Comparison of cox models in detecting factors affecting healing rate of dengue hemorrhagic fever

Abstract

Dengue Hemorrhagic Fever (DHF) is an epidemic of disease that usually becomes a benchmark of good or bad condition of the environment and health facilities of a region. When the disease is outbreak and does not get serious treatment, it causes death. One of the first steps that can be done to deal with this disease is to know the factors that affect the healing rate of DHF patients. This study aims to detect factors affecting the healing rate of DHF patients by using Cox regression analysis based on Cox model family, such as Cox Proportional Hazard (PH) model, extended Cox model with one and two heaviside function, and stratified Cox model and get best model for this study. The data used is secondary data consisting of 107 in-patients DHF period January-December 2016 at Regional General Hospital dr. Zainoel Abidin Banda Aceh, Indonesia. Based on Cox Proportional Hazard model, the variable that significantly affect the patient’s healing rate is Age. Based on the Extended Cox model with one heaviside function, the variables that significantly affect the patient’s healing rate are Age, Sex, Number of Platelets, Clinical Degrees III and Number of Leukocytes multiplied by the time function. Based on the Extended Cox model with two heaviside functions, the variables that significantly affect the patient’s healing rate are Age, Sex, Number of Platelets, Clinical Degrees III and Number of Leucocytes multiplied by second time function. Based on the stratified Cox model, there are no variables that significantly affect the patient’s healing rate. The best model based on Akaike Information Criterion value is the Stratified Cox model.

Effects of winter climate on high speed passenger trains in Botnia-Atlantica region

“…A likelihood ratio test or Schoenfeld residuals can be used to examine whether a covariate satisfies the assumption or not [1]. If not, a heaviside function g(t) can be chosen to fix the proportionality violation [14]. The purpose of a heaviside function is to partition the observational time into intervals and make the proportionality assumption be valid within each interval.…”