Chapter 4 74 Strengths and limitations The study findings should be considered in the context of its strengths and limitations. The review was conducted predominantly according to best-practice methodologies, which included protocol pre-registration, development of the search strategy with help of an information specialist, review of multiple databases, a focus on adjusted estimates, no language restriction and contextualization of the findings within the GRADE strength of evidence framework, which benefits the generalizability of the findings of this systematic review. The study has several limitations worth noting. First, despite the previously mentioned facts regarding study design for prognostic factor research, we chose to include both prospective cohort studies and cross-sectional studies in this systematic review. Our goal was not only to gather prognostic evidence, but also to gain insight into all possible variables that were studied. A risk is that the evidence may be downgraded by the cross-sectional studies, however, the subgroup analysis performed by design showed no subgroup differences between cohort studies and the other studies. Also, the level of evidence may have been downgraded because the risk of bias was assessed for all study designs using the QUIPS tool. For cross-sectional studies, the National Institutes of Health (NIH) Quality Assessment Tool for Observational Cohort and Cross- sectional studies is recommended [95]. However, the use of two different tools may complicate the comparison of the risk of bias between studies. Since the signaling questions of both the QUIPS and the NIH have much overlap, we chose to use only the QUIPS tool. The risk of bias now compares well, only the QUIPS assesses cross-sectional studies more strictly than the NIH would. This may have slightly lowered the level of evidence as indicated by GRADE. Second, only the lead researcher (ER) screened all the titles and abstracts, since there was less than 5% difference between the results in the first 25% of screening between ER and AD. This could have limited the validity of the screening process. Third, when undertaking a systematic review, there is always a risk of publication bias where negative studies of predictors not being associated with adherence might be less likely to get published. To reduce this problem a grey literature search of unpublished work was performed. None of the included studies were identified using this search strategy. Further, due to a power issue only of two prognostic factors a funnel plot could be created and the Egger's test for small study effects performed. So, publication bias was not included in the grading of the evidence. This may have influenced the quality of the evidence. Finally, most of the studies did not use validated measures for assessing exercise adherence which limits the strength of their findings. This particular point is indicative of a wider issue around measuring exercise adherence. Implications The present study provides information on possible relevant prognostic factors of home-based exercise adherence in patients with chronic diseases and their
RkJQdWJsaXNoZXIy MjY0ODMw