Institute:Office of National Coordinator (ONC) Workforce Training Curriculum
Component:The Culture of Health Care
Unit:Evidence-Based Practice
Lecture:Phrasing the clinical question
Harm and prognosis
Slide content:Clinical Prediction Rules Use of results of multiple tests to predict diagnosis (Adams, 2012) Best evidence establishes rule in one population and validates in another independent one Examples of clinical prediction rules: Predicting deep venous thrombosis (DVT) (Wells et al., 2000; Wells, Owen, Doucette, Fergusson, & Tran, 2006; Righini , 2013) High sensitivity, moderate specificity Better for ruling out than ruling in disease Coronary risk predictionnewer risk markers do not add more to known basic risk factors (Folsom et al., 2006) Inconsistent results for prognostic ability of popular risk prediction models ( Siontis , 2012) 10
Slide notes:10 We finish this lecture with some discussion about clinical prediction rules. We dont go into great detail, but many of you who are regular readers of the medical literature have probably seen papers where they are used. The idea behind clinical prediction rules is that we use results from multiple tests, in quotes here because the information used in clinical prediction rules includes not only things like blood tests and x-rays but also the presence of certain clinical findings, signs, and symptoms. All of these different pieces of data are used to predict the diagnosis. There are rules for critically appraising clinical prediction rule studies, and in essence, the best evidence for clinical prediction rules will establish the rule in one population and then validate it in another independent one. As an example, the prediction of deep vein thrombosis, or DVT [D-V-T] is very important clinically because there are no medical tests that can definitively diagnose this condition. DT is a blood clot in the deep veins of the lower extremities, which as all clinicians know, puts the patient at risk for the clot breaking off and causing a pulmonary embolismwhich can be serious and even fatal. Unfortunately, there are no tests that are both highly sensitive and specific for DVT, and so its helpful to try to develop clinical prediction rules that give us confidence in the diagnosis or help us rule out the diagnosis when we are seeing a patient who might have this condition. The prediction rule for DVT that Wells and colleagues developed has high sensitivity but moderate specificity. This is probably helpful because having high sensitivity, its good at ruling out disease, more so than at ruling it in. And with something as serious as DVT that can predispose to pulmonary embolism, its probably more important to be confident that we ruled out the disease rather than ruled it in. There are many other areas where clinical prediction rules have been applied. One study looked at predicting coronary artery disease by looking at all the different so-called markers that have been proposed for coronary artery disease in recent years. Interestingly, this study found that none of these newer risk markers add more to known basic risk factors, such as cholesterol level, family history, hypertension, and diabetes. Another study showed limited evidence and inconsistent results about the relative prognostic ability of the most popular risk prediction models for cardiovascular disease. The techniques of clinical prediction rules can be used to evaluate new markers for disease as they are developed as well as to compare results between studies.