Prognostic factors of adherence 67 Table 1 Continued Health-system related prognostic factors of exercise adherence No prognostic factors - - - - 1Phase 1 and 2 studies and hence downgraded by 1. 2Significant heterogeneity among studies but all studies have same direction of effect and hence not downgraded. 3Serious limitations and hence downgraded by 1. 4No limitations or inconsistency and hence not downgraded. 5Inconsistency and hence downgraded by 1. 6Publication bias could not be assessed except for self-efficacy and exercise history, therefore not included the grading. CI: Confidence Interval; OR: odds ratio Patient-related factors Patient-related prognostic factors reported in more than one study included, selfefficacy, exercise history, intention, motivation, attitude, Perceived Behavioral Control (PBC) and perceived benefits. Higher self-efficacy, having an exercise history, motivation and PBC were the patient-related factors to be predictive of exercise adherence. High-quality evidence suggested that having higher PBC predicted higher exercise adherence. Moderate-quality evidence suggested that higher self-efficacy and having an exercise history predicted higher exercise adherence. The pooled ORs showed a significant better adherence rate; selfefficacy OR = 1.58 (95%CI 1.27, 1.97; I2 = 82%), exercise history OR = 4.05 (95%CI 1.10, 0; I2 = 76%), motivation OR = 1.25 (95%CI 1.12, 1.39; I2 = 69% and PBC OR = 1.21 (95%CI 1.07, 1.36; I2 = 0%) (Figure 2). To address the heterogeneity, we performed subgroup analyses. When examining the covariates disease (cancer vs. other) and study design (cohort studies vs. other studies) for self-efficacy and exercise history a significant subgroup difference could not be found (self-efficacy p = 0.47 and p = 0.67; exercise history p = 0.15 and p = 0.51). Forest plots revealed three outliers in the prognostic factor self-efficacy; Albert et al. [37], McCaul et al. [35] and Caetano et al. [42], studies with, high, high and moderate risk of bias. In a sensitivity analysis these three studies were removed. By removing these three studies, subgroup differences remained not significant (p = 0.57 and p = 0.51). However, the I2 of the subgroup cohort studies went from 76% to 49%. The heterogeneity between studies is mainly due to the studies of Albert et al., McCaul et al. and Caetano et al. The forest plot revealed two outliers in the prognostic factor exercise history: Cheng et al. [43] and Woodgate et al. [70], both studies with high RoB. By removing these studies heterogeneity dropped to I2 = 45% and subgroup differences remained not significant (p = 0.16 and p = 0.88), indicating that heterogeneity was mainly due to the studies of Cheng et al. and Woodgate et al.
RkJQdWJsaXNoZXIy MjY0ODMw