基于神经网络混合效应模型的循证2型糖尿病监测研究

慢病控制管理现状不理想，其发病增长快，疾病负担和死因构成逐年增加，严重影响患者生活质量并造成社会及家庭负担。国内慢病的现况调查和健康管理研究较多，但针对糖尿病的监测指标随访分析始终是研究难点且一般采用传统线性混合模型或生存分析方法。这些方法假定指标的线性趋势且数据分布满足特定要求或对基线风险有先验假设，但实际资料常不符合假定而并不适用。因此，探索糖尿病血糖监测随访分析方法是糖尿病健康管理研究中亟待解决的问题。本项目以2型糖尿病患者为研究对象，采用流行病与卫生统计学方法、循证医学和数据挖掘技术，针对复杂的非线性关系、随机效应和交互效应，探索基于神经网络混合效应模型的随访预测建模方法，提高血糖监测随访质量，以利开发合理的监测规则与健康管理策略。本研究为准确预测2型糖尿病随访控糖水平及其并发症提供科学依据、分析方法与循证实践参考，对提高患者生存质量、减轻疾病负担、健康管理等方面具有重要现实意义。

The current situation on control and management of chronic diseases is serious. Its incidence is increasing rapidly with the disease burden and cause of death increasing year by year, which seriously affects the quality of life of patients and causes social and family burden. The health management research on chronic diseases is hot with many available cross-sectional investigations in China. However, the follow-up analysis on monitoring indicators such as blood glucose has always been the research challenge, and it generally used the traditional linear mixed model or survival analysis methods. These methods should assume the linear trend of the data and data distribution meeting specific requirements or some presuppositions of the baseline risks. However, the conditions of the data in practice often do not comply with the assumptions. Therefore, it is a problem to be solved to explore the follow-up analysis methods on monitoring indicators in the study of health management of diabetes. This project is to take patients with type 2 diabetes as the participants, by using the methods of epidemiology and health statistics, evidence-based medicine and the data mining technology, focusing on complex nonlinear relationship, random effect and interaction effect, to explore follow-up monitoring modeling methods on type 2 diabetes based on neural network mixed effect model, which aims to improve the monitoring and follow-up quality of type 2 diabetes, and develop reasonable rules of monitoring and health management strategy. The project is expected to provide a scientific basis, analysis methods and evidence-based practice for complications and monitoring of type 2 diabetes, which has important practical significance on improvement of the survival quality among patients, reduction of the burden of diseases, health management and control, etc.