RESUMO
This paper was designed to determine if the Mahalanobis-Taguchi System (MTS) applied to the delirium-evidence-based bundle could detect medical patterns in retrospective datasets. The methodology defined the evidence-based bundle as a multidimensional system that conformed to a parameter diagram. The Mahalanobis distance (MD) was calculated for the retrospective healthy observations and the retrospective unhealthy observations. Signal-to-noise ratios were calculated to determine the relative strength of detection of 23 delirium preindicators. This study discovered that the sufficient variation in the CAM-ICU assessment, the standard for delirium assessment, would benefit from knowledge of how different the MD is from the healthy average. The sensitivity of the detection system was 0.89 with a 95% confidence interval of between 0.84 and 0.92. The specificity of the detection system was 0.93 with a 95% confidence interval between 0.90 and 0.95. The MTS applied to the delirium-evidence-based bundle could detect medical patterns in retrospective datasets. The implication of this paper to a biomedical research is an automated decision support tool for the delirium-evidence-based bundle providing an early detection capability needed today.