This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
Physical activity (PA) for older adults has well-documented physical and cognitive benefits, but most seniors do not meet recommended guidelines for PA, and interventions are lacking.
This study evaluated the efficacy of a 12-week Internet intervention to help sedentary older adults over 55 years of age adopt and maintain an exercise regimen.
A total of 368 sedentary men and women (M=60.3; SD 4.9) were recruited, screened, and assessed online. They were randomized into treatment and control groups and assessed at pretest, at 12 weeks, and at 6 months. After treatment group participants rated their fitness level, activity goals, and barriers to exercise, the Internet intervention program helped them select exercise activities in the areas of endurance, flexibility, strengthening, and balance enhancement. They returned to the program weekly for automated video and text support and education, with the option to change or increase their exercise plan. The program also included ongoing problem solving to overcome user-identified barriers to exercise.
The multivariate model indicated significant treatment effects at posttest (
These results suggest that an online PA program has the potential to positively impact the physical activity of sedentary older adult participants. More research is needed to replicate the study results, which were based on self-report measures. Research is also needed on intervention effects with older populations.
Physical activity (PA) for older adults increases active life expectancy while limiting the development and progression of disabling conditions and chronic disease [
Although the most effective PA intervention mediators are yet to be determined, theory-based behavioral interventions promoting adoption of exercise as a lifetime habit are recommended [
Computer-technology with multimedia interfaces has the potential to provide cost-effective personalized home PA interventions [
Shaping a PA intervention to an older adult audience requires a thoughtful approach because seniors may have decade-old habits and attitudes to change, and they may have functional limitations due to age or medical conditions. Further, based on rates of Internet adoption, seniors may be less prone than younger age groups to adopt new technological approaches, but the tide is changing as baby boomers age and use of the Internet becomes more commonplace. While seniors have been the slowest age group to use the Internet, they have been the fastest growing population segment in recent years, and as of April 2012, 53% of American adults aged 65 and older use the Internet or email, and 70% of them use the Internet daily [
The intervention in this research was based on the theory of planned behavior [
Consequently, the program was designed to provide information and support that would enhance knowledge, attitudes, self-efficacy, and behavioral intentions to participate in exercise activities on a regular basis. Using the criteria recommended for theoretically driven web exercise programs by Doshi and colleagues [
Stage of change theory [
In this study, we developed and tested a stand-alone 12-week Internet intervention designed to improve self-reported PA of sedentary older adults. The randomized design (Clinicaltrials.gov NCT01579240) evaluated self-reported changes in exercise across four domains: endurance, stretching, strengthening, and balance. We hypothesized that the intervention would be linked to improvement in the above PA domains and to theoretically relevant mediators of behavior change (eg, attitudes, self efficacy, behavioral intentions) and that user acceptance would be positive. This was a “real-world” effectiveness trial [
The intervention, entitled
The interactive framework was developed in consultation with professionals experienced in the design and implementation of research-based exercise programs for older adults. Care was taken to include only exercises that the participants could do safely on their own with minimal equipment.
At the initial 1-hour start-up session,
During the first visit, participants were asked to identify their personal goals and their perceived benefits of increased exercise. Users provided information on their recent exercise history and then categorized themselves (ie, Beginner, Intermediate, Advanced) for each of the four activity types. Next, the program helped the user build a tailored weekly exercise plan for each activity type (eg, Beginner stretching: 5 stretches, 3 days per week; Intermediate endurance: 30 minutes per day, 5 days per week). Given the sedentary target population, the program was not designed to provide aerobic exercise. Since we could not control for inaccurate self-categorization of fitness level (eg, sedentary individual selects Advanced for a category), the activities were designed conservatively. That is, Advanced levels were slight increases in frequency or duration from Intermediate levels, and participants were encouraged by text on their printouts to postpone finishing an activity if it seemed too difficult.
For each activity type, users committed to exercises (eg, type of stretches from a list; type of endurance activity from a list) and scheduled the days of the week when they would do each one. For each of the four activity types, users responded to multiple-choice questions about their confidence in achieving their exercise plan for the following week. If they were not confident, they were interactively asked to adjust either the intensity or duration of that exercise (eg, stretching: decrease number of stretches or days per week; endurance: decrease number of days or minutes per day), and they were queried until they interactively expressed confidence that they could meet their weekly commitment. Users also were encouraged to print the schedule for their reference. The printout included personal goals, next week’s exercise plan and blank exercise tracking sheets, guidelines, and safety tips for each activity type. Finally, the session was summarized by the video narrator who extolled the benefits of following the exercise plan and invited the user back in a week for the next session. While participants could visit the website as often as they liked (eg, to read articles or print out personal exercise plans linked to the user’s ID and password), the next program session was available no sooner than 1 week after the completion of the previous session.
At return visits, users were welcomed back and given video and text support for returning. Based on an interactive self-report about success in adhering to exercise commitments from the previous session, users were appropriately praised and encouraged to continue their efforts. For those who reported no progress, the coach’s message was upbeat, offering praise for coming back, and encouragement to try again. At each visit, the user was offered tailored video support on overcoming self-identified exercise barriers (eg, too tired, lack motivation, lack skills, etc). Each week, new educational material was presented to engage users and enhance their knowledge about how to make exercise a habit over time. Based on the user’s self-reported progress and motivation, changes to the exercise plan from the previous week were recommended, if appropriate. As before, the users selected exercise amounts and schedules for each activity and affirmed their confidence to meet the commitments. At the 12th visit, users were encouraged to maintain their exercise program into the future, making it a habit.
The study was a randomized controlled trial on the Internet with three assessments: pretest (T1), postintervention at 12 weeks after pretest (T2), and 6-month follow-up (T3; see
Research design with participation level from recruitment to T3 assessment.
After approval by an Institutional Review Board for protection of human subjects (IRB), the study was conducted entirely on the Internet. Participants were recruited via a mixture of online recruitment strategies (eg, listservs, advertising on a website for seniors), flyers, newsletters, and announcements supported by service agencies, senior centers, and worksites. Interested individuals linked to an information website, which offered a link to an online screening questionnaire to determine eligibility (
The online screening questionnaire asked respondents a total of 14 questions about current PA levels (ie, frequency and duration of exercise), desire to exercise more (ie, yes/no), demographics (ie, age, gender, race/ethnicity, employment status, computer use), a working email address, and access to a computer Internet connection. Participants were required to be at least 55 years of age, with a reported desired to engage in more PA. Maximum self-reported exercise levels were: (1) no more than 60 minutes per week of moderate exercise, defined as exercise that increases heart rate, with (2) no bouts of continuous exercise lasting 35 minutes or more. Each respondent answered questions from the 7-item Par-Q [
Individuals who qualified for the study read and agreed to an online informed consent. They then provided contact information, which was checked for fraud before they were randomized by the database into Tx and Ctrl groups. Blinding of the research team to the participants’ research condition was unnecessary.
Our previous Internet research has found a few applicants who attempted to screen-in to a study by providing false information. Consequently, in this study, participant data were checked against our database of about 6000 records from previous Internet study applicants, for fraudulent information (eg, same name or IP address shows inconsistent age, gender, or ethnicity). Screened-in participants providing suspicious data were telephoned, and if the inconsistencies were not resolved, the individual was excluded from the study. A total of 38 were dropped from the study, including 19 before T1, who were not randomized, and 19 after randomization (13 Tx; 6 Ctrl) who were discovered between T1-T3 and were then dropped from the study. Personal privacy was protected with unique user ID and passwords once a participant was accepted into the study and provided contact information. Only passwords provided to Tx participants could link to the intervention.
After completing T1, Tx group participants were mailed a web-enabled CD-ROM (WECD) and emailed log-in information to the
Tx group members were asked by a flyer in the WECD mailer to visit the website within a week. One week after a visit to
Twelve weeks after T2, and 6 months after T1, all participants still enrolled in the study were emailed a link to the T3 assessment. After completion of T3, Ctrl participants and individuals who initially were screened out of the study, but who expressed interest in using the site, were emailed a link to access the
The protocol for prompting individuals who failed to submit surveys or to complete intervention visits included up to 5 emails over a 1-month time period. They were followed by a single phone call attempting to verify that technical difficulties were not responsible for the lack of participant communication. The individual was then dropped from the study if participation was not re-established. This protocol was developed with the approval of our IRB in other studies. We believe that it allowed for contentious follow-up of consented participants without undo harassments.
All participants were mailed a $25 check after submitting each survey. Participants in the Tx group did not receive a financial incentive to use the intervention website.
The T1, T2, and T3 assessments were adapted from our previous Internet research on sedentary factory workers [
Each participant’s self-reported current activity level was measured with 2-item sets addressing the frequency and duration of intentional physical activities that included: (1) cardiovascular activities to increase heart rate (eg, walking briskly, swimming, bicycling, or mowing the lawn), (2) stretching activities to improve flexibility, (3) strength building activities, and (4) balance enhancement activities. For each category, one item asked “In a typical week, how many days do you intentionally…?”, and a pull-down menu offered choices between 0-7 days a week. The second statement asked “How many minutes do you typically … on each of those days?” and a pull-down menu offered choices of from 5-60+ minutes in 5-minute increments. Items were scored to reflect minutes per week of each activity. Scores showed substantial skew, so a log to base 10 transformation was applied.
For a sedentary individual, an increase in the number of physical activities, even if they were not categorized as intentional exercise, would indicate an improvement over a sedentary lifestyle. Participants were asked to report on activities they engaged in during the previous week. They were presented with a list of 16 typical physical activities of older adults: yard work, housework, doing exercises, toe raises or stretches, dancing alone as a physical activity, going for a walk for 10 minutes or more, using the stairs instead of an elevator, parking farther away from the store, exercising with others, playing with children, attending activity classes, dancing/square dancing, bowling or other active games, going to a museum, park, or mall, playing golf, and other physical activities. A count of the number of activities engaged in during the previous week was computed for analysis.
The SF-12 is a 12-item survey that has proven useful in monitoring health outcomes [
The Body Mass Index (BMI) has been used as a way to classify sedentary (physically inactive) individuals with an average body composition by the World Health Organization (WHO) [
The theory of planned behavior suggests that an individual’s attitude and knowledge will shape self-efficacy and intention [
The importance of behavioral self-efficacy to exercise adherence is supported by both social cognitive theory [
The theory of planned behavior suggests that behavioral intentions can predict exercise behavior [
We found no research to adequately measure the motivation of sedentary individuals to exercise, but improvement on this variable should be linked with an increase in PA. Consequently, we adapted the motivation item used by Irvine et al [
A positive change in perceived ability of an individual to perform day-to-day activities was hypothesized to be a measure of improved physical fitness. To assess these capabilities, scales from previous research, eg, [
One goal of the intervention was to change perceptions about possible barriers to participating in physical activity, which research suggests are the reasons many individuals fail to adopt and/or maintain exercise habits, eg, [
If the research intervention was successful, a progression of participants along the continuum of change would be expected. Stage of change (SOC), ie, precontemplation, contemplation, action, maintenance, which assesses an individual’s readiness to adopt new behavior, was measured using four items developed by Marcus, Rossi, Selby, Niaura, and Abrams [
User acceptance of the intervention was measured with ratings of perceived satisfaction. Tx group participants responded to additional items relating to their subjective opinions of the
A total of 405 participants, including 200 Tx and 205 Ctrl group participants were randomized into the study after consenting, and a total of 368 participants completed the T1 assessment including Tx (n=178) and Ctrl (n=190;
The two experimental groups were compared on baseline characteristics and pretest outcome measures. With respect to baseline characteristics, the only significant difference was obtained for race/ethnicity: compared to the Ctrl participants, Tx participants were less likely to be Caucasian, ie, 53% vs. 64%; chi-square (1, N=368) = 4.46,
Over the course of the study, a total of 84 (62 Tx; 22 Ctrl) of the 405 randomized participants were unresponsive to repeated prompts and were dropped from the study, and 19 participants (13 Tx; 6 Ctrl) were removed as fraudulent during the 6-month period between T1-T3 assessments. Of the Tx group participants, only 145 of the 178 who submitted the T1 assessment logged on to initially use the intervention, and 6 of those participants did not complete Visit 1. A total of 92 (73.6%) of those completing T3 assessments (ie, 51.7%) from the T1 Tx group completed all 12 sessions.
Thus, out of the 178 Tx Group participants at T1, 125 (70.2%) eventually remained in the study to T3. A total of 305 participants (125 Tx group; 180 Ctrl group) submitted a T2 assessment, and 302 (125 Tx group; 177 Ctrl group) submitted a T3 assessment. Overall, T1-T3 attrition was (368-302)/368 = 17.9%.
A significantly higher attrition rate was obtained for the Tx condition compared to the Ctrl condition with 30% vs. 7%; chi-square (1, N=368) = 32.84,
Rates of missing study outcomes ranged from 0-1% at T1, 18-21% at T2, and 19-22% at T3. The full-information maximum likelihood estimators assume data are at least missing at random (MAR). It is not possible to know for sure that data are MAR because information about the value of the missing data is not available. However, given the abovementioned significant associations between attrition and study outcomes at baseline, the MAR assumption appears less tenable. Therefore, the main outcome analyses were conducted with (1) available data (ie, “complete cases”, n=294 to 300, dependent outcome), and (2) one fully-imputed dataset that included all 368 study participants. Since the inclusion of additional predictors in the imputation model can reduce bias and make the MAR assumption more plausible [
Sequential regression multiple imputation (SRMI [
Results indicate that the available data approach and imputed data approach resulted in a similar pattern of results. Following the intent-to-treat approach, results from the imputed model are reported below.
As mentioned, 125 (70%) of the original Tx group remained in the 12-week study. The mean number of visits to the website for these individuals was 15.2 visits (SD 9.02). The mean total time spent using the program summed across all visits was 123.4 minutes (SD 185.98), and the mean time spent per visit was 9.66 minutes (SD 10.48). Participants each accessed an average of 2.92 (SD 4.30) program segments designed to help overcome specific perceived barriers to exercise.
A 2 x 2 (condition by race/ethnic minority status) MANCOVA was conducted on the posttest outcome measures in which the pretest outcome measures were included as covariates. The dependent measures included: (1) physical activity measures, (2) SF-12 physical and mental composite measures, (3) BMI, and (4) psychosocial measures (
An overall multivariate model was tested at posttest, followed by univariate models for each outcome measure. Partial eta-square was used as the estimate of the effect size; values of .01, .06, and .14 represent small, medium, and large effect sizes, respectively [
ANCOVA results for the outcome measures.
Outcome measure/ Condition | T1-T2 condition effect | T1-T3 condition effect | ||||
|
|
Eta2 a |
|
|
Eta2 a | |
Cardiovascular activities | 26.32 | <.001 | .067 | 19.38 | <.001 | .050 |
Stretching activities | 25.71 | <.001 | .070 | 23.36 | <.001 | .060 |
Strengthening activities | 42.70 | <.001 | .105 | 19.03 | <.001 | .050 |
Balance activities | 37.26 | <.001 | .092 | 32.37 | <.001 | .081 |
Activities min/wk | 25.4 | <.001 | .068 | 17.98 | <.001 | .049 |
SF-12 physical | 5.19 | .023 | .015 | 7.56 | .006 | .021 |
SF-12 mental | 11.41 | .001 | .032 | 9.51 | .002 | .026 |
BMI (kg/m2) | 1.04 | .309 | .003 | 4.42 | .036 | .012 |
Attitudes/Knowledge | 10.16 | .002 | .028 | 8.57 | .004 | .024 |
Self-efficacy | 9.08 | .003 | .025 | 13.05 | .001 | .036 |
Behavioral intentions | 38.99 | <.001 | .1 | 23.38 | <.001 | .063 |
Motivation to exercise | 21.22 | <.001 | .057 | 22.47 | <.001 | .06 |
Ability to exercise | 4.14 | .043 | .012 | 6.85 | .009 | .019 |
Barriers to exercise | 8.67 | .003 | .024 | 8.26 | .004 | .023 |
a Partial eta-square (effect size): .01 small, .06 medium, .14 large.
To examine the maintenance of program effects at follow-up, an overall 2 x 2 MANCOVA model was tested comparing the two conditions on the follow-up outcome measures, controlling for pretest measures, followed by univariate ANCOVA models. The multivariate model at follow-up was significant in which the Tx participants were found to maintain large gains compared to the Ctrl participants,
Stage of change groupings are compared in
Stages of change groups by condition at T1, T2, and T3.
|
|
Pre-contemplation | Contemplation | Preparation | Action | Maintenance | |||||
|
N | % | N | % | N | % | N | % | N | % | |
|
|||||||||||
|
Treatment | 3 | 2 | 91 | 52 | 24 | 13.6 | 7 | 4 | 51 | 29 |
|
Control | 3 | 2 | 97 | 52 | 37 | 19.8 | 11 | 6 | 39 | 21 |
|
|||||||||||
|
Treatment | – | – | 14 | 11 | 19 | 15.4 | 28 | 23 | 62 | 50 |
|
Control | 8 | 5 | 57 | 33 | 24 | 13.9 | 19 | 11 | 65 | 38 |
|
|||||||||||
|
Treatment | 1 | 1 | 16 | 13 | 15 | 12 | 13 | 10 | 80 | 64 |
|
Control | 4 | 2 | 46 | 26 | 30 | 17 | 17 | 10 | 79 | 45 |
a Chi square (4, N=363) = 5.12,
b Chi square(4, N=296) = 28.79,
c Chi square (4, N=301) = 13.61,
A dose-response analysis was conducted to examine whether the level of exposure to the program was significantly associated with pretest-posttest change in the outcome measures for the participants assigned to the treatment condition. A composite dose measure was created by standardizing the total time spent across all sessions, the number of page views, and the number of sessions, and computing the mean value of the three standardized scores (Cronbach alpha = .79). A composite outcome measure was created by calculating pretest-posttest gain scores for each of the 13 outcome measures, standardizing each gain score, and computing the mean value of the standardized gain scores (Cronbach alpha = .85). The correlation between the composite dose measure and the composite gain score was significant,
On a 7-point scale (eg, not at all satisfied – extremely satisfied), participants reported they were quite satisfied with the program (M=5.6, SD 1.3), the program was very easy to use (M=5.9, SD 1.2), the overall information was very helpful (M=5.9, SD 1.2), the articles provided by the program were very helpful (M=5.7, SD 1.3), and they would be very likely to recommend the program to a friend or family member (M=5.7, SD 1.4). On a 5-point scale to determine opinions about the number of sessions, 89 participants (70%) rated “Just Right,” and 27 (21%) rated “A few too many”.
This randomized effectiveness trial to evaluate the
The strength of the results presented here may actually be diluted somewhat by the measurement process. Treatment subjects who were late completing courses or had stopped taking them were exposed to relatively less of the intervention before the T2 assessment. Additionally, the Ctrl group results generally improved over time, possibly indicating that the assessment process might have brought about reactive effects. Just being exposed to those questions might have sensitized the Ctrl group to improve their level of physical activity and related cognitions.
While defining and identifying a sedentary population of older adults is an inexact science, the screening was designed to disqualify any applicant who engaged in moderate exercise for more than 60 minutes per week, which is 40% of the minimum recommended level [
The fact that 92 of the 139 participants (66%) who completed Visit 1 finished all 12 sessions without payment to do so, suggests that the PA intervention was engaging enough for a majority to follow through for 12 weeks. This is a very positive sign, but we found no attrition results from comparable Web research against which to measure our results. As indicated by the dose-response analysis, greater program utilization resulted in greater change in outcomes, which has been reported by other [
Dropouts in this study had lower attitude/knowledge scores about the benefits of exercise, and they had higher perceptions of barriers to exercise. While the results presented here must be viewed cautiously until validated by other research, they might offer a hint as to why sedentary individuals fail to start to engage in a Web PA program. Other research attributes attrition from PA programs to unrealistic participants’ expectation [
This research also presents data on recruitment success using an informational website and automated online screening (
Also of potential interest to other researchers is the incidence of fraud reported here. Of the 424 individuals who screened in and agreed to the informed consent, 19 were dropped because of fraudulent information before being randomized, and another 19 subjects were discovered to be fraudulent and were dropped between T1-T3. Thus, 9% of those who screened in provided false information. We have experienced roughly similar numbers in other Internet studies, causing us to set up our fraud database to cross-check participant information. Some individuals are repeat offenders, and we even had one fraudulent participant in another study complain to us when confronted on the telephone that removal from a study constituted mistreatment by us because she said, “I’m only trying to make a living.” We believe that the potential for fraudulent participation in research studies on the Internet is an important issue, but we are aware of no research into the frequency of occurrence or steps to minimize it.
The current results must be viewed cautiously because we have no evidence that the participants actually engaged in PA or provided accurate information. Additionally, some of the measures were not validated in other research, physical functionality was not measured beyond the SF-12 physical sub-scale, and the follow-up period was somewhat limited (ie, 6 months). Follow-up studies of 1-2 years, using participant exercise logs and verifiable measures of PA (eg, treadmill testing; 6-minute walk testing), and functionality (eg, Physical Functional Performance Test [
Participants were a relatively young population of older adults (ie, M=60 years of age), and they tended to be employed, educated, and frequent computer users with at least a middle-class income. Less educated, lower income, rural, and ethnic populations might be less likely to have Internet in their homes [
The higher attrition rate among treatment participants compared to control participants is another limitation that may have biased the study findings. However, because experimental conditions did not interact significantly with any of the baseline participant characteristics in predicting study attrition, the potential confounding due to differential attrition would appear to be minimal. Furthermore, the use of maximum likelihood estimation for missing data would help to reduce potential biases associated with study attrition.
Despite limitations, this research demonstrates that a theoretically based stand-alone Internet exercise program that tailors content according to users’ preferences and interests can increase self-reported PA and be well received by sedentary older adults. This type of intervention can be available to users 24/7 on the Internet, making it a potentially cost-effective PA tool that can reach large numbers of people. The results are impressive considering that the study was not conducted as part of a larger health promotion campaign, which might have provided additional support and encouragement for the participants and which might have decreased attrition. Still, more research is needed to understand factors associated with using Internet interventions to maintain engagement in PA over time.
Demographic information for participants.
Pretest, posttest, and follow-up descriptive statistics (untransformed values are reported for cardiovascular activities, stretching activities, strengthening activities, and balance).
CONSORT-Ehealth Checklist V1.6.2 [
This research was funded by a grant to the first author from the US National Institutes of Health, National Institute on Aging (R44AG20002). Special thanks to Sid Stahl, PhD, and Marcia Ory, PhD from NIA. The project required the efforts of a multidisciplinary team. Molly Billow and Ellie Price led on evaluation logistics. The program development team included Molly Billow, Gretchen Boutin, Rob Fightmaster, Rob Hudson, Brian Johnson, Beth Machamer, Jennifer Monte, Neil Moyse, Percy Perez, Ellie Price, Diana Robson, Lee Amberwood, Ann Benbow, PhD, Ker Cleary, Brian Enos, Nancy Hawkins, PhD, David Kerr, Michael Manocchia, PhD, Nelda Mier PhD, Eva Montee, and Patricia Smith. Dennis Ary, and Ann Glang made helpful comments on earlier versions of the manuscript, and Elizabeth Greene helped prepare it for publication. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Aging or the National Institutes of Health.
Blair Irvine was the grant PI. He is employed as a Research Scientist at ORCAS, a health care technology company that creates self-management programs to improve physical and emotional well-being. The software is not for sale, and he and the other authors derive no financial benefit from development of the software or from publication of this research.