By Jackson Cathey, Carson Herman, Michelle Tetro, Melody Gao

Department of Statistics, Duke University

Abstract

As primary care providers deal with a multitude of chronic conditions, cardiovascular disease may often be overlooked. Finding cardiovascular disease early in a patient increases treatment options and reduces the likelihood of further complications. Thus, it is important that providers have tools at their disposal to assess patients’ odds of heart disease, given past medical histories and vital signs in clinic. Utilizing patient data from the Cleveland Heart Clinic, we developed a predictive model for assessing the odds that a patient evaluated in clinic may have an underlying heart condition. We selected each variable based on its relative capacity for prediction of heart condition, then further analyzed the contributes of this feature in the model. The final model includes the sex of the patient, their self-reported experience of exercise-induced angina, and resting ECG results. This model can be used in the clinic to estimate a patient’s odds of heart disease. Unlike current models used in primary care clinics, this model uses fewer variables and specifically assesses the odds of a current condition, as opposed to the future development of cardiovascular disease. Testing directly for heart disease is costly and time-consuming, so it’s useful to build a predictive diagnostic model.