Parkinson’s disease (PD) is the second most common neurodegenerative disease worldwide that mainly affects motor system. Treatment given to PD patients may have further complications effect such as dyskinesias and motor fluctuation. People with PD receiving medication often experience complications. It is interest to identify factors associated with the complications. Data on 215 people with PD obtained from the Parkinson’s Progression Markers Initiative (PPMI) database, retrieved on April, 4^th 2018 were analysed. Total scores for the Movement Disorder Society-Unified Parkinson Disease Rating Scale (MDS-UPDRS) Part 1, 2 , and 3, and Hoehn Yahr scale (HY) as motor aspects, Geriatric Depression Scale (GDS) and Scales for Outcomes in Parkinson's disease-Autonomic (SCOPA-AUT) as non motor aspects were used as the explanatory variables.

 

Poisson regression is commonly used for count data. However, this method requires the data to be equidispersed, which is not always fulfilled in the case of overdispersion. Overdispersion may exist when there is excess zero in the data. A two stage regression model, Zero-Inflated Poisson (ZIP) regression, might solve this problem of overdispersion by identifying the structural zero at the first stage, then model the Poisson counts as the second stage. Thus, in this paper, we proposed ZIP regression to model the frequency of experiencing motor complications in people with PD.

 

Bayesian approach offers several advantages over the frequentist approach, such as accomodation of uncertainty (coefficient parameters of regression) in the model through the prior specification. Therefore, in this paper, the parameters in ZIP regression were estimated using the Bayesian approach. Sampling from the posterior distribution of the parameters of interest is conducted using Monte Carlo Markov Chain-Gibbs Sampling (MCMC-GS). The result shows that total score of MDS-UPDRS Part 2 and HY are negatively associated with people for no need medication, while the opposite is observed for total score of MDS-UPDRS Part 3. Furthermore, in the second stage of the model, we gain insight on factors associated with the frequency of motor complications due to the medication. Total score of Part 3 is negatively associated with frequency of complications, while the opposite trend is observed for HY.

Keywords: Bayesian, Excess Zero, Gibbs Sampling, Parkinson’s Disease, UPDRS, Zero-Inflated Poisson.