Cognitive impairment is a common disease among middle-aged and elderly people, which seriously affects health outcomes and quality of life, and carries a risk of progressing to severe stages such as dementia. Early identification is beneficial for timely intervention and treatment. This study proposes a new model for predicting cognitive impairment that integrates static and dynamic data, including medical, demographic, and social relationship features. It combines Granger causality with deep learning and uses multiple metrics to evaluate model performance. The performance comparison results between our model and the baseline model demonstrate that our model's predictions have a certain level of accuracy. In addition, causal features derived from Granger causality analysis are used to identify cognitive impairments. Statistical analysis shows that the selected features have statistical significance, further verifying the robustness of our model and its potential for predicting cognitive impairment.