Competing risk model for dengue epidemiology in Sri Lanka: modeling length of stay in hospital

Abstract

This paper focuses on exploring methods for analyzing length of stay in hospital and the discharge endpoint for dengue patients reported from high risk areas of Sri Lanka during the period of 2006-2008. Due to the length of stay being related to different endpoints, the response of length of stay is modeled with the concept of competing risk. A popular competing risk model of sub distribution proportional hazard model was fitted to find out the associated factors for different end points of the discharge competing risk model. Moreover this study concludes that the effect of age, ethnicity, dengue classification, district and platelet count are important to the discharge status of the competing risk model.

Keywords

• competing risk model, hazard, length of stay, subdistribution proportional hazards model