Abstract

Despite a critical shortage of transplantable organs, recovered organs may still go unused after repeated center-level declines, logistical delays, and uncertainty during the allocation process. Although organ quality is an important factor, non-use also reflects an information-systems problem: transplant centers make time-sensitive acceptance decisions under clinical uncertainty, operational constraints, and limited visibility into which centers are most likely to accept a difficult-to-place organ. Prior work on hard-to-place kidney offers shows that donor-level and system-level factors, including cold ischemia time and offer timing, are associated with discard risk (Narvaez et al., 2018). This TREO proposes a predictive decision-support framework for reducing organ non-use in transplant allocation systems. The project reframes non-use from a static discard outcome into a time-sensitive donor-center routing problem. The empirical foundation will draw on OPTN/UNOS STAR files and, subject to data availability and access approval, offer-level donor-center matching and acceptance information. STAR files provide patient-level information on transplant recipients, deceased and living donors, and waiting-list candidates, and require a data use agreement (Organ Procurement and Transplantation Network, n.d.). The proposed system estimates an acceptance confidence score for each donor-center pair by integrating donor and organ-quality indicators, logistical constraints, historical center behavior, and allocation timing. Methodologically, the project will compare interpretable logistic regression with XGBoost-based prediction and SHAP explainability. If sequential offer-level data are available, recurrent neural networks may be evaluated as an extension for modeling time-dependent offer trajectories. A threshold-based decision simulation will assess whether predicted acceptance scores can reduce low-probability offer cycles while preserving clinical judgment, policy constraints, and patient-level priorities. The resulting IT artifact is a donor-center acceptance scorecard that presents acceptance likelihood, key explanatory factors, and routing-relevant risk signals for time-sensitive allocation decisions. This work contributes to MIS research by showing how predictive analytics can reduce information friction in high-stakes healthcare allocation systems. Practically, the framework may support faster, more transparent routing decisions, reduce avoidable delays, and improve organ utilization without replacing clinical expertise.

Download

Share

COinS