Use of a conditional quantiles method to predict future health outcomes based on the trajectory of pediatric end-stage liver disease (PELD) scores

Pediatric patients with advanced liver diseases often require liver transplantation. To maximize the efficiency and effectiveness of this procedure in saving lives, the United Network of Organ Sharing (UNOS) uses the results of the pediatric end-stage liver disease (PELD) scoring system to prioritize the list of pediatric patients who are awaiting a transplant. In the analysis reported here, we used data derived from pediatric patients who had a primary diagnosis of biliary atresia, were awaiting a liver transplant, and had PELD scores reported in the Standard Transplant Analysis and Research (STAR) database of UNOS. We used a conditional distribution quantiles method to predict a patient’s future distribution of 90-day PELD scores on the basis of his or her PELD scores in the past 30 days. Because this method takes into account both the scores and the rate of change of scores, it is able to demonstrate how patients with the same current scores may have different distributions of scores in the future. To examine the quality of our predictions, we compared our estimated distribution of 90-day PELD scores with the observed distribution of 90-day PELD scores. We used the diagnostic plot to assess the overall goodness of fit of the model. Public health significance: Transplantation is an effective treatment for patients with end-stage liver diseases. Use of an efficient and effective method to allocate organs among pediatric candidates remains a major challenge. We proposed a method to more accurately estimate the future health condition using not only the current information but also the path of reaching the current status. If clinicians and policy makers will adopt the method to improve the current organ allocation policy, organs will be used more efficiently and eventually save more lives.\u