Device-procedure attribution of reported harm in robotic-assisted bronchoscopy: a MAUDE study informing safety surveillance of robotic surgical systems.

Postmarket surveillance of robotic surgical systems requires distinguishing harm associated with platform performance from harm arising through procedural, anatomical, anesthetic, or perioperative pathways. We evaluated narrative-supported attribution in US Food and Drug Administration Manufacturer and User Facility Device Experience (MAUDE) reports involving the Ion and Monarch platforms for robotic-assisted bronchoscopy, a navigation-dependent minimally invasive thoracic intervention. Death- and injury-designated reports received through 31 May 2026 were reviewed using a prespecified framework. Two independent reviewers assessed whether patient harm was documented and classified eligible reports as device-related, procedure-related, mixed device-procedure, or indeterminate; disagreements were adjudicated by a third reviewer. Among 1,691 reports, 1,638 (96.9%) contained narrative evidence of harm. Among harm-eligible reports, narrative-supported attribution was procedure-related in 1,429 (87.2%), device-related in 27 (1.6%), mixed device-procedure in 31 (1.9%), and indeterminate in 151 (9.2%). Harm-eligibility agreement was 99.7% (κ = 0.952), whereas attribution agreement among reports judged eligible by both reviewers was 87.8% (κ = 0.587). Indeterminate attribution was more frequent among death- than injury-designated reports (34.5% vs. 7.4%); no death-designated report met the prespecified criteria for device-related attribution. Submitted narratives therefore more often supported procedural or perioperative mechanisms than explicit malfunction-to-harm pathways, but this finding neither establishes operator error nor excludes undocumented device contribution. Mechanism-based attribution may improve triage for engineering investigation, procedural quality review, and evidence preservation across robotic surgical systems.