Cardiopulmonary bypass (CPB) oxygenators provide essential intra-operative gas exchange. Although rare, CPB oxygenator failures can cause catastrophic morbidity and mortality, and published analyses of CPB-specific oxygenator failures are limited. We aimed to study reports of CPB oxygenator failures to observe patterns to inform preparation and device improvement.

We retrospectively reviewed Manufacturer and User Facility Device Experience (MAUDE) reports (April 2015–April 2025) of CPB oxygenator failures resulting in injury or death. ECMO events, malfunctions without harm, and duplicates were excluded. Manufacturer, device problem, and patient problem were compared between injury and death cohorts using chi-square or Fisher’s exact tests. Narrative reports underwent supervised natural language processing (NLP) to identify clinical themes.

A total of 388 events met criteria: 279 injuries (71.9%) and 109 deaths (28.1%). Events were commonly related to six manufacturers. Manufacturer B and Manufacturer D devices were disproportionately represented in deaths (43.1% vs 11.1% of injuries, p < 0.001 and 12.84% vs 5.38% of injuries, p = 0.017), while Manufacturer A was more frequently linked to injuries (60.6% vs 22.9%, p < 0.001), although the incidence of failure for each manufacturer could not be calculated. The remaining manufacturers were evenly distributed among report types. Device problem categories differed by outcome: “Unexpected Device Operation” was significantly more common in deaths (7.3% vs 1.4%, p = 0.0054), whereas “Human Factors” were more common in injuries (11.8% vs 4.6%, p = 0.036). NLP identified intra-operative performance/monitoring failures, hemolysis and thromboembolic events, and post-procedural evaluation/quality management as common descriptions of device failure situations.

Using MAUDE reports, CPB oxygenator failures causing harm show distinct manufacturer, device problem, and outcome patterns but incidence could not be determined. Reports of unexpected operational failures and thromboembolic complications are more often fatal. NLP enhances detection of clinically relevant patterns, supporting targeted safety strategies and device improvements.