Development and Validation of a Next-Generation Time-Driven Activity-Based Costing Model for Quantifying Uncertainty: An Empirical Study of Fuzzy-Logic Time-Driven Activity-Based Costing Using Surgical System Data

Background/Objectives: Under fixed-fee reimbursement such as Japan’s Diagnosis Procedure Combination/Per Diem Payment System (DPC/PDPS), hospitals bear the cost of operations that run longer than planned, yet conventional time-driven activity-based costing (TDABC) reports a single cost per case. This study developed and validated a fuzzy-logic TDABC (FL-TDABC) model that expresses case-level time uncertainty as interpretable intervals. Methods: Data from 4674 gastrointestinal surgeries at a Japanese acute-care hospital (fiscal years 2020–2024) were analyzed. Four pre-operative attributes (surgical index rank, technical grade, number of anesthesiologists, and urgency) define 96 possible rules; for each rule, triangular fuzzy numbers for surgical, anesthesia, and room occupancy times were built from the observed minimum, mode, and maximum and defuzzified by the centroid method. Estimates were validated against realized costs in a temporal hold-out (rules from FY2020–2022 applied to FY2023–2024) and 5-fold cross-validation, and 21 alternative specifications were tested. Results: The FL-TDABC centroid exceeded realized cost by ¥36,080 per case (10.2%; Cohen’s d = 0.558). The premium arose entirely from the maximum-time vertex, vanished under percentile bounds, and did not improve point accuracy out of sample. The intervals were informative: the observed range of each rule covered the recorded room occupancy time of 97.5% of hold-out cases and the α = 0.5 cut 76%, interval width increased with surgical complexity, and planning room time at the α = 0.5 upper bound reduced overruns from 44% to 10% of cases. Conclusions: The value of FL-TDABC lies in interval information that can inform scheduling buffers, staffing, and risk assessment, not in a higher or more accurate point estimate.