Abstract:
Dengue is a common mosquito-borne tropical disease caused by a virus. It is a life threatening disease since it sometimes leads to death within a short period of time. Multilevel modeling is a form of statistical modeling when data is at different levels. Due to dengue seriousness and risk being more similar for patients within districts than between districts, there is correlation between patients within districts. Thus district has to be taken as a cluster variable. A frequently encountered response in epidemiological studies is the length of Stay (LOS) of a patient, which measures the time until the event of interest occurs. Complexity arises with the different states/destinations of the time event and competing risk modelling is a more appropriate method for handling such states. The association of platelet count and length of stay of a dengue patient leads to the joint modelling approach for analyzing the dengue patients. Formulation criteria for the joint model with clustered data is to link these models through two sub models that is by using the multilevel multinomial logistic model for the LOS of dengue patients with different destinations and multilevel continuous model for the log platelet count. The linkage between two responses was derived by sharing a common random effect. Factors that have an effect on different destinations of LOS are, time indicators, year, age, classification, rainfall, temperature and humidity, while age, sex, classification, year place treated, rainfall, temp and humidity are associated factors for the log platelet count of dengue patients. Moreover, supremacy of joint modelling was proved by the AIC and BIC values over two separate univariate models.
