Comparison of four strategies for diagnosing deep vein thrombosis: a cost-effectiveness analysis.

PURPOSE: Four strategies for the diagnosis of deep vein thrombosis have been validated recently. The strategies use various combinations of assessment of a patient's clinical probability of having deep venous thrombosis, serial lower limb venous compression ultrasonography, and measurement of plasma D-dimer levels. We compared the cost-effectiveness of these diagnostic strategies.

MATERIALS AND METHODS: We performed a formal cost-effectiveness analysis using a decision-analysis model. Outcomes considered were costs per patient, 3-month quality-adjusted survival, number of lives saved per 1,000 patients, and incremental costs per quality-adjusted life-year (QALY) gained.

RESULTS: Under baseline conditions, with a 24% prevalence of deep vein thrombosis in tested patients, the effectiveness of all strategies was similar (4.6 to 4.8 lives saved per 1,000 patients managed). The most expensive strategy was serial ultrasound (repeat ultrasound on day 7 in all patients with a normal initial ultrasound) at a cost-effectiveness of $10,716 per additional QALY. Performing a repeat ultrasound only in patients with an elevated D-dimer level (serial ultrasound with D-dimer) was somewhat less expensive at $10,281 per additional QALY. Taking clinical probability into account by repeating ultrasound only in patients with an intermediate clinical probability of deep vein thrombosis (risk-based serial ultrasound) yielded further savings and cost $10,090 per additional QALY. The least expensive and most cost-effective option was to perform D-dimer as the initial test, followed by a single ultrasound if the D-dimer level was abnormal, and by phlebography in patients with a normal ultrasound and a high clinical probability of deep vein thrombosis (D-dimer with risk-based single ultrasound) at $8,897 per additional QALY. This strategy allowed a 17% reduction in incremental costs compared with the most expensive algorithm and reduced resource consumption (70 ultrasound procedures per 100 patients managed vs 130 to 170 with the other diagnostic strategies).

CONCLUSIONS: Combining clinical probability and D-dimer with a single ultrasound in the diagnostic workup of patients with possible deep vein thrombosis is highly cost-effective, allowing a reduction in costs and resource use without any substantial increase in mortality. Serial ultrasonography is less cost-effective.