Objective To establish and validate a risk prediction model for cognitive frailty in community elderly. Methods 526 elderly people taking physical examinations were recruited in a community health service center in Guangzhou by convenience sampling from August 2020 to July 2021. They were divided into a modeling group (368 cases) and a validation group(158 cases). Data were collected by a general information questionnaire and cognitive frailty assessment tools. Logistic regression was used to determine the influencing factors,and R software was used to establish a nomogram model for predicting the risk of cognitive frailty. Bootstrap method was used for internal validation of the model,and the validation group was used for external validation. C statistic and calibration curve were used to evaluate the prediction performance of the model. Results The model variables included IADL,self-rated health,daytime mental state,the number of chronic diseases,age,nutritional status and physical exercise. The AUROC of the model was 0.920(95%CI:0.892~0.947),the best cutoff value was 0.401;the sensitivity was 79.7%;the specificity was 89.1%;The C statistics of internal and external validation were 0.910 (95%CI:0.863~0.936) and 0.850(95%CI:0.785~0.915),respectively;calibration curve and Brier score showed good fit. Conclusion The prediction model has a good degree of discrimination and calibration,which can intuitively and easily screen the elderly at high risk of cognitive frailty in the community,and provide references for early screening and intervention.