Construction of the risk predition model of mild cognitive impairment in hospitalized elder patients

Abstract

Objective To explore the influencing factors of mild cognitive impairment (MCI) in hospitalized elderly people, construct and compare the relative risk prediction model of multi-group MCI.Methods A convenient sampling method was used to select the elderly hospitalized in the geriatrics department of the First Affiliated Hospital of Xinjiang Medical University from January 2023 to September 2023. The multivariate Logistic regression, decision tree and neural network were constructed, and the influencing factors of MCI were analyzed. The area under curve (AUC) of receiver operating characteristic (ROC) was adopted to compared the performance of three sets of prediction models.Results A total of 992 hospitalized elderly patients were included, and the detection rate of MCI was 21.17%. The analysis results of multivariate Logistic regression model, decision tree model and neural network model all showed that age, cerebrovascular disease and education level were the main influencing factors of MCI, and the multiple Logistic regression model and neural network model also showed that the score of daily living ability below 60 was also the influencing factors of MCI. The prediction accuracy of multivariate Logistic regression prediction model was 89.1%, the AUC of ROC curve was 0.933[95%CI(0.916, 0.950)], the sensitivity was 0.881, the specificity was 0.852, and the Yoden index was 0.733. The prediction accuracy of decision tree prediction model was 86.1%, AUC was 0.908[95%CI(0.888, 0.927)], the sensitivity was 0.919, the specificity was 0.753, and the Yoden index was 0.672. The prediction accuracy of the neural network model was 88.7%, the AUC was 0.933[95%CI(0.915, 0.950)], the sensitivity was 0.876, the specificity was 0.861, and the Yoden index was 0.737. The prediction results of the three groups of models were more than 70%, and the prediction efficiency were good.Conclusion Increasing age, shorter years of education, cerebrovascular disease, and decreased ability to perform daily living increase the risk of MCI in older adults. Multivariate Logistic regression, decision tree and neural network models can fully explore the influencing factors of MCI from different levels, and the effective combination of multiple models can fully understand the interaction between different factors, providing references for early screening and intervention of MCI.

Keywords

• mild cognitive impairment
• multivariate logistic regression model
• decision tree model
• neural network model
• prediction model
• elderly