Multistate health transition modeling using neural networks
提出一种结合神经网络与广义线性模型的新方法，用于估计和预测健康状态间的转移强度，利用中国长寿调查数据验证其优于单独模型，并给出不同人群的预期寿命估计。
Abstract This article proposes a new model that combines a neural network with a generalized linear model (GLM) to estimate and predict health transition intensities. We introduce neural networks to health transition modeling to incorporate socioeconomic and lifestyle factors and to allow for linear and nonlinear links between these variables. We use transfer learning to link the models for different health transitions and improve the model estimation for health transitions with limited data. We apply the model to individual‐level data from the Chinese Longitudinal Healthy Longevity Survey from 1998 to 2018. The results show that our model performs better in estimation and prediction than standalone GLM and neural network models. We provide new estimates of the life expectancies for a range of population subgroups. We also describe a wide range of possible applications for further health‐related research, including risk prediction using health claim data and mortality prediction based on individual‐level mortality data.