Prediction-Informed Prioritization for Admission Decisions in High-Stakes Settings

Amy Ward Professor University of Chicago Abstract: Predictions based on historical information increasingly inform priority-based admission decisions in capacity-constrained systems. However, better predictive accuracy does not necessarily lead to better admission decisions when individuals have heterogeneous resource-use durations and service outcomes. In this paper, we study how prediction errors affect decision quality in high-stakes settings where offline data is available and online exploration is infeasible. We do this in the context of a multiclass service system in which individuals’ classes are unobservable at admission and must be predicted from observed covariates. We first show that a priority score policy solves both the ground-truth and estimated admission control problems. We then develop an evaluation metric, Score@K̂, for prediction-model selection that accounts for the endogenous admission boundary and operational cost consequences of prediction errors. We also develop a ranking-based uncertainty-quantification framework that exploits the priority structure to assess individual decision confidence. Finally, using data from incarceration-diversion programs in seven Illinois counties, we construct a simulation testbed environment to (1) show how our Score@K̂ metric can help with prediction model selection in a realistic setting, and (2) demonstrate the impact of capacity on individual decision confidence. Bio: Amy R. Ward is the Rothman Family Professor of Operations Management at the University of Chicago Booth School of Business. She received her Ph.D. degree from Stanford University in 2001. She is a fellow of the INFORMS Manufacturing and Service Operations Management (M&SOM) Society (elected June, 2023). She is the Editor-in-Chief for the journal Operations Research (term began 1/1/2024), which is a flagship journal for the INFORMS community. She was previously Editor-in-Chief for the journal Operations Research Letters (term 4/1/2021-1/1/2024). Earlier, she held the position of Chair of the INFORMS Applied Probability Society (term 11/2016-11/2018), and was one of the keynote speakers at the 21st INFORMS Applied Probability Society Conference. She was the 2025 Faculty Commencement Speaker at the Chicago Booth Graduation Ceremony for its Evening and Week-end MBA Program, the Sokolov EMBA Program, and the Masters in Management Program Graduates (see here). She has given more than 50 University Seminars, and multiple conference keynote presentations (most recently at the second workshop on Artificial Intelligence and Machine Learning in Operations Management Research at Banff in 2026, see here).