Validation of the Canadian Clinical Probability ... - Wiley Online Library

13 downloads 11316 Views 183KB Size Report
venous thrombosis was calculated using the Canadian ... Canadian Probability Model for Acute. Deep Vein ..... programmed computer decision-support tool, use.
561

ACAD EMERG MED • June 2002, Vol. 9, No. 6 • www.aemj.org

CLINICAL INVESTIGATIONS Validation of the Canadian Clinical Probability Model for Acute Venous Thrombosis G. Patrick Shields, MD, Sam Turnipseed, MD, Edward A. Panacek, MD, MPH, Norman Melnikoff, MD, Robert Gosselin, MT, Richard H. White, MD Abstract Objective: To validate the predictive value of the Canadian clinical probability model for acute venous thrombosis, which, to the best of the authors’ knowledge, has not been done in emergency department (ED) settings outside of Canada. Methods: Demographic and clinical information, rapid D-dimer testing, and venous ultrasound imaging were obtained among patients presenting with clinically suspected venous thrombosis at a university-affiliated ED. A diagnosis of deep venous thrombosis (DVT) was made based on venous ultrasound test results or objectively documented venous thromboembolism during a 12-week follow-up period. The probability of venous thrombosis was calculated using the Canadian clinical probability model. Results: Among 102 patients, 17 (17%) were diagnosed as having venous thrombosis initially or during the three-month follow-up period. The

frequency of venous thrombosis among patients categorized as having high probability was 10 of 17 [59%, 95% confidence interval (95% CI) = 35% to 82%], 6 of 44 (14%, 95% CI = 4% to 24%) with intermediate probability, and 1 of 41 (2%, 95% CI = 0.1% to 11%) with low probability. This compares with respective values of 49%, 14%, and 3%, reported by Canadian researchers in an ED study. Forty-one of 102 (40%) patients had an alternate diagnosis as likely or more likely than venous thrombosis, but only three (7%, 95% CI = 2% to 18%) of these had venous thrombosis. Conclusions: Use of the Canadian probability model for DVT in this ED resulted in effective risk stratification, comparable to previously published results. Key words: DVT; clinical probability; acute venous thrombosis; validation. ACADEMIC EMERGENCY MEDICINE 2002; 9:561–566.

Recent studies have documented the utility of using clinical probability models as aids in the management of patients with suspected deep venous thrombosis (DVT) or pulmonary embolism.1–3 Canadian researchers have promoted the use of a probability model for acute venous thrombosis based on nine clinical elements (Table 1).2 In a prospective trial, use of this probability model in an emergency department (ED) setting led to successful categorization of a large cohort of patients into high-, intermediate-, and low-probability groups.4 It is not clear, however, whether the predictive value of this model is comparable when used by emergency physicians in other settings. A clinical probability model is useful if it leads to better management decisions or cost savings.5,6 Clinical probability models have been tested in conjunction with the use of rapid D-dimer testing

among patients with suspected DVT7–14 and among patients with suspected pulmonary embolism.8 These initial studies suggested that patients with both low clinical probability of thromboembolism and a negative rapid D-dimer test have a very low three-month incidence of developing objectively documented venous thromboembolism: less than 2%. The present study was designed to validate in an ED setting the predictive value of the Canadian clinical probability model for venous thrombosis. In addition, we investigated the use of rapid D-dimer testing using SimpliRed, a rapid qualitative agglutination test that uses a small sample of whole blood.

From the University of California, Davis (GPS, ST, EAP, NM, RHW), and the UC Davis Clinical Laboratory (RG), Sacramento, CA. Received August 6, 2001; revision received November 28, 2001; accepted December 6, 2001. Presented at the Society of General Internal Medicine annual meeting, Washington, DC, May 2000. Address for correspondence and reprints: Richard H. White, MD, Division of General Medicine, Suite 2400, PSSB, 4150 V Street, Sacramento, CA 95817. Fax: 916-734-2732; e-mail: [email protected].

METHODS Study Design. This was a prospective study of consecutive patients presenting to an urban academic ED with signs or symptoms suggesting the diagnosis of DVT in the lower extremity. This study was approved through the Human Subjects Review Committees at the University of California, Davis. Study Setting and Population. Patients over 18 years old who presented to the University of California, Davis, ED with signs or symptoms suggestive of lower-extremity venous thrombosis were el-

562

Shields et al. • PROBABILITY MODEL FOR DVT

TABLE 1. Canadian Probability Model for Acute Deep Vein Thrombosis Criteria 1. Active cancer (ongoing treatment for ⱕ6 months, or palliative) 2. Paralysis, paresis, or recent plaster immobilization of the lower extremities 3. Surgery 3 days 4. Thigh plus calf swelling on the affected side 5. Tenderness along deep venous system 6. Affected calf >3 cm larger than other side (measured 10 cm below tibial tuberosity) 7. Pitting edema (greater in affected leg) 8. Collateral superficial veins (no-varicosities) 9. Alternative diagnosis as likely or greater than venous thrombosis

Score 1 pt 1 1 1 1

pt pt pt pt

1 pt 1 pt 1 pt ⫺2 pts

Probability of venous thrombosis: Low probability = 0 or less points Intermediate probability = 1 or 2 points High probability = 3 or more points

igible for entry into the study. Exclusion criteria were: 1) a prior history of acute venous thrombosis or pulmonary embolus, 2) signs and symptoms of acute pulmonary embolus, 3) current use of an oral anticoagulant, 4) a positive pregnancy test, or 5) failure to complete adequate testing in the ED. Study Protocol. In a convenience sample of patients presenting with signs or symptoms suggesting venous thrombosis, emergency physicians were urged, but not required, to complete a one-page data collection form that included demographic information and clinical findings used in the Canadian clinical probability model (Table 1).1 A specific question on the form asked ‘‘Is there an alternate diagnosis as or more likely than venous thrombosis that might explain the patient’s findings, and, if so, please write the name of this alternate condition.’’ It was known that none of the ED staff used the Canadian probability model, and none knew the scoring rules to stratify different probability groups using this tool. The data collection form had to be completed prior to ordering any laboratory or radiology studies. All patients in whom venous thrombosis was excluded were followed up at the end of the threemonth follow-up period by a phone call to the patient, and by chart review. Primary endpoints were the presence of objectively documented venous thrombosis or pulmonary embolus within the three-month period. If a patient was hospitalized at another institution, records were requested and reviewed. Patients who had an initially negative venous ultrasound study and no clinical symptoms of thromboembolism, as confirmed by a follow-up telephone call, or objectively documented throm-

boembolism during the three-month follow-up interval, were categorized as not having venous thromboembolism. Measurements. After the initial clinical assessment, each physician was asked to order a rapid D-dimer test (SimpliRed, Agen Inc, Brisbane, Australia), a lower-extremity venous ultrasound imaging test, and any further tests deemed necessary. A final diagnosis of venous thrombosis was made based on the venous ultrasound test results. Possible pulmonary embolism was evaluated using leg ultrasound combined with either ventilation–perfusion lung scanning, helical tomogram scan, or pulmonary angiography. Venous ultrasound testing was performed in the radiology department or in the noninvasive vascular laboratory (HDI 5000, ATL Ultrasound, Bothell, WA) by a trained technologist. Findings were interpreted by the attending radiologist. The principal criterion for venous thrombosis was non-compressibility of a venous segment from the common femoral to the distal popliteal vein.9 Calf veins were not systematically evaluated. A diagnosis of pulmonary embolism was made using any of the following criteria: a high-probability ventilation–perfusion lung scan, a helical computed tomogram of the lung showing pulmonary embolization, or a positive pulmonary arteriogram. During the time of this study, the clinical laboratory frequently performed paired enzyme-linked immunosorbent assays (ELISAs; D-dimer Gold, American Diagnostica, Greenwich, CT) on plasma samples from patients being tested using the rapid D-dimer assay. The normal reference range for both the ELISA and a negative SimpliRed assay is