A Validation of New Methods for Estimating Healthcare Costs Adjusted for End-of-Life Cost and Censoring

OBJECTIVES: Much of the data utilized in cost studies is subject to censoring due to short study periods and patient deaths. Estimating mean total costs using traditional statistical methods without considering censoring can bias long-term healthcare cost estimates. This study aims to propose a method to adjust for both censoring and end-of-life costs when estimating cumulative costs and validate this method using simulation.

METHODS: We generated ideal artificial data via simulation, where all patients were followed for five years. For the cost data generation, the following parameters and distributions were considered: population censoring rate of 10%, mortality rate of 4%, healthcare costs following a normal distribution (mean 100, standard deviation 10), and end-of-life healthcare costs following a normal distribution (mean 600, standard deviation 30). After adjusting the simulated data by removing healthcare costs incurred within a specific period before death, we applied four different censoring methods: Available-sample (AS) estimator, Complete Case (CC) estimator, Bang and Tsiatis (BT) method, and Zhao and Tian (ZT) method. We compared the estimated 5-year cumulative healthcare costs with and without end-of-life cost adjustments.

RESULTS: The cumulative cost, after adjusting for end-of-life costs, decreased compared to the original unadjusted data. The ZT method estimated the highest cumulative cost, followed by BT, AS, and CC when end-of-life costs were not adjusted. The ranking of the adjusted cumulative costs was similar, but the differences between BT and ZT, as well as AS and CC, decreased.

CONCLUSIONS: Additional adjustments for end-of-life costs, based on the pattern of cost increases before death, should be considered to determine the most accurate method for estimating cumulative healthcare costs. Furthermore, further analyses based on empirical data should be carried out to evaluate the practical validity of the proposed method. Adjusting for end-of-life costs in addition to censoring is expected to enhance the accuracy of healthcare cost estimates.