Predicting long-term maintenance of physical activity in older adults

Background. Considerable research has established that self-efficacy is a consistent correlate of physical activity. Additional factors, such as exercise-induced affect, social support, and value judgments, have also been identified as having the potential to influence adherence to activity. This study examined the utility of such variables in predicting the long-term exercise behavior of older adults. Method. In the context of a 6-month randomized controlled trial with an 18-month follow-up, we tested the extent to which adherence during the trial, affective responses to exercise, exercise value, and social support from the exercise group had either a direct or indirect effect through the mediation of exercise self-efficacy on physical activity levels at 6- and 18-month follow-up in a sample (N = 174) of older adults (M age, 66 years). Structural equation modeling tested several models of exercise prediction. Results. The best fitting model indicated significant paths from social support, affect, and exercise frequency to efficacy at the end of the program. Efficacy, in turn, was related to physical activity at 6- and 18-month follow-up. The model accounted for 40% of the variance in 18-month activity levels. Conclusion. This prospective study provides support for the inclusion of social cognitive variables in models of exercise adherence and highlights the pivotal role of self-efficacy in long-term exercise behavior. Subsequent trials are called for to replicate and extend these findings. (C) 2003 American Health Foundation and Elsevier Science (USA). All rights reserved.