Identify hospitalization cost drivers of traumatic fracture patients in China using quantile regression and backpropagation neural network

Abstract

Abstract Objective Analyze the factors associated with hospitalization costs of traumatic fracture patients.Methods Data for the retrospective analysis was extracted from the first pages of inpatient medical records in Zhuhai, China. The sample consisted of 31503 patients hospitalized for traumatic fractures between January 1, 2018 and December 31, 2020. We first compared differences in hospitalization costs between subgroups, followed by quantile regression and backpropagation neural network to investigate the key drivers of the hospitalization costs.Results The median hospitalization cost for traumatic fracture patients was ¥13528.2. The mean length of stay was 13.77 days. Quantile regression showed that higher hospitalization costs from the Quantile 0.1 to the Quantile 0.9 significantly correlated with advanced age, more severe types of fracture, operation, comorbidity, longer length of stay, higher level of hospital, and payment with Medicare. Backpropagation neural network indicated that the length of stay, operation level and hospital level were the most important predictors of hospitalization costs.Conclusion Quantile regression and backpropagation neural network yielded valuable information on the factors affecting the hospitalization costs of traumatic fractures in China. Findings suggested that interventions aiming to reduce length of stay contributed to reducing the economic burden associated with traumatic fractures.