Incidence and risk factors for central venous access device failure in hospitalized adults: A multivariable analysis of 1892 catheters

Background: Central venous access devices (CVADs) allow intravenous therapy, haemodynamic monitoring and blood sampling but many fail before therapy completion. Objective: To quantify CVAD failure and complications; and identify risk factors. Designs, Settings and Participants: Secondary analysis of multicentre randomised controlled trial including patients aged ≥16 years with a non-tunnelled CVAD (NTCVAD), peripherally-inserted central catheter (PICC) or tunnelled CVAD (TCVAD). Primary outcome was incidence of all-cause CVAD failure (central line-associated bloodstream infection [CLABSI], occlusion, accidental dislodgement, catheter fracture, thrombosis, pain). Secondary outcomes were CLABSI, occlusion and dislodgement. Cox regression was used to report time-to-event associations. Results: In 1892 CVADs, all-cause failure occurred in 10.2% of devices: 49 NTCVADs (6.1%); 100 PICCs (13.2%); 44 TCVADs (13.4%). Failure rates for CLABSI, occlusion and dislodgement were 5.3%, 1.8%, and 1.7%, respectively. Independent CLABSI predictors were blood product administration through PICCs (hazard ratio (HR) 2.62, 95% confidence interval (CI) 1.24–5.55); and in TCVADs, one or two lumens, compared with three to four (HR 3.36, 95%CI 1.68–6.71), intravenous chemotherapy (HR 2.96, 95%CI 1.31–6.68), and diabetes (HR 3.25, 95%CI 1.40–7.57). Independent factors protective for CLABSI include antimicrobial NTCVADs (HR 0.23, 95%CI 0.08–0.63) and lipids in TCVADs (HR 0.32, 95%CI 0.14–0.72). NTCVADs inserted at another hospital (HR 7.06, 95%CI 1.48–33.7) and baseline infection in patients with PICCs (HR 2.72, 95%CI 1.08–6.83) were predictors for dislodgement. No independent occlusion predictors were found. Modifiable risk factors were identified for CVAD failure, which occurred for 1-in-10 catheters. Strict infection prevention measures and improved CVAD securement could reduce CLABSI and dislodgement risk.