This proposal describes a 5 year mentored training program in patient oriented clinical research. The PI has completed a residency and research fellowship in Emergency Medicine and is seeking further training to study Bayesian probability, diagnostic test assessment, resource utilization, and classification tree analysis techniques to predict presence or absence of pulmonary embolism (PE). The most common symptoms of PE, chest pain and shortness of breath are included as the chief complaints of an estimated 10 million people in US emergency departments annually. Testing for PE must be done in conjunction with estimation of the pretest probability that the clinician believes to be existent at the time of patient presentation. There are no uniformly used or accepted means of estimating pretest probability. Two methods have been suggested in North American populations, the Wells score and the Charlotte criteria, but these give only ranges of probability estimates and include a series of questions, which must be memorized. They have been uncommonly used outside of their derivation populations. There has also been a recent proliferation of blood based screening tests for PE such as the quantitative D-dimer test. If normal, in low pretest probability patients, these rapid, highly sensitive, low specificity tests alone may be used to reduce the likelihood of PE to below a safe threshold. The ease of ordering a blood test to screen for PE may result in significant changes in overall test utilization, radiological test use, and overall disease diagnosis. This prospective observational study will quantify the effects of a new quantitative D-dimer on these three parameters, as well as test the hypothesis that low risk patients (Wells and Charlotte criteria) with a negative D-dimer have a very low prevalence of PE (<1.0%) as determined by imaging test results, and 3 month follow-up. To improve the safety, accuracy and efficiency of pretest probability assessment and D-dimer testing, a new model identifying low risk patients (goal: <5% pretest probability) will be derived from this data by non-linear modeling and classification tree analysis. Dr. Charles Bennett will be the mentor for this project and is a national leader in hematological and oncology research. He has significant experience in mentoring junior investigators and an emphasis in disease detection and outcome prediction to ensure development of the PI toward that of an independently funded investigator.
Showing the most recent 10 out of 22 publications
