Importance: In China, the societal burden of cognitive impairments continues to increase as the country ages, but our knowledge remains limited regarding how to accurately predict future cognitive impairment at the individual level for preventative interventions. Identifying the most predictive risk factors and socioeconomic groups where predictions are less accurate would provide a foundation for developing targeted prediction models that can identify elderly at high risks of future cognitive impairments. Objectives: To quantify how well demographics, instrumental activities of daily living, activities of daily living, cognitive tests, social factors, psychological factors, diet, exercise and sleep, chronic diseases, and three recently published prediction models predict future cognitive impairments in the general Chinese population and among male, female, rural, urban, educated, and uneducated elderly. Design: The Chinese Longitudinal Healthy Longevity Survey (CLHLS) is a prospective cohort study of elderly Chinese from 23 provinces. Individual information from the 2011 CLHLS survey was used to predict if participants would become cognitively impaired by follow-up in 2014. Setting: Population-based. Participants: 4047 CLHLS participants 60 years of age or older without cognitive impairments at baseline were included. Main Outcome: Cognitive impairment was identified through the Chinese language version of the Mini Mental State Examination (MMSE). Predictive ability was quantified using the AUC, sensitivity, and specificity across 20 repeats of 10-fold cross validation where the target variable was an indicator of cognitive impairment 3 years from the baseline survey. Results: A total of 337 (8.3%) elderly Chinese became cognitively impaired by the follow up survey. The risk factor groups with the most predictive ability in the general population were demographics (AUC, 0.78, 95% CI, 0.77-0.78), cognitive tests (AUC, 0.72, 95% CI, 0.72-0.73), and instrumental activities of daily living (AUC, 0.71, 95% CI, 0.70-0.71). Demographics, cognitive tests, instrumental activities of daily living, and all three re-created prediction models had significantly higher AUCs when making predictions among women compared to men and among the uneducated compared to the educated. Dietary factors, which have yet to be included in prediction models in China, had more predictive power (AUC, 0.59, 95% CI, 0.58-0.60) than activities of daily living (AUC, 0.57, 95% CI, 0.56-0.57), psychological factors (AUC, 0.58, 95% CI, 0.57-0.59), and chronic diseases (AUC, 0.53, 95% CI, 0.52-0.53). Conclusion and relevance: This study suggests that demographics, cognitive tests, and instrumental activities of daily living are the most useful risk factors for predicting future cognitive impairment among elderly Chinese. However, the most useful risk factors and existing models have lower predictive power among male, urban, and educated elderly Chinese. More efforts are needed to ensure that equally accurate risk assessments can be conducted across different socioeconomic groups in China.