Xuxi Zhang

among the studies. 48 This statistical heterogeneity can be quantified, but there is usually “uncertainty about the clinical sources of this heterogeneity and how important the differences really are”. 49 Subgroup analyses could be adopted to explore the heterogeneity in the meta ‐ analysis. 48 In Chapter 6 and 7, heterogeneity was moderate to high across the studies included in the meta ‐ analyses. Subgroup analyses were only partly able to explain this variation. A high level of heterogeneity was still observed in some subgroups. Further studies are needed to confirm our findings in some subgroups with high heterogeneity. Clinical meaning of (changes of) the TFI scores In Chapter 3 , statistically significant differences in frailty scores between baseline and follow ‐ up were found. Future studies should explore what changes of the TFI scores correspond to clinically meaningful changes regarding the level of frailty. 50, 51 9.3 RECOMMENDATIONS FOR FUTURE RESEARCH Frailty We recommend further studies on the effect of the frequency and intensity levels of PA in order to determine the optimum level of PA required to prevent the progression of physical, psychological and social frailty among older adults. Our findings indicated that the TFI can be applied as an instrument to study frailty in community ‐ dwelling older adults. We propose to study whether (and how) application of the TFI can be applied in community medicine to identify high risk groups and to promote health. The application of the TFI in health care practice will benefit from the establishment of general population norms or reference scores. 52 Also, the use of the TFI in other settings such as the hospital setting to identify high risk groups and to promote health is recommended. Previous studies suggested targeting modifiable risk factors at midlife might reduce the occurrence of frailty at later ages. 53 54 Therefore, we suggest future research on the development and evaluation of interventions that target people at an earlier age. Chronic conditions Our results showed that peer support can be effective regarding glycemic control for patients with T2DM. We recommend further studies to verify the results of our subgroup analyses (e.g. Telephone ‐ dominant ‐ intervention and Home ‐ visit ‐ intervention). Peer support is complex and could be influenced by many factors like culture, psychology, emotion and social environment. Therefore, we suggest further research to take these contextual factors into consideration. As our study was the first to report variation among different types and different follow ‐ up intervals of interventions targeting modifiable risk factors of diabetic retinopathy, we suggest replication studies to confirm our findings. We also recommend more studies on the effectiveness of interventions targeting various modifiable risk factors in prevention and control of diabetic retinopathy. 9 229 General discussion