Forecasting New Cases of Bipolar Disorder Using Poisson Hidden Markov Model

Background: Bipolar disorder (BD) is a major public health problem. In time series count data there may be over dispersion, and serial dependency. In such situation some models that can consider the dependency are needed. The purpose current research was to use Poisson hidden Markov model to forecast new monthly BD instances. Methods: In current study the dataset including the frequency of new instances of BD from October 2008 to March 2015 in Hamadan Province, the west of Iran were used. We used Poisson hidden Markov with different number of conditions to determine the best model according to Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC). Then we used final model to forecast for the next 24 months. Results: Poisson hidden Markov with two states were chosen as the final model. Each component of dependent mixture model explained one of the states. The results showed that the new BD cases is increase over time and due to forecasting results number of patients for the next 24 months comforted in state two with mean 85.15. The forecast interval was approximately (56, 100). Conclusion: As the Poisson hidden Markov models was not used to forecast the future states in other prior researches, the findings of this study set forward a forecasting strategy as an alternative to common methods, by considering its deficiencies.

Abstract licence: CC BY-NC

A person counts once a day per paper. We keep no cookie and no address: a reader is a hash that changes every day. Any month under five reads shows no number.