroke) except diabetes were significant in predicting the likelihood to attend a meeting when compared to the base category of smoking. A meeting held at a school was significantly more likely to impact whether the subjects would talk to family or friends before attending a meeting and watching a video was significantly less likely to be discussed with a family or friend. (See Table 2 where the significant unstandardized partial regression slopes (B), standard errors (SE), and significance levels are presented.)

Table 2. Regression Analyses – Likelihood to Acquire Information, Participate in a Meeting, Talk to Friends/Family
original image

ANOVA of Significant Variables

Based on the regressions in the above models it can be seen that health issue was an important significant predictor variable across all 3 dependent variables. A one-way ANOVA was run in order to examine the mean scores for each level of this independent variable. LSD post hoc comparison was used to determine which pairs of means differed significantly. The one-way ANOVA results indicated that health issues differed significantly on all 3 of the dependent variables. The consistent finding was that on post hoc analysis all of the health issues were significantly different from smoking, but no other consistent patterns were found. As can be seen in the Table 3 all subjects were less likely to acquire information, meet about or talk to family or friends about smoking.

Table 3. ANOVA.
original image

Interaction Effects

We also have found several interaction effects. The first interaction effect was found between format of information (e.g. pamphlet) and health issue (e.g. high blood pressure). The second interaction effect was the location of the survey implementation (library or senior center) and the likelihood of the subject attending a meeting. Our analysis indicates that subjects from the library sites are less likely to participate in a group class than subjects from the senior centers. Perhaps those older adults who frequent senior centers are more inclined to participate in group activities because they prefer socializing. Older adults who frequent libraries may think of the library as a destination and not a source of social interaction

Additional Finding Related to Age

Last we looked at the differences for age and its effect on the 3 dependent variables and we also examined age for its effect on computer use. It was found that age was negatively correlated with all 3 dependent variables, that is older subjects were significantly less likely to acquire information, meet about health topics and talk to family or friends. As age increased the subjects reported significantly less library, computer and internet use.

Implications and Next Steps

Based upon this research and its current analyses, the implications for providers of health information, including libraries, health care agencies, and senior services, include the following:

  • 1.Printed materials in concise form are preferred by older adults.
  • 2.Particular health topics may be more preferred by certain segments of the older adult population.
  • 3.Assistance in using health information resources through group meetings or classes may not be accepted by those who go to libraries but may be acceptable to those who use senior centers.
  • 4.The difference that family and friends might make in a health information seeking setting is not clear.

Our next steps include analysis of focus group interviews with information providers, and, intersections between the two phases.

Future Research Considerations

Some settings are associated with significantly more likelihood of acquiring information, e.g., school, and others are not, which implies that the information grounds concept needs further study. The lack of significance of reading level and community types indicate that these two concepts need additional study within this context.

Acknowledgements

This research was funded by the Institute of Museum & Library Services.

Ancillary