Background: Peripherally inserted central catheters (PICCs) have been associated with venous thromboembolism (VTE) and are a major cause of upper extremity deep vein thrombosis (DVT). However, mechanisms to identify patients at greatest risk of PICC-associated VTE are limited.
Methods: Using data from the Michigan Hospital Medicine Safety consortium, patients with PICCs that experienced symptomatic, image-confirmed upper-extremity DVT were identified. A logistic, mixed effect, two-stage model with hospital-specific random intercepts was used to identify factors associated with PICC-DVT. Points were assigned to each predictor, stratifying patients into four classes for risk of PICC-DVT. Validation was performed by internal bootstrapping with results expressed as odds ratios (OR) and 95% confidence intervals (CI).
Results: Of 22,056 patients that received PICCs, 478 (2.2%) developed symptomatic PICC-DVT were included in the analysis. Risk factors associated with PICC-DVT following two-stage modeling included: history of DVT; use of a multi-lumen PICC; active cancer; presence of another CVC when the PICC was placed; and a white blood cell count greater than 12,000 at the time of PICC insertion (Table 1). Thrombosis rates were 0.8% for class I, 1.7% for class II, 2.9% for class III and 6.9% for class IV. The risk classification rule was significantly associated with VTE risk (p<0.0001), with ORs of PICC-DVT of 1.83 (95 % CI: 1.30, 2.57), 3.30 (95 % CI: 2.43, 4.49) and 8.27 (95 % CI: 5.69, 12.01) for risk classes II, III and IV vs. risk class I, respectively (Table 2). The area under the receiver-operating-characteristics curve was 0.72, with an estimated optimism of 0.02 from the bootstrap internal validation data, suggesting excellent performance.
Conclusions: The Michigan Risk Score offers a novel way to predict and categorize risk of PICC-VTE and advances the field in several ways. First, it can help identify patients at high-risk of thrombosis when considering PICC placement. Second, the tool may help inform surveillance or testing among those at high-risk in the setting of vague symptoms. Finally, the risk score could also help inform duration of anticoagulation in patients with PICC-associated thrombosis, with consideration of extended courses in those at highest risk of this event. External validation of this rule with studies that explore implementation in real-world practice are needed.