A Machine Learning Approach to Adaptive EWMA Control Charts: Insights From Cardiac Surgery Data

This study explores using a structured quality framework to enhance patient care quality and safety in healthcare environments. Assessing the performance of healthcare services is a complex task due to the wide range of health conditions among patients, making it difficult to establish consistent evaluation metrics. The study employs a support vector machine (SVM) regression model to address this issue, which helps analyze and manage patient risk factors effectively. To enhance the monitoring of healthcare service quality, this study introduces a new SVM‐based adaptive risk‐adjusted exponentially weighted moving average (SVM‐ARAEWMA) control chart. This chart is designed using the residuals generated by the SVM regression model, allowing for a more accurate assessment of variations in patient outcomes. This study demonstrates how SVM effectively assesses patient conditions by applying this methodology to real‐world data from cardiac surgery patients. Compared to traditional risk‐adjusted exponentially weighted moving average (RAEWMA) and adaptive RAEWMA (ARAEWMA) control methods, the SVM‐ARAEWMA chart is more effective in identifying critical changes in healthcare quality and patient outcomes.