Optimizing workflow of urgent stroke endovascular intervention: A focused lean six sigma project

Introduction: Urgent endovascular intervention is currently accepted as the primary and critical therapeutic approach to patients whose acute ischemic stroke results from a large arterial occlusion (LAO). In this context, one of the quality metrics most widely applied to the assessment of emergency systems performance is the "doorto-puncture" (D-P) time. We undertook a project to identify the subinterval of the D-P metric causing the most impact on workflow delays and created a narrowly focused project on improving such subinterval. Methods: Using the DMAIC (i.e., define, measure, analyze, improve and control) approach, we retrospectively reviewed our quality stroke data for calendar year (CY) 2021 (i.e., baseline population), completed a statistical process control assessment, defined the various subintervals of the D-P interval, and completed a Pareto analysis of their duration and their proportional contribution to the D-P interval. We retooled our workflow based on these analyses and analyzed the data resulting from its implementation between May and December 2022 (i.e., outcome population). Results: The baseline population included 87 patients (44 men; mean age = 67.2 years). Their D-P process was uncontrolled, and times varied between 35-235 minutes (Mean = 97; SD = 38.40). Their door to angiography arrival (D-AA) subinterval was significantly slower than their arrival to puncture (AA-P) (73.4 v. 23.5 minutes; p < 0.01), accounted for 73% of the average length of the D-P interval. The group page activation to angiography arrival (GP-AA) subinterval accounted for 41.5% of the entire D-AA duration, making it the target of our project. The outcome population originally consisted of 38 patients (15 men; mean age = 70.3 years). Their D-P process was controlled, its times varying between 43-177 minutes (Mean = 85.8; SD = 34.46), but not significantly difference than the baseline population (p = 0.127). Their target subinterval GP-AA varied between 0-37 minutes and was significantly improved from the baseline population (Mean = 13.21 v. 29.68; p < 0.001). Conclusions: It seems feasible and reasonable to analyze the subinterval components of complex quality metrics such as the D-P time and carry out more focused quality improvement projects. Care must be exercised when interpreting the impact on overall system performance, due to unexpected variations within interdependent subprocesses. The application of a robust and comprehensive LSS continuous quality improvement process in any CSC will have to include individualized focused projects that simultaneously control the different components of overall system performance.