A0584
Title: Quantile residual life regression analysis of HIV/AIDS patients in Korea
Authors: Soomin Kim - Yonsei University (Korea, South) [presenting]
Sangwook Kang - Yonsei University (Korea, South)
Abstract: An HIV patient's residual lifetime is a major point of interest for both the patient and their physicians. While existing analyses on patient survival make forecasts based on data collected at the start of the study, residual lifetime analysis allows for a dynamic analysis based on added data collected up to a certain point in time. Since data on patient survival time shows a long-tailed distribution to the right, the median rather than the mean provides a more useful summary statistic of distribution. This study utilizes modeling of the quantile, including the median as a special case. Using data from the HIV/AIDS prospective cohort study in Korea, we propose statistical inference procedures that model the residual lifetime of HIV patients until they develop dyslipidemia. In this model, we model the quantiles of HIV patients remaining lifetime based on longitudinal biomarkers such as CD4 cells count, which is an important biomarker for HIV patients. To increase the computational efficiency in variance estimation, we propose an induced smoothing approach for the non-smooth estimating functions based on a check function. The proposed estimators are shown to have desirable asymptotic properties. Simulation experiments demonstrated that they perform reasonably well under finite samples.
