Neural network-based prediction of consented organs utilization

Abstract

Background: Organ donation is coordinated by organ procurement organizations’ staff, along with hospitals. This procedure requires various steps and intricate coordination between all parties to obtain consent from the family and recover the organs. However, organs are not always recovered despite consent, and these cases are referred to as consented but not recovered. This research aims to predict the likelihood of a consented case with no recovered organ occurrence, which might affect the chances of organ recovery, and thus transplant. This also allows the clinical team to be more focused on cases with more likelihood of recovery. Methods: A two-phase analysis was performed to evaluate assessment methods and to see if the best method was affected by a larger dataset. The first phase used case histories from 752 cases collected from January 2018 to October 2019; the second phase used 1,476 cases from 2016 to 2019. Data included brain death and cardiac death patients for both phases, where 10% are consented and not recovered cases. The variables considered are related to donor characteristics and the approach for donation and referral process data points. Statistical analysis of the main factors in terms of the donation outcome was conducted. Backpropagation, resilient propagation, and globally convergent propagation models were constructed to predict consented but not recovered organs. Results: The data were preprocessed, and various imputation techniques were applied. The globally convergent propagation model yielded slightly better performance for both data sets (RMSE 0.015 and accuracy of 0.80 at 30 hidden neurons). The results indicated that medical reasons ruling out recovery formed the most common factor that prevented donation and the time between death and approach as the least significant. Conclusions: These findings will help devise a better strategy to robustly predict and constitute the first step to reduce consented but not recovered cases through understanding the underlying causes. This informed decision process will contribute in the long run to the increase of organ utilization and donation conversion rates.