The purpose of the present study was to examine the ability of several conceptually related clusters of variables to predict adherence of older adults who were at risk for disability to the LIFE-P physical activity intervention. Our results indicate that demographic variables, disease burden, self-reported symptoms, physical functioning and social cognitive variables together predicted 10% of the variance in percent attendance during both the adoption phase and the transition phase. Furthermore, when we added percent attendance in the adoption phase to the prediction of percent attendance in the transition phase, the amount of explained variance increased from 10% to 21%. During the maintenance phase, the conceptually-based clusters of predictor variables accounted for 13% of the variance in predicting frequency of physical activity; however, this estimate increased to 46% when adding attendance to center-based visits to the model for the transition phase. Clearly, prior behavior in relation to center-based attendance is an important predictor of how frequently participants engage in physical activity when involved at home in an independent environment (Rejeski et al 1997
There was a lack of consistency in the variables that predicted adherence across the various phases of LIFE-P. For example, in the adoption phase, the presence of lung disease and low efficacy to manage barriers to physical activity met the inclusion criteria for the composite model. In the transition phase, both poorer performance on the baseline 400 m walk and presence of a pacemaker were significant predictors of percent attendance in the composite model, whereas predictors in the maintenance phase for this model included the presence of a pacemaker and higher levels of self-reported fatigue. Moreover, the variance accounted for by any of the three composite models never exceeded 13%. We view this evidence as encouraging from a public health perspective in that the LIFE-P intervention appears to be tolerated quite well by diverse subgroups of older adults.
The observed pattern of the individual predictor variables across the different phases of the intervention in LIFE-P is supportive of past research (McAuley 1993
). For example, although self-efficacy has been consistently found to be a reliable predictor of physical activity in older adults (Brassington et al 2002
; McAuley et al 2003
), it has been found to play different roles at different phases in physical activity interventions (McAuley et al 1993
; Oman and King 1998
; Brassington et al 2002
; McAuley et al 2003
). Similar to the present study, McAuley and colleagues (2003)
also found that self-efficacy was a significant predictor of adherence during the adoption phase of an older adult physical activity intervention, not during the transition phase. This finding is consistent with the position that self-efficacy is most influential in predicting behavior during challenging situations (Bandura 1997
) such as attempting to integrate physical activity into the lives of sedentary older adults who have compromised physical function. As physical activity becomes more habitual, different variables become important such as past experience or complications from co-morbidities (McAuley et al 1993
; Brassington et al 2002
; McAuley et al 2003
). Interestingly, in contrast to a study by McAuley and his colleagues (2003)
, the present investigation did not find self-efficacy to be a significant predictor of adherence during the maintenance phase of the intervention; only past attendance and energy/fatigue were significant predictors among LIFE-P participants. We can only speculate about the failure of self-efficacy to predict adherence in the maintenance phase of LIFE-P. In this regard, the adherence problems in the maintenance phase may be rooted in the lack of desire to be physically active as opposed to lacking confidence in the ability to do so, a hypothesis that is indirectly supported by the powerful role that physical activity behavior in the transition phase had in predicting adherence in the maintenance phase – a partial R2
The contribution that the predictor variables had on explaining variance in physical activity attendance is consistent with our previous work with older adults with knee osteoarthritis (OA) (Rejeski et al 1997
). In that study, the predictor variables also accounted for ~10% of the variance in attendance in the adoption phase. Additionally, the most consistent and potent predictor of attendance in LIFE-P was exercise behavior in the previous phases of the trial, a finding that also parallels data from our research on older adults with knee OA. As we mentioned in the introduction, the repetition of intentional behavior is related to habit formation (Maddux and DuCharme 1997
). However, it is also true that participants with better adherence are exposed to greater and more consistent mastery experience. Mastery experience is the most potent source for enhancing self-efficacy beliefs (Bandura 1997
) and self-efficacy is a known determinant of physical activity behavior (McAuley et al 2003
The role that CHF had on days spent in suspended status is intriguing. One might be tempted to conclude from these data that older adults who have CHF are not good candidates for a physical activity intervention similar to the one employed in LIFE-P. However, we have examined the consequence of having CHF on return from suspended status. The evidence suggests that the probability of returning from suspended status in this subgroup is no worse than other causes of suspended status. Thus, instead of using CHF as an exclusion criterion, we would argue that these individuals might have the most to gain from being physically active.
We conclude that the adherence of older adults at risk for disability to physical activity interventions is not related to differences in demographic profiles. Similarly, there was not strong, consistent evidence that adherence is related in a consistent manner to comorbidities, level of physical functioning, physical symptoms, or even social cognitive variables related to functioning and physical activity that exist prior to the onset of an intervention. Although the percent variance accounted for by the models in each phase of the intervention was statistically significant, there was not a single instance where the composite models explained more than 13% of the variability in adherence. These results are heartening in that the physical activity intervention appears to have been well tolerated by diverse subgroups of older adults. In contrast, prior behavior accounted for an additional 11% of the variance in the transition phase and an additional 33% in the maintenance phase. These data underscore the importance of being proactive in countering nonadherence and in responding quickly to individuals who begin to develop patterns of nonadherence. We would caution readers to recognize that the results of this study are limited to older adults in the target population who are motivated to volunteer for a 12 month randomized controlled trial. We do not yet know, but must conduct research on theoretically relevant predictors of adherence to longer term interventions. To this end, our aim is to conduct a large multi-center trial that builds on the experience acquired in LIFE-P.