Modeling Nurse Performance and Error Risk With Neural Networks: The Moderating Role of Decision‐Making Styles

Human error in emergency departments (EDs) arises from contextual human factors and individual cognitive differences, yet these streams are often studied separately. This study models ED nurse performance and error risk using artificial neural networks (ANNs), with decision‐making styles (DMSs) as a moderating variable. Twenty factors contributing to human error (FCHE) were identified and refined by ED experts. Data from 239 emergency nurses in three Iranian military hospitals over 1 year included incident‐based FCHE scores, a 40‐item performance instrument, and Driver et al.‘s DMS questionnaire. Reliability and validity were confirmed, and aspect weights were derived using the Best–Worst Method. Multilayer perceptron (MLP) and radial basis function networks were trained with leakage‐free fivefold cross‐validation. The MLP achieved an MSE of 0.0584 on a held‐out 15% test set, outperforming regression baselines. Efficiency scores were validated against DEA and sensitivity analysis ranked influential factors by DMS, supporting targeted staffing and decision‐support design.