Pre-Test Probability for Assessment for Pulmonary Embolism Grant uri icon

abstract

  • Pretest probability assessment (PTP) plays a pivotal role in the evaluation of pulmonary embolism (PE) in the emergency department (ED). Current methods of PTP assessment produce broad categories of risk that tend to overfit individual patients and not reduce test-ordering. The primary hypothesis of this application is that the PTP of a life-threatening process can be accurately estimated by matching the clinical attributes of an individual patient to an identical profile of attributes shared by patients drawn from a large, reference database. The reference database must contain detailed data from an appropriate spectrum of patients previously evaluated for that specific process and for whom outcomes are encoded. The computerized system returns all matched patients and estimates PTP as the proportion of matched patients with an outcome of the disease in question. To accumulate a large database from ED patients undergoing objective testing for PE, previous work developed an electronic dataform (e-form) to collect clinical data using PDAs and desktop computers in conjunction with a website to upload entered data. In the present study we will employ the previous technology to collect clinical data on ED patients evaluated for PE in our ED, in concert with a large, multicenter study. Our aim is to take part in the prospective collection of PE e-forms on consecutive patients evaluated for PE in our ED, and to complete 45-day follow-up on all patients. Patients will be enrolled based upon implicit determination of enough clinical suspicion of PE to compel the clinician to order a contrast-enhanced CT angiography of the chest (CT A), a ventilation-perfusion lung scan of the chest or a quantitative D-dimer (either rapid ELISA or immunoturbidimetric). The prospective PE e-form will be completed in the ED and 45-day follow-up will be completed. Both the prospective and follow-up PE e-forms are uploaded to the BreathQuant Medical Systems website. Prior to allowing upload of data, the website determines if all mandatory fields are completed to ensure 100% complete data. The second aim is to perform statistical analysis on the database using classification and regression tree analysis. Statistical testing will be performed on the reference database using classification and regression tree (CART) analysis to determine which variables should be used as attributes in the matching process. We will use CART to select 8-10 attributes that will be used to assess PTP of PE by returning all patients in the database with the identical profile of these attributes and calculating the percentage with a positive 90 day outcome. The database will ultimately be used to help physicians estimate a tailor-made pretest probability of PE.

date/time interval

  • July 2006 - June 2007