Expanding the Reach and Impact of
|Table of Contents|
|Executive Summary (Stand-Alone)|
|Preface: A Vision of e-Health Benefits for All|
|Chapter 1. Introduction|
|Chapter 2. Mapping Diversity to Understand Users’ Requirements for e-Health Tools|
|Chapter 3. Assessing the Evidence Base for e-Health Tools for Diverse Users|
|Chapter 4. Strategic Factors in Realizing the Potential of e-Health|
|Chapter 5. Partnerships for Meaningful Access|
|Appendix 1. Environmental Scan of 40 e-HealthTools|
|Appendix 2. Project Interviewees, Experts Consulted, and Reviewers|
|Appendix 3. Chapter 3 Literature Review Summary|
|Appendix 4. A Comparison of Internet Use and Health Status of Populations That Experience Health Disparities|
Chapter 3. Assessing the Evidence for e-Health Tools for Diverse Users (Part 2)
Usage Over Time
Studies that monitored login rates showed that logins were most frequent in the beginning of the intervention. They also found that participants used the programs less frequently and/or did not complete all modules as time passed (Clarke, Reid, Eubanks, et al., 2002, TR#7; Glasgow et al., 2003, TR#13; Irvine, Ary, Grove, et al., 2004, TR#17; McKay et al., 2001, TR#22; McKay et al., 2002, TR#21; Pinto et al., 2002, TR#27; Tate et al., 2001, TR#34; Tate et al., 2003, TR#33). Four studies found evidence of a dose-response relationship, with increased use leading to better outcomes (Celio, Winzelberg, Wilfley, et al., 2000, TR#5; Delichatsios et al., 2001, TR#9; Frenn, Malin, Bansal, et al., 2003, TR#42; McKay et al., 2001, TR#22). However, Pinto et al. did not find this effect (2002, TR#27).
Although the decline in usage may indicate some level of dissatisfaction, users in the intervention groups had higher login rates than persons in the computer-based control groups throughout the duration of the studies (McKay et al., 2001, TR#22; Tate et al., 2001, TR#34; Tate et al., 2003, TR#33). Further, the studies by Glasgow et al. (2003, TR#13) and McKay et al. (2002, TR#21) used multiple intervention groups. Similarly, they found that not only did the intervention groups use the program more than the control groups, but also the intervention groups that included a social support component had more logins than the other intervention groups.
There is almost no information on how this decrease in utilization compares to what might occur in traditional face-to-face interventions. The only exception is that McKay et al. reported that their dropout rate of 16 percent was “somewhat” higher than a similar intervention conducted in person (2002, TR#21).
Researchers identify several factors with the sites and users that might have caused attrition. Participants in a study by Napolitano et al. reported that because the Web site did not change over time, they did not need to return (2003, TR#23). Lenert and Cher reported that their site was too complex, relied too heavily on text, and required too much self-direction to locate pertinent information (1999, TR#65). They further hypothesized that people who enroll in an Internet-based program may not be as committed as those who enroll in traditional face-to-face interventions. McKay et al. thought that the Internet might be more conducive to surfing behavior and less to use of a single site (2001, TR#22). Developing Web sites that keep users coming back is a challenge (Glasgow et al., 2003, TR#13), and more research is needed to determine how to stimulate ongoing use (McKay et al., 2001, TR#22).
Other studies have identified some strategies that can be used to attract and keep users. Bowen, Ludwig, Bush, et al. found that the use of e-mail cues increased the number of women who logged in to a breast cancer information site (2003, TR#50). They found that the most common reason for nonusage was finding the time to get online. Feil et al. found no difference in attrition between groups receiving a $10 incentive and groups receiving a $20 incentive, and no difference in response to follow-up using either e-mail or regular postal service reminders (2003, TR#10). Although large numbers of people search the Internet and see many advantages to the Internet as a channel for health information, research has yet to focus on what will hold the interest of diverse sets of users and motivate them to return to a tool again and again.
Applicability is related to utility and outcomes. Because most research studies treat e-health tools as an intervention, studies typically are designed to measure the impact of the tools on a wide range of outcomes, ranging from changes in knowledge to health status. Many different types of tools were found to produce different types of positive outcomes. The findings summarized here are from studies using control group comparisons, either in randomized clinical trials or quasi-experimental designs. Only one study involved the evaluation of a commercial Web site (Womble, Wadden, McGuckin, et al., 2004, TR#38).
Knowledge and Information Needs
e-Health tools have been found to increase knowledge in a wide range of areas, including:
Gustafson et al. found that race, education level, and insurance status interacted with use of CHESS (2001, TR#15). This system helped women of color, more than Caucasian women, to overcome the perception of unmet information needs and increase their perception of participation in their own health care. Education levels and health insurance status were found to interact in the same way as race and ethnicity, with women with less education and less health insurance receiving more benefit. McTavish et al. found that women of color used a CHESS discussion group differently than white women in that the communications by women of color focused more specifically on information about breast cancer and its treatment, whereas white women were more likely to discuss daily life or offer mutual support (2003, TR#88).
Attitudes and Beliefs Theorized to Mediate Behavior Change
Positive changes in attitudes and beliefs were seen in the following areas as a result of interacting with e-health tools:
Two randomized controlled trials measured perceived social support and showed that it can be affected (Barrera et al., 2002, TR#2; Gustafson et al., 2001, TR#15). One of these studies examined a multifunctional program (CHESS), so the relative contribution of the support components cannot be determined (Gustafson et al., 2001, TR#15). Barrera et al. found that those in the support conditions (social support alone and combined social support with coach) increased their perceptions of the availability of social support as compared to the information-only control group or the group that had access to a “personal coach” (2002, TR#2).
Two studies examined decision support tools designed to be used as an adjunct to clinical care. Green et al. studied the effect of using a computer-based decision aid about breast cancer susceptibility and genetic testing (2004, TR#14). Those in the intervention group interacted with the computer and received genetic counseling; the control group received only genetic counseling. After using the computer program, women with a low risk of breast cancer were able to reduce their perceived risk of getting breast cancer and their intention to undergo genetic testing, and this perceived risk was further reduced after the genetic counseling session. At baseline, more than 80 percent of women in both groups indicated their intention to receive genetic testing; at follow-up, only 19 percent had actually undergone testing.
Chewning et al. studied the effect of a computer-based contraceptive decision aid designed to promote effective selection and contraceptive use in sexually active adolescent girls during visits to family planning clinics (1999, TR#6). The decision aid was evaluated in two clinics, one with a primarily Caucasian population (Madison, Wisconsin) and the other with a primarily African American population (Chicago, Illinois). They found that significantly more of those in the intervention group in Chicago followed through with their intention to use oral contraceptives as compared to the Chicago control group, with a similar but statistically nonsignificant trend in Madison.
Use of specific e-health tools has been shown to affect health behaviors as follows:
Two studies compared their findings to objective outcome goals. Although Baranowski et al. (2003, TR#39) and Frenn et al. (2003, TR#42) found that they were able to positively impact the dietary habits of study participants, the improvements were not enough to meet dietary guidelines.
Researchers have used a variety of e-health tools to affect health outcomes. The results, which are mixed, are summarized in the following.
Weight Loss. Two studies by Tate et al. found that an Internet-based weight-loss program led to significant weight loss in overweight adults (2001, TR#34; 2003, TR#33). Harvey-Berino et al. found no difference in weight loss between those using an online program as compared to those attending an in-person group (2002, TR#16). Womble et al. compared weight loss in overweight women who were randomly assigned to use a commercial dieting site (eDiets.com) or a weight-loss manual (2004, TR#38). In the strictest analysis of data, they found that the group using the manual lost significantly more weight than the group using eDiets.com.
Pregnancy. In a study of contraceptive use, there were no differences between control and intervention groups in the discontinuation of oral contraceptives. There was a statistically nonsignificant trend toward decreased pregnancy in Madison for those who used the computer-based decision aid, but no difference between groups in the Chicago sample (Chewning et al., 1999, TR#6).
Mental Health and Quality-of-Life Outcomes. Proudfoot et al. found decreased levels of depression and anxiety in people with those conditions (2003, TR#28). Clarke et al. found no effect of their Internet program on depression; however, process evaluation showed low usage of the program overall (2002, TR#7). Winzelberg et al. found significant changes in measures of depression, stress, and cancer-related trauma in women with breast cancer, but no difference in anxiety or coping for women (2003, TR#37). A possible explanation is that the intervention was not designed to affect these measures directly. Smith and Weinert found no differences between study groups on psychosocial and quality-of-life measures in women with diabetes, although this may be due to a small sample size (2000, TR#32). The participants did report that the project provided a great deal of support and feelings of connectedness. No changes in quality-of-life measures were found in adults with type 2 diabetes (Glasgow and Toobert, 2000, TR#12). Both groups (eDiets.com and manual) in the study by Womble et al. showed improvements in quality-of-life measures and less depression during the course of the study, but there were not significant differences between the groups (2004, TR#38).
Physiological Measures. Modest changes were found in cholesterol and lipid ratios along with small reductions in glycosylated hemoglobin (HbA1c) levels in adults with type 2 diabetes (Glasgow et al., 2003, TR#13; Glasgow and Toobert, 2000, TR#12), but no change was found in these measures in a study by McKay et al. (2002, TR#21). No difference was found in blood pressure, glucose, lipids, or lipoproteins between groups in the Womble et al. (2004, TR#38) study.
Possible Negative Outcomes
Some researchers have posited possible negative effects, such as increased depression or social withdrawal, from Internet use. Several studies show that those who seek help in online communities may have more serious conditions than those who do not (Beebe, Asche, Harrison, et al., 2004, TR#47; Epstein, Rosenberg, Grant, et al., 2002, TR#55; Erwin, Turk, Heimberg, et al., 2004, TR#56; Houston, Cooper, and Ford, 2002, TR#44). However, these studies were not randomized controlled trials. It is not clear that Internet use is the cause of this greater impairment. It is equally possible that those who need support and lack it in their face-to-face relationships are trying to attain support via the Internet (Beebe et al., 2004, TR#47).
Another area of concern relates to the possibility that patients could become distressed or anxious by something they read as a result of having electronic access to their medical records (Tang et al., 2003, TR#76; Masys et al., 2002, TR#68). Tang et al. used hyperlinking to link medical terms to a dictionary to improve patient understanding, but they did not evaluate the impact of this feature (2003, TR#76). Masys et al. set up safeguards, including a toll-free hotline number, to protect patients; however, they found that this concern was unfounded for this group of participants (2002, TR#68). Participants using SPPARO (System Providing Patient Access to Records Online), a Web-based online medical record, did not report any negative effects (Ross et al., 2004, TR#30).
Cost Savings and Return on Investment
Although not part of the “Five A’s” framework, described at the beginning of this chapter, the effect of e-health tools on costs and return on investment for healthcare organizations, insurers, employers, and the Government is of strong interest in the policy and healthcare communities.
Researchers are beginning to calculate the financial impacts of the use of e-health tools. Krishna et al. provided evidence that using an e-health tool for asthma self-management education is cost-effective (2003, TR#18). This study showed reductions in emergency department visits in the intervention group that translated into a savings of approximately $907.10 per child as compared with a savings of only $291.40 per child for the control group. Other indirect savings were discussed but not calculated. For example, the children in the intervention group used a significantly lower average dose of inhaled corticosteroids by their third clinic visit, thus leading to a reduction in medication expenditures. In addition, they reduced school absences during the study period by an average of 5.4 days per child per school year as compared with 1.6 days for children with asthma in the control group. These indirect savings would be realized by working parents and their employers.
In a randomized clinical trial, 59 children and adolescents, age 8 to 16, improved their self-care and reduced their emergency clinical utilization after playing Packy & Marlon, a health education and disease management video game (Lieberman, 2001, TR#20). They reduced diabetes-related urgent and emergency visits by 77 percent after 6 months of access, compared to no reduction in clinical utilization in a control group of youngsters with diabetes who used an entertainment video game with no health content.
Ross et al. found no difference in hospitalizations or mortality between patients who used SPPARO and those who did not have access (2004, TR#30). Those who used SPPARO did have more emergency department visits; however, these did not temporally relate to use of SPARRO.
e-Health tools can also result in savings by enabling patients to perform monitoring tasks that professionals would do. For example, Finkelstein et al. demonstrated that lung function test results collected during home asthma telemonitoring were comparable to those collected under the supervision of trained professionals (2001, TR#11).
Summary and Discussion
This chapter provides a review of recent research pertaining to e-health tools and factors affecting their use by diverse population segments. Overall, the research continues to inspire a sense of promise for these tools as many positive findings have been reported across different categories of tools with a wide variety of components. The lack of diversity in the samples used in these studies, however, makes very clear one of the key messages of this report. The body of knowledge about which groups will engage with and benefit from e-health implementation is thin and must be developed using a model of diversity if the tools are to achieve their potential as public health interventions. This section summarizes the research reviewed in this chapter and examines the limitations and challenges of current research.
The Body of Research
Existing research on e-health tools clusters around two broad areas: (1) evaluation of public domain e-health tools and Internet use, and (2) development and evaluation of specific tools developed and tested in research settings. Research on tools in the public domain includes quality assessments and readability analyses of online content, content analyses of online communities, and surveys and observations about how people use the Internet.
The general public appears satisfied with the information and support online; however, content analyses find that the quality of the information is less than optimal. Furthermore, readability and other access issues may make online use difficult for members of diverse populations. Evaluation of e-health tools can benefit users by improving the quality and effectiveness of the tool, minimizing the chance of harm, promoting innovation in the tools, conserving resources, and allowing users to make informed choices about tools (Eng, Maxfield, Patrick, et al., 1998). Only one study evaluated a widely available commercial e-health tool (eDiets.com) in a randomized controlled trial, the results of which were not favorable.
The second broad area of research focuses on the development and evaluation of specific e-health tools. These studies provide information about the usability, efficacy, and effectiveness of the tools. The quantity and quality of the research is uneven across topics and tools. Some areas, such as tools for behavior change, are theory-based and have generated sound research and evaluation to support their use. Many multiple randomized controlled studies across several health topics have found positive outcomes. Other tools, such as healthcare tools, that are emerging in response to market and policy demands do not yet have much of a scientific basis to suggest that they will have their intended effect. Most of the research on these tools is focused on satisfaction and usability.
Unfortunately, many research-based tools are not widely distributed or easily accessed by the general public. It is important to bring evidence-based e-health tools to those who can benefit from them. The reverse is also true. It is just as important to use the findings about what people actually need, desire, and do while online to guide the development of research-based e-health tools. Much work remains to be done to bridge the gaps between these areas. Chapter 4 discusses this topic in greater detail.
Although the literature review and the scan of tools in the field identified a large number of tools, there are no standard, accepted definitions for purposes or components of tools for consumers. In general, the tools tend to be multicomponent programs that have been designed for many purposes: to inform, provide support, aid behavior change, assist decisionmaking, help manage disease, and facilitate interaction with the healthcare system. Some research studies clearly describe the tool being studied; others provide only vague descriptions. Some tools with similar stated purposes have notably different components. The wide range of tools reflects the array of burgeoning and exciting possibilities that can be offered through electronic media, but it also makes the comparison of different studies and future replications difficult.
More needs to be known about e-health tools, including the identification of critical components and combinations of components as well as the optimal conditions for use of these tools. Individual studies may answer one or two questions about use, but there is not yet a body of research that indicates who should use these tools, when, where, how frequently, and how intensively. Factors that lead to user adoption and ongoing use as well as factors that lead to attrition also need to be identified.
It is encouraging that many studies have found positive changes in knowledge and intention after just one interaction. Findings on actual behavior change and health outcomes have been less clear. However, many of these studies may not have provided interventions with enough frequency or intensity to bring about desired changes in these areas.
Key Findings of the Review by Access, Availability, Appropriateness, Acceptability, and Applicability
Millions of people are using the Internet for health-related purposes, and estimates can be made about the deployment of e-health tools in large, closed systems, such as the VA’s My HealtheVet. Beyond this, little is known about actual uptake and use of e-health tools. Few if any data exist on the distribution of e-health tools across the population or within subgroups. Population and subgroup data on level of interest in and attention to these tools also are not available. Large numbers of e-health tools have been developed, but it is not known how many people know about these tools, how many are using these tools, and how many could be influenced to try them. The ability of interested users to locate and access these tools, particularly those with a credible research basis, is also unknown.
A major issue that emerges from this review is the limited external validity of much of the research, as so many of the studies utilized convenience samples or required computer ownership. This approach has led to a disproportionate amount of information on Caucasian women with higher education. Even when studies reported the demographics of their samples, most did not analyze their findings according to these variables. A few exceptions exist, such as the findings from CHESS, in which women of color, women who were less educated, and women with less health insurance appeared to derive greater benefits from interacting with CHESS (Gustafson et al., 2001, TR#15). Similarly, Oenema and Brug found that respondents with less education seemed to have benefited more from the tailored nutrition feedback than did those with higher education (2003, TR#25). Frenn et al. also found evidence that their intervention had a differential effect based on race and gender of users (2003, TR#42). The lack of diversity in the research samples and evidence of differential effects based on demographics suggest major gaps in our knowledge about how to address issues of access as well as the acceptability and appropriateness of personal e-health tools for diverse segments of the population.
Some tools have been recently developed that target special populations, and some of these were developed with input from the target audience. These studies show that with careful attention to cultural, literacy, and technological needs, successful tools can be developed for and used within these subpopulations (Campbell et al., 1999, TR#4; Jantz et al., 2002, TR#45). User-centered design and usability research, along with participatory research methods, can be used to bridge the gap between what designers and researchers envision and what the ultimate end users actually find engaging and helpful. It is critical to seek input about the diverse needs of all potential users during tool development and ensure that they are represented in the evaluation studies.
Any review in this area should consider how technology is used in the research projects. The studies that required participants to use their own computers found that the capabilities of users’ technology can vary tremendously. At times, researchers have found that participants were not always able to access all parts of the programs being tested. These kinds of studies are important because they help determine the feasibility of delivering e-health tools over the Internet. Other studies had participants interact with an e-health tool in a lab or clinical setting. This allows for potentially greater representation in the study sample, helps minimize potential technical problems, and gives an idea of the efficacy of a tool, that is, its success under very controlled conditions. Information from both of these kinds of studies is important for building the knowledge base for e-health tools.
Findings from the studies in the Acceptability section reveal that people like e-health tools and generally find them easy to use. There does seem to be a decline in usage over time, but the declines were not as steep as those found in the control conditions. It is not known how this decline compares to other intervention formats, such as in-person educational or therapeutic programs. Several researchers have ideas about why dropoffs occur; they posit that sites are too complex or not dynamic enough. Research will need to continue to investigate these factors. A research path would be to examine what personal qualities lead to preferences for online interventions or whether differences exist between those who seek help online and those who seek face-to-face interventions.
The studies in this section found many positive findings, but some design issues deserve further mention.
Measures. These studies showed a strong reliance on self-reported data to document change. Typically, self-reported data are considered weaker than other types of objectively collected data and subject to bias. Because participants tend to make their responses more socially desirable, the effects may be overstated. Also, many of the studies use questionnaires or adapt existing questionnaires without reporting reliability or validity. This could affect findings in unknown ways. To establish firmly the effectiveness of these tools, researchers must continue to develop and utilize objective, reliable, and valid measures.
From a health literacy perspective, an equally important issue may be the mismatch in understanding between researchers and study participants about what is being measured. The health literacy construct highlights the frequent gap in understanding between health professionals and nonprofessionals. Particularly when the use of technology is involved, attitudes, beliefs, and expectations may play an important role in shaping how users interact with the systems and report data.
Frequency, Duration, and Intensity. The studies examined a variety of tools under a variety of conditions. Some studies exposed participants to the intervention for only one short session; others made a Web site available to users over a specified period of time. Because of the differences in the tools, it is difficult to compare the effects of frequency, duration, and intensity across studies. There does appear to be a dose-response relationship in which those participants who showed the greatest use of a tool also showed the greatest benefit. No studies formally manipulated the frequency, duration, or intensity of use.
Types of Control Groups. The types of control groups used in these studies varied. Some control groups received no intervention. Others received treatment as usual, which might include in-person contact or informational brochures. It is possible that the positive effects of such comparisons in these studies are due to the use of the computer itself rather than the specific intervention.
Studies are beginning to appear that have control groups using alternative computer-based activities. For example, while the intervention group in the study by Jantz et al. used a program about nutrition, the control group interacted with a program on household budgeting (2002, TR#45). This type of comparison allows researchers to make a stronger case for attributing findings to the computer-based intervention itself rather than the novelty of the channel. Gustafson et al. points out that some of the benefits seen in their study may be due to loaning participants a computer, although they dispute this because their data showed significant actual use of the CHESS program (2001, TR#15). Further evidence is seen in the study by Barrera et al. in which the control group had computer access, but did not show the same benefits as the intervention groups (2002, TR#2).
Capitalizing on Digital Technology for Research. Although evaluation of e-health tools shares many similarities with evaluation of other health-related media, some unique opportunities are specific to the use of digital technology. Research is beginning to capitalize on these attributes. For example, several studies used computer-based assessments that can streamline the data collection and entry process. Anecdotal evidence suggests that this approach can be a less threatening way of collecting data from populations with low literacy. Other studies have used online tracking systems that can help determine if participants actually used the programs and in which areas they spent their time. This type of process information can be very important in helping to determine what users find attractive and which program components are effective.
The research enterprise will need to be harnessed in a more coordinated and focused manner to ensure access and the availability of appropriate tools for people who want and need them. As noted in Chapter 1, “doing better” in the application of e-health tools to population health improvement means finding the best approaches to create tools that are “participatory, deeply meaningful, empathetic, empowering, interactive, personally relevant, contextually situated, credible, and convenient” (Neuhauser and Kreps, 2003). Meeting these requirements will entail much greater attention to the use of participatory research methods and samples that reflect population diversity than demonstrated in the current body of research.
Endnote: Search Terms
The following search terms were used in the search strategy for Chapter 3:
Health Information: A preprogrammed PubMed search was conducted under Healthy People 2010 Objective 11-4—Increase the proportion of health-related World Wide Web sites that disclose information that can be used to assess the quality of the site—using the following search terms: (internet/standards[majr] AND (web OR website OR websites) AND (quality assurance OR quality control[mesh] OR confidentiality[mesh] OR privacy[mesh] OR ethics[mesh] OR health education/standards[mesh] NOT letter[pt] AND English[1a].
Behavior Change/Prevention: (Internet OR computer OR CD-ROM OR interactive multimedia) AND (behavior change OR health promotion OR prevention)
Online Communities: (Online OR Internet OR computer-mediated) AND (communities OR chat groups OR chat rooms OR listservs OR discussion groups OR support groups) AND health
Healthcare Tools: Personal electronic health record, personal electronic medical record, electronic messaging. Searches also were conducted for research related to specific healthcare tools as identified in the expert interviews.
Decision Support: Decision support, decision support tools, decision support AND online, decision aid
Disease Management: Disease management, disease management health tools, self-care tools, consumer health management tools
|Back to Top||Chapter 4 >|