Back to 2026 Abstracts
Analysis Of Perioperative Workflows During Lower Extremity Vascular Procedures: A Multisite Analysis Using Ambient Computer Vision
Alan B. Lumsden, Rose M. McCullough, Theoren Loo, Dora Z. Zatyko, Nathaniel Hilger.
Houston Methodist, Houston, TX, USA.
Background:Workflow variation contributes to differences in efficiency, outcomes, and cost. Despite growing interest in standardization, temporal data about perioperative phases are rarely captured systematically. We used an ambient computer vision system to quantify time spent in perioperative workflows across six hospital sites performing high-volume lower extremity interventions.
Methods: We retrospectively analyzed cases from six sites in one academic health system (Sept 2022-Aug 2025). High-resolution wall-mounted cameras captured surgical progression, and machine learning algorithms passively timestamped key perioperative phases: anesthesia induction, patient preparation, final preparation, active procedure, post-operation, patient exit, room cleanup, and setup. For each phase, medians and interquartile ranges were calculated by site, with durations normalized for cross-site comparison. Kruskal-Wallis tests assessed between-site differences; site-segment combinations with <10 observations were excluded.
Results: All perioperative phases showed statistically significant variation across sites (Table, Figure). Patient preparation at Site C was substantially longer, with a median of 20.0 minutes versus 9.0-10.0 minutes elsewhere. Final preparation and active procedure times also varied, with Site F consistently shorter. Active procedure times ranged from 31.0 minutes at Site F to 63.0 at Site C. Turnover phases revealed further differences: room cleanup was longest at Site A (25.0 minutes) and shortest at Site D (10.0), while setup ranged from 19.0 to over 33.0 minutes.
Conclusions:Perioperative workflows vary substantially across hospitals, even within the same health system. We focused on lower extremity interventions as common, standardizable procedures. Variability stems from staffing, protocols, execution, and patient factors, while documentation remains largely manual and retrospective. We previously validated ambient computer vision for basic measures (patient in/out, draping); this study extends it to a standardized, multi-hospital system with expanded data points for actionable insights. Identifying time loss enables recognition of performance gaps and data-driven benchmarks supporting standardization and targeted improvement. Ambient computer vision offers a scalable, objective tool for surgical workflow analysis and OR efficiency optimization, with surgeon involvement essential to ensure metrics reflect meaningful outcomes and quality of life.
Back to 2026 Abstracts
