Abstract: Objective To explore the application of the growth mixed model of identifying the latent classes of the elderly people with mild cognitive impairment and identifying the cognitive trajectory of different classes. Methods The growth mixed model (GMM) was constructed by using the follow-up data of the elderly with mild cognitive impairment over the age of 65. Bayesian information criterion(BIC) and the posterior probability were used to determine the number of latent classes and to conduct model evaluation, respectively. Results The elderly with mild cognitive impairment presented three different cognitive trajectories.The initial cognitive function of class 1 was low and decreased rapidly in the short time,and then rebounded in the middle, finally descended to a certain degree and tended to be slow, which was called the "high risk population of dementia", accounted for 4.46% of the total population. Class 2 had a lower cognitive function and a sharp decline in cognitive ability with age, which was called "low risk population of dementia", accounted for 33.28% of the population. Class 3 had higher cognitive function, and the cognitive function decreased slowly with age, which was called "normal aging population", accounted for 62.26% of the total population. Conclusions The study reveals the heterogeneity of the cognitive development trajectory in the elderly population with mild cognitive impairment, which helps health promotion personnel develop interventions for high-risk groups early to reduce the incidence of Alzheimer's disease.