A Decision Support Model for the Necessity of Cardio-Angiography, A Result of Generalized Additive and logistic Regression Model

Fatemeh Rezaeisharif; Azadeh Saki; Ali Taghipour; Mohammad Tajfard

doi:10.18502/jbe.v7i2.6712

Fatemeh Rezaeisharif Department of Biostatistics, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
Azadeh Saki Department of Biostatistics, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
Ali Taghipour Department of Epidemiology, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.
Mohammad Tajfard Department of Health Education and Health Promotion, School of Health, Mashhad University of Medical Sciences, Mashhad, Iran.

DOI: https://doi.org/10.18502/jbe.v7i2.6712

Keywords: logistic regression; Generalized additive model; Decision support system; Angiography; hs-CRP

Abstract

Introduction: Angiography is used as the gold standard for diagnosis of coronary artery disease (CAD). It is an invasive procedure and has several complications. Also, some patients refuse angiograms for reasons such as fear, high cost, and loss of trust in physician diagnosis. The negative results of this test is more than a third. Therefore, having a statistical predictive model for estimating the risk of CAD, as an evidence-based support system, can help the physician and patient decide on the necessity of angiography.

Aims: In this study we aimed to find an evidence-based supportive model for decision making on the necessity of angiography in people who were candidates for angiography by the physician after initial tests.

Methods: In this study, 1187 patients who had been referred to Ghaem Hospital of Mashhad University of Medical Sciences and diagnosed with physicians after initial tests were enrolled. Demographic data, lipid and blood sugar levels, and the history of underlying disorders were variables that were studied in the statistical model fitting. Initially, generalized additive models were used singularly for quantitative predictors, then by applying right transformations on the predictor variables we entered them simultaneously in logistic and count regression models. These two models were fitted to the data using R software and then compared in terms of predictive accuracy.

Findings: Generalized additive models showed that the relationship between CAD with the hs-CRP level was not monotone. Exploratory analyzes showed that 62% of people with hs-CRP level <3 and 50% of people with hs-CRP levels between 3 and 6 were suffered from the CAD. The highest rate of CAD was seen in the range of 6-8 (93%) but with increasing the hs-CRP level to above 8 it decreased to 70%. The relationship between age and the risk of CAD was S-shaped. Risk of CAD in diabetic subjects with normal FBS was equal to that of nondiabetic subjects with normal fasting blood sugar. The age, gender, diabetes, FBS, and hs-CRP were significant in both models (p <0.05). The area under the ROC curve was upgraded to 81 for the logistic model.

Conclusion: The most important finding of this exploratory study was that out of 232 patients with hs-CRP level between 6 to 8, 217 (93%) had coronary artery occlusion, for these subjects the probability of occluding a coronary artery was 0.93. The present study also showed that if the blood sugar is controlled in patients with diabetes, then this disease will not be a risk factor for patients with coronary artery occlusion. The logistic regression model presented in this study is recommended as the final model to support decision-making about the necessity of angiography.