Predicting Obstructive Coronary Artery Disease in Patients Undergoing Coronary Angiography for Stable Angina

Accurately identifying the pre-test probability of obstructive coronary artery disease (CAD) in patients presenting with chest pain is key in determining the next steps in management. The Diamond-Forrester model, or an updated version, are currently recommended to calculate a patient’s pre-test probability of obstructive CAD in both the American and European guidelines, respectively. However, studies show there continues to be overutilization and inappropriate use of both invasive and non-invasive diagnostic testing in patients presenting with chest pain. Therefore, an updated model to predict the presence of obstructive CAD could help physicians better select patients who may benefit from further testing and avoid expensive testing in those at low risk for whom testing is unlikely to alter their treatment. Using prospectively collected data from patients undergoing coronary angiography for the evaluation of chest pain in the Coronary Angiogram Database of South Australia (CADOSA), we constructed multivariable logistic regression models using patient demographics (age, sex), clinical history (chest pain duration, quality, location, associated symptoms, precipitants, relievers), and co-morbidities (hypertension, dyslipidemia, diabetes, smoking, family history, cerebrovascular disease, peripheral vascular disease) to predict the presence of obstructive CAD. Prior to model development, multiple imputation methods were used to generate 25 randomly imputed datasets. Logistic regression models were built on each imputed data separately and the results were pooled to obtain final beta weights. Among 4,004 patients evaluated with coronary angiography for chest pain, just under fifty percent (1988/4004) were found to have obstructive CAD on coronary angiography and the other half (2016/4004) did not. Age, sex, chest pain duration, classic chest pain relievers (rest and nitroglycerin in <5 minutes), classic chest pain precipitants (exercise, emotional stress, cold weather), hypertension, dyslipidemia, diabetes, smoking status, and family history of CAD were all predictive of the presence of obstructive CAD. In a model containing only these variables, there was good calibration and discrimination (average c-statistic 0.739). Using a large, contemporary registry, we developed a model to determine pre-test estimates of obstructive CAD in patients presenting with chest pain. Importantly, the model contains variables which are readily available at initial patient presentation.

Name thesis:degree_name

M.S. (Masters of Science)

Level thesis:degree_level

Masters

Discipline thesis:degree_discipline

Bioinformatics (UMKC)

Grantor

University of Missouri--Kansas City

Year dc:date.issued

2021

Author dc:creator

• Thomas, Merrill

Advisor dc:contributor.advisor

• Spertus, John A.

Handle dc:identifier.uri

https://hdl.handle.net/10355/84168

OAI identifier oai:identifier

oai:mospace.umsystem.edu:10355/84168