Objective To investigate the prevalence of frailty in diabetic patients and determine the predictive validity of frailty assessment instrument in poor outcome. Methods The cohort studies about poor outcomes of frailty in diabetic patients were searched during January 2001 and April 2019 in the following electronic bibliographic databases, PubMed, Embase, Web of Science, PEDro, and the Cochrane Library, CINAHL, CNKI, Wanfang, CBM and VIP Data. The data were extracted according to inclusion and exclusion criteria. Meta-analysis was conducted by STATA 15.1 software, and descriptive analysis was conducted for data that cannot be combined. Results A total of 12 cohort studies, containing 563221 participants were included. The reported prevalence of frailty ranged from 0.3% to 74.0%. The frailty condition identified by frailty assessment tools were significantly associated with poor outcomes, such as mortality, hospitalization, disability and fall. The meta-analysis revealed that compared with other tools, FRAIL Scale scores were the most widely used tool in diabetic patients, and showed good predictive value for death and hospitalization risks. Conclusion The prevalence of frailty by different tool assessments varies widely in elderly patients with diabetes mellitus. The FRAIL scale is preferred for primary screening for elderly frail patients with diabetes in the community. However, due to limited studies included, prospective cohort study with large sample is still needed to verify the conclusion.