Abstract:
This paper focuses on exploring methods and analyzing the survival pattern of clustered dengue data
reported from high risk areas in Sri Lanka during the period 2006 to 2009. Due to dengue cases being
clustered within districts resulting in cluster correlation, the response of survival was modeled in a
multilevel framework. As the data consists of several missing values this paper further investigates
multilevel multiple imputation as a method to handle the partially observed dengue dataset appropriately.
A Discrete Time Hazard Model via standard logistic model has been suggested to model the survival of
dengue patients. Results indicate that there is an impact from the clustering variable, district and from
different types of dengue infections, place treated initially, Packed Cell Volume and White Blood Cell
count on the response of interest.
