http://heb.sagepub.com/Health Education & Behavior

http://heb.sagepub.com/content/early/2014/08/22/1090198114543004The online version of this article can be found at:

 DOI: 10.1177/1090198114543004

published online 25 August 2014Health Educ BehavPrabu David, Michael L. Pennell, Randi E. Foraker, Mira L. Katz, Janet Buckworth and Electra D. Paskett

How Are Previous Physical Activity and Self-Efficacy Related to Future Physical Activity and Self-Efficacy?  

Published by:

http://www.sagepublications.com

On behalf of: 

  Society for Public Health Education

can be found at:Health Education & BehaviorAdditional services and information for    

  http://heb.sagepub.com/cgi/alertsEmail Alerts:

 

http://heb.sagepub.com/subscriptionsSubscriptions:  

http://www.sagepub.com/journalsReprints.navReprints:  

http://www.sagepub.com/journalsPermissions.navPermissions:  

at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from

Health Education & Behavior 1 –4© 2014 Society for PublicHealth EducationReprints and permissions: sagepub.com/journalsPermissions.navDOI: 10.1177/1090198114543004heb.sagepub.com

Brief Report

Physical activity (PA) is integral to health and wellness (Kokkinos, 2008). Despite this, only 36% to 46% of women aged 45 years and older in the United States meet PA guide-lines of 30 minutes of activity per day on most days of the week (Centres for Disease Control and Prevention, 2010). An extensive body of research is available on promoting PA and constructs from social cognitive theory (Bandura, 1977b), such as outcome expectations, setting goals, self-regulation, and self-efficacy (SE) are widely used in inter-ventions. Among these constructs, SE has received considerable attention as a key determinant of sustained PA (McAuley & Blissmer, 2000).

Though SE can be conceptualized both as a trait and a state (Bandura, 1977a), in much of the extant research, SE has been examined as a trait or individual difference and found to be a robust predictor of PA. SE can also be concep-tualized as a within-individual state that fluctuates over time, potentially having a varying effect on PA and in these studies SE is typically measured at intervals that are weeks apart (e.g., Blanchard et al., 2007). For example, in one study, four assessments over a period of 6 months were used to model changes in SE among older adults (McAuley, Jerome, Marquez, Elavsky, & Blissmer, 2003). However, to our knowledge, a fine-grained analysis of day-to-day variation in SE and its relationship to daily PA have not been reported. Furthermore, despite the strong relationship between SE and

PA in cross-sectional analysis, longitudinal analysis suggests that SE decreases over time (Blanchard et al., 2007) and the relationship between SE and PA weakens over time (e.g., McAuley et al., 2003). To further understand the temporal interplay between SE and PA, day-to-day variations in SE and PA were examined over 12 weeks.

Fluctuations in SE are inevitable (McAuley & Blissmer, 2000), and over the course of an intervention the relationship between SE and PA can take on different patterns (Dishman et al., 2005). SE can lead to success in reaching PA goals, which in turn can reinforce SE, thus creating a cycle of

543004 HEBXXX10.1177/1090198114543004Health Education & Behavior XX(X)David et al.research-article2014

1The Edward R. Murrow College of Communication, Washington State University, Pullman, WA, USA2Division of Biostatistics, College of Public Health, Ohio State University, Columbus, OH, USA3Division of Epidemiology, College of Public Health, Ohio State University, Columbus, OH, USA 4Division of Health Behavior and Health Promotion, College of Public Health, Ohio State University, Columbus, OH, USA5Department of Kinesiology, University of Georgia, Athens, GA, USA6Department of Internal Medicine, College of Medicine, Ohio State University, Columbus, OH, USA

Corresponding Author:Prabu David, Washington State University, CADD 101, Pullman, WA 99164, USA. Email: prabu.david@wsu.edu

How Are Previous Physical Activity and Self-Efficacy Related to Future Physical Activity and Self-Efficacy?

Prabu David, PhD1, Michael L. Pennell, PhD2, Randi E. Foraker, PhD3, Mira L. Katz , PhD4, Janet Buckworth, PhD5, and Electra D. Paskett, PhD6

AbstractSelf-efficacy (SE) has been found to be a robust predictor of success in achieving physical activity (PA) goals. While much of the current research has focused on SE as a trait, SE as a state has received less attention. Using day-to-day measurements obtained over 84 days, we examined the relationship between state SE and PA. Postmenopausal women (n = 71) participated in a 12-week PA intervention administered via cell phone and monitored their daily PA using a pedometer. At the end of each day, they reported their state SE and number of steps. Using a longitudinal model, state SE was found to be a robust predictor of PA even after accounting for trait SE and other covariates. The findings offer insights about the temporal relationship between SE and PA over the course of an intervention, which can be of interest to researchers and intervention designers.

Keywordsbehavioral intervention technology, health behavior change, mixed models, self-efficacy

at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from

2 Health Education & Behavior

reinforcement that is akin to “success-breeding-success.” On the other hand, failure to reach PA goals can lead to erosion in SE, which in turn can lead to further failure, thus perpetu-ating a downward spiral of “failure-breeding-failure.” In light of these possibilities, a better understanding of the rela-tionship between SE and PA was pursued by examining the paths (labeled a through f) in Figure 1, which were estimated from three models summarized in the hypotheses.

Hypothesis 1: Yesterday’s SEt − 1 is a function of yester-day’s PAt − 1 behavior (c).Hypothesis 2: Today’s SEt is a function of yesterday’s PAt − 1 (f) and yesterday’s SEt − 1 (b).Hypothesis 3: Today’s PAt is a function of yesterday’s PAt − 1(a), yesterday’s SEt − 1 (d) and today’s SEt (e).

Method

In all, 71 postmenopausal women participated in a 12-week PA intervention administered via cell phone and an interactive voice response (IVR) system and 39 completed the interven-tion. Inclusion criteria for the study were body mass index of 25 to 40 kg/m2, postmenopausal status, access to a cell (mobile) phone for 12 weeks during the intervention, and willingness to walk at least 30 minutes a day on most days. The study protocol, including the informed consent procedure, was approved by the Ohio State University Institutional Review Board in June 2007. Data collection took place between January 2008 and March 2009 and data analysis was completed in March 2013. Details of the study, including design, intervention, and attrition have been previously reported (David et al., 2012).

Study participants were provided with a daily-steps goal and wore a pedometer to monitor their steps. The goals were gradually increased every week until the 10,000 steps/day goal was reached. Two daily phone interactions with the IVR were scheduled. An outbound call from the IVR went out to

the participants’ cell phone between 7 a.m. and 5 p.m. during a 2-hour time block identified by each participant. The out-bound call was essentially a brief assessment that included three items: whether the participant had walked or planned to walk that day, confidence in achieving the steps goal for the day, and a general item on whether the participant was hav-ing a good or bad day. In addition, participants called the IVR every evening to report their daily step count, state SE, and received an automated intervention message (e.g., “Did you know that there are a lot of health benefits from walk-ing? Walking is good for your heart, lungs, bones, and walk-ing helps prevent cancer, too).

Intervention messages were based on relevant constructs from social cognitive theory (Bandura, 1977a), goal-setting theory (Locke & Latham, 1984), problem-solving theory (D’Zurilla & Nezu, 2007), and the transtheoretical model (Prochaska & Norcross, 2001). A unique messaging theme was used each week (e.g., SE, outcome expectancies, increas-ing knowledge and awareness, self-monitoring, awareness of barriers, coping with negative thoughts, problem solving, goal setting, and social support). Daily messages, which were between 15 to 30 seconds in duration, reinforced the weekly theme. Each week, a bar chart summarizing step counts from the preceding seven days was sent to the participant via e-mail or surface mail. A brief message of reinforcement or encour-agement was added to the summary chart and the goal for the following week was provided. In addition, a summary of the intervention messages heard during the week was included.

Measures

Trait SE was assessed at baseline through a 15-item instrument (Garcia & King, 1991). Items (e.g., I could walk when I am tired, I could walk when my schedule is hectic) were rated on a 0% to 100% scale (0% = Cannot do it at all, 100% = Certain that I can do it) (Cronbach’s α pre = .95). State SE was obtained at the end of each day along with PA and an overall assessment of the day. State SE was measured with the following item: “On a scale of 1 to 7, where 1 = not at all confident, and 7 = very confident, how confident are you that you can achieve your steps goal tomorrow?” PA was obtained using the follow-ing prompt: “Please check your pedometer and enter the num-ber of steps walked today followed by the # sign.” In addition, participants also provided the following assessment: “Taking into consideration everything that happened today, on a scale of 1 to 7, where 1 = very bad day and 7 = very good day, how would you rate today?” which was introduced as a covariate.

Data Analysis

Over the 84-day intervention, participants completed 66% of the calls to the IVR. An analysis based only on completed calls could yield biased estimates if the reasons for missed calls were related to measured predictors (Ibrahim, Chu, & Chen, 2012). To avoid these potential biases, missing call

Figure 1. Standardized coefficients (standard error) of direct effects predicted in Hypotheses 1 to 3.Note. PA = physical activity; SE = self-efficacy. Solid paths significant at p < .05.

at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from

David et al. 3

data were imputed using the Markov chain Monte Carlo method in SAS PROC MI (SAS Inc., Cary, NC). The Markov chain Monte Carlo method was chosen because missingness was nonmonotonic. PROC MI ANALYZE was used to com-bine results from 30 different imputed data sets.

Imputation was performed based on our model for Hypothesis 3. Missing PA was imputed using yesterday’s end of day SE and today’s SE, missing yesterday’s end of day SE was imputed using today’s SE and today’s PA, and missing today’s SE was imputed using yesterday’s end of day SE and today’s steps. Yesterday’s PA was not included in the imputa-tion because this would have required incorporating random effects which are not possible given our missing data struc-ture and capabilities of commercial software. Baseline psy-chometric measures (exercise goals, exercise planning, negative exercise thoughts, social support from family and friends, and walking SE) and anthropometrics (waist/hip ratio, pulse rate, and body mass index) were also used in imputing the missing PA and SE data.

Linear mixed models (SAS PROC MIXED) were used to test hypotheses. Prior to analysis, each variable was stan-dardized by subtracting the overall mean and dividing by the sample standard deviation. For Hypothesis 1, yesterday’s end of day SE was regressed on yesterday’s steps as fixed effect and participant and participant-by-steps interaction were modeled as random effects. For Hypothesis 2, today’s SE was regressed on yesterday’s end of day SE and yester-day’s steps as fixed effects. Instead of random effects, within participant correlations in Hypothesis 2 were accounted for by using autoregressive, AR(1), covariance structure for residual errors. Finally, for Hypothesis 3, today’s steps was regressed on yesterday’s steps, yesterday’s SE and today’s SE. As in the Hypothesis 2 analysis, correlations between steps measured on the same subject were accounted for using an AR(1) covariance structure for the residual errors, which provided a lag-1 correlation estimate that could be used to assess the relationship between steps walked on adjacent days. Subsequently, trait SE and type of day were added to each model as covariates and the results were similar.

Results

As predicted in Hypothesis 1, yesterday’s PA was positively associated with yesterday’s SE (β = .12, SE = .02). Yesterday’s SE, in turn, showed a lagged effect and was positively asso-ciated with today’s SE (β = .54, SE = .03), which supports Hypothesis 2, whereas yesterday’s PA was not significantly associated with today’s SE. The third hypothesis focused on today’s PA as a function of yesterday’s PA, yesterday’s SE, and today’s SE. Yesterday’s PA was significantly related to today’s PA (lag-1 correlation = .29, SE = .02) and yesterday’s SE was significantly related to today’s PA (β = .07, SE = .03). After controlling for the effects of yesterday’s SE and yester-day’s PA on today’s PA, the effect of today’s SE on today’s PA was significant (β = .30, SE = .02).

The above analysis was repeated without multiple impu-tation (MI) and the conclusions were similar, with one excep-tion. Because the imputation ignored the relationship between yesterday’s and today’s steps (due to limitations of existing software), the correlation between the two was stronger in the complete case analysis (r = 0.47) than in the MI analysis (r = 0.36), though the relationship was signifi-cant in both analyses (p < .001). Finally, given that SE is a crucial variable in this study, we examined whether PA was different during the week when daily messages emphasized SE in comparison to the rest of the weeks and found that there were no significant difference.

Discussion

The first hypothesis predicted that the number of steps on a particular day will influence SE on that day. Unlike trait SE that focuses more on between-person differences, emphasis of this hypothesis was on the fluctuations in SE from proxi-mal PA. When SE is measured along with a behavioral report of PA, a positive association between the two is to be expected and the results support the prediction that SE shifts up or down with a corresponding increase or decrease in the number of steps reported. The direct association between state SE and PA was significant after accounting for trait SE and participant’s “type of day.”

The second hypothesis sheds light on the carry-over effect of SE from one day to another. Although SE has been found to have a robust effect on PA, our findings emphasize the importance of state SE, which is strongly influenced by carry-over effects from the previous day. Although yester-day’s SE had a significant relationship with today’s SE, yes-terday’s PA did not contribute to today’s SE.

The third hypothesis focused on today’s PA as a function of three components: today’s SE, yesterday’s SE, and yester-day’s PA. The findings reveal that SE reported on the same day as PA has as much influence as yesterday’s PA, thus rein-forcing the robust tie between SE and PA. The direct effect of yesterday’s SE on today’s PA was weak. Yesterday’s SE seemed to operate on today’s PA mainly as an indirect effect mediated through today’s SE; to demonstrate this, we removed today’s SE from our Hypothesis 3 model and found that the effect of yesterday’s SE on today’s PA increased dra-matically (from .07 to .24).

To our knowledge, a fine-grained analysis of daily fluctua-tions in PA and SE has not been previously reported. To fill this gap, the relationship between SE and PA was examined using daily reports collected over 84 days. SE is a robust pre-dictor of success in achieving PA goals and, many health interventions use various approaches to bolster SE to achieve PA success. In the past, various tailored interventions have been implemented based on between-individual differences in SE. The findings from this study offer empirical evidence to extend tailoring to within-individual fluctuations in SE. A better understanding of the carry-over effects of SE and PA

at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from

4 Health Education & Behavior

presented here may be the first step toward fine-grained tai-lored interventions in the future. Our findings were consistent with other studies that have found that PA can be increased through interventions administered via mobile technologies, such as personal digital assistants (PDAs), smartphones, and telephones (King et al., 2013). By combining other features, such as global positioning system (GPS) and accelerometer, more relevant motivational messages that are sensitive to the relationship between SE and PA can be offered (Taraldsen, Chastin, Riphagen, Vereijken, & Helbostad, 2012).

Small sample, high attrition, and reliance on MI to gener-ate data are some limitations of this study. Nonreporting of intervention activities and more than 30% attrition are com-mon in physical activity interventions and this study is no exception. Future studies should consider follow-up of par-ticipants to examine whether missing data can be attributed to chance (e.g., participant forgot to report data) or is corre-lated to outcomes (e.g., not achieving steps goal). Also, dynamic modeling using feedback loops could offer further insight on the interplay between SE and PA.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by the Breast Cancer Research Foundation.

References

Bandura, A. (1977a). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215.

Bandura, A. (1977b). Social learning theory. Englewood Cliffs, NJ: Prentice Hall.

Blanchard, C., Fortier, M., Sweet, S., O’Sullivan, T., Hogg, W., Reid, R., & Sigal, R. (2007). Explaining physical activity levels from a self-efficacy perspective: The physical activity counseling trial. Annals of Behavioral Medicine, 34, 323-328.

Centres for Disease Control and Prevention. (2010). U.S. physical activity statistics. Atlanta, GA: Author.

D’Zurilla, T. J., & Nezu, A. M. (2007). Problem-solving therapy: A positive approach to clinical intervention (3rd ed.). New York, NY: Springer.

David, P., Buckworth, J., Pennell, M., Katz, M., DeGraffinreid, C., & Paskett, E. (2012). A walking intervention for post-menopausal women using mobile phones and interactive voice response. Journal of Telemedicine and Telecare, 18, 20-25.

Dishman, R. K., Motl, R. W., Sallis, J. F., Dunn, A. L., Birnbaum, A. S., Welk, G. J., & Jobe, J. B. (2005). Self-management strat-egies mediate self-efficacy and physical activity. American Journal of Preventive Medicine, 29, 10-18.

Garcia, A., & King, A. (1991). Predicting long-term adherence to aerobic exercise: A comparison of two models. Journal of Sport & Exercise Psychology, 13, 394-410.

Ibrahim, J. G., Chu, H., & Chen, M.-H. (2012). Missing data in clin-ical studies: Issues and methods. Journal of Clinical Oncology, 30, 3297-3303.

King, A. C., Hekler, E. B., Grieco, L. A., Winter, S. J., Sheats, J. L., Buman, M. P., & Cirimele, J. (2013). Harnessing differ-ent motivational frames via mobile phones to promote daily physical activity and reduce sedentary behavior in aging adults. PLoS ONE, 8(4), e62613.

Kokkinos, P. (2008). Physical activity and cardiovascular disease prevention: Current recommendations. Angiology, 59(2 Suppl.), 26S-29S.

Locke, E. A., & Latham, G. P. (1984). Goal setting for indi-viduals, groups, and organizations. Chicago, IL: Science Research.

McAuley, E., & Blissmer, B. (2000). Self-efficacy determinants and consequences of physical activity. Exercise and Sport Sciences Reviews, 28, 85-88.

McAuley, E., Jerome, G. J., Marquez, D. X., Elavsky, S., & Blissmer, B. (2003). Exercise self-efficacy in older adults: Social, affective, and behavioral influences. Annals of Behavioral Medicine, 25, 1-7.

Prochaska, J. O., & Norcross, J. C. (2001). Stages of change. Psychotherapy: Theory, Research, Practice, Training, 38, 443-448.

Taraldsen, K., Chastin, S. F., Riphagen, I. I., Vereijken, B., & Helbostad, J. L. (2012). Physical activity monitoring by use of accelerometer-based body-worn sensors in older adults: A systematic literature review of current knowledge and applica-tions. Maturitas, 71, 13-19.

at Bobst Library, New York University on October 18, 2014heb.sagepub.comDownloaded from