Nonparametric estimation of medical cost quantiles in the presence of competing terminal events

Medical care costs are commonly used by health policy-makers and decision-maker for evaluating health care service and decision on treatment plans. This type of data is commonly recorded in surveillance systems when inpatient or outpatient care service is provided. In this paper, we formulate medical cost data as a recurrent marker process, which is composed of recurrent events (inpatient or outpatient cares) and repeatedly measured marker measurements (medical charges). We consider nonparametric estimation of the quantiles of cost distribution among survivors in the absence or presence of competing terminal events. Statistical methods are developed for quantile estimation of the cost distribution for the purposes of evaluating cost performance in relation to recurrent events, marker measurements and time to the terminal event for different competing risk groups. The proposed approaches are illustrated by an analysis of data from the Surveillance, Epidemiology, and End Results (SEER) and Medicare linked database.