A1558
Title: A Bayesian hierarchical hidden Markov model for infectious diseases time series
Authors: Geoffrey Singini - University of Malawi (Malawi) [presenting]
Samuel Manda - University of Pretoria (South Africa)
Abstract: In infectious disease forecasting, hidden Markov statistical time series models are used to understand the distribution of the observed disease data conditional on the hidden states and the transitions between states. For example, in human immunodeficiency virus (HIV) disease, hidden states (viral latency or activation) could be associated with observed states (unsuppressed or suppressed viral load). Using hidden Markov models could help draw a complete picture of HIV development. However, in many applications, the observations of the disease progression within a subject may be correlated. Thus, estimating the hidden Markov model parameters between and within subjects could improve the model's capabilities. Moreover, the observed disease data could have been collected from different sources, necessitating a joint model of multiple data streams. A Bayesian hidden Markov model that incorporates subject-level and data source heterogeneity is developed. The proposed methodology is demonstrated with extensive simulation studies and an application to HIV time series data from multiple data sources in Malawi.
