Objective: To develop a prediction model for mild cognitive impairment (MCI) risk among the elderly, so as to provide a tool for MCI early screening. Methods : From July 2022 to September 2024, a multi-stage stratified random cluster sampling method was used to recruit permanent residents aged ≥65 years from the Xinjiang Uygur Autonomous Region as study participants. Data on sociodemographic characteristics, nutritional status, body composition indices, bone mineral density, and handgrip strength were collected through questionnaires and physical examinations. Sarcopenia was defined based on appendicular skeletal muscle index and handgrip strength. MCI was assessed using the Mini-Mental State Examination, with adjustments for educational level. Participants were randomly divided into a training set and a validation set in a 7∶3 ratio. LASSO regression and multivariable logistic regression models were employed to screen for predictors and construct an MCI risk prediction model. The predictive performance of the model was evaluated using receiver operating characteristic (ROC) curve and decision curve analysis (DCA). Results: A total of 1 641 participants were surveyed, including 755 males (46.01%) and 886 females (53.99%). The majority of participants were aged 65-<75 years, comprising 1 154 individuals (70.32%). MCI was detected in 517 participants, corresponding to a detection rate of 31.51%. Resultsfrom LASSO regression and multivariate logistic regression analysis showed that residence (rural, OR = 2.323, 95% CI: 1.682-3.210), age (75-<85 years, OR = 1.405, 95% CI: 1.019-1.937; ≥85 years, OR = 3.655, 95% CI: 1.696-7.875), educational level (primary school, OR = 0.341, 95% CI: 0.247-0.472; junior high school, OR = 0.255, 95% CI: 0.160-0.408; high school, OR = 0.286, 95% CI: 0.154-0.531; bachelor's degree or above, OR = 0.120, 95% CI: 0.041-0.351), history of alcohol consumption (yes, OR = 3.216, 95% CI: 2.164-4.779), risk of malnutrition (yes, OR = 1.464, 95% CI: 1.064-2.014), sarcopenia (yes, OR = 3.197, 95% CI: 2.332-4.385), and waist-to-hip ratio (abnormal, OR = 1.540, 95% CI: 1.159-2.048) were identified as predictive factors for MCI among the elderly. In the training set, the area under the ROC curve, sensitivity, and specificity were 0.788, 0.719, and 0.712, respectively. In the validation set, the corresponding values were 0.784, 0.913, and 0.542, respectively. DCA demonstrated that the model provided a higher clinical net benefit for predicting MCI risk when the risk threshold probability ranged from 0.124 to 0.764. Conclusion The prediction model developed in this study demonstrates good discriminative ability and clinical utility, indicating its substantial value for predicting the MCI risk among the elderly.