Time Series Models for an Exposure-Response Relationship Problem in Mental Healthcare

Abstract

The problem of the exposure-response relationship between environmental factors and mental illness is gaining attraction in recent years. This chapter explores the various time series approaches that can be applied to solve the exposure-response relationship. In the problem of predicting psychiatric facility admissions based on environmental factors, there is a lagged association between the daily concentration of environmental variables and hospital admission, which is non-linear. The Poisson generalized linear regression in conjunction with the distributed lag non-linear model is utilized to explore this non-linear and lagged effect. The various deep learning approaches employed for addressing the exposure-response relationship are discussed in this chapter. The performance of various time series techniques is illustrated with the help of a dataset based in Bangalore City, India.