Multilevel joint competing risk models

Joint modeling approaches are often encountered for different outcomes of competing risk time to event and count in many biomedical and epidemiology studies in the presence of cluster effect. Hospital length of stay (LOS) has been the widely used outcome measure in hospital utilization due to the be...

Full description

Saved in:
Bibliographic Details
Published in:Journal of physics. Conference series 2017-09, Vol.890 (1), p.12132
Main Authors: Karunarathna, G H S, Sooriyarachchi, M R
Format: Article
Language:English
Subjects:
Clusters
Dengue fever
Epidemiology
Logit models
Maximum likelihood estimates
Modelling
Multilevel
Parameter estimation
Physics
Regression models
Risk
Citations: Items that this one cites
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Joint modeling approaches are often encountered for different outcomes of competing risk time to event and count in many biomedical and epidemiology studies in the presence of cluster effect. Hospital length of stay (LOS) has been the widely used outcome measure in hospital utilization due to the benchmark measurement for measuring multiple terminations such as discharge, transferred, dead and patients who have not completed the event of interest at the follow up period (censored) during hospitalizations. Competing risk models provide a method of addressing such multiple destinations since classical time to event models yield biased results when there are multiple events. In this study, the concept of joint modeling has been applied to the dengue epidemiology in Sri Lanka, 2006-2008 to assess the relationship between different outcomes of LOS and platelet count of dengue patients with the district cluster effect. Two key approaches have been applied to build up the joint scenario. In the first approach, modeling each competing risk separately using the binary logistic model, treating all other events as censored under the multilevel discrete time to event model, while the platelet counts are assumed to follow a lognormal regression model. The second approach is based on the endogeneity effect in the multilevel competing risks and count model. Model parameters were estimated using maximum likelihood based on the Laplace approximation. Moreover, the study reveals that joint modeling approach yield more precise results compared to fitting two separate univariate models, in terms of AIC (Akaike Information Criterion).
ISSN:1742-6588
1742-6596
DOI:10.1088/1742-6596/890/1/012132