Modeling the Number of Obstructed Coronary Arteries in Adults Aged 50 Years and Older: A Comparison of Poisson, Negative Binomial, Zero-Inflated, and Hurdle Models

Background: Coronary artery disease (CAD) is a leading cause of morbidity and mortality worldwide. Modeling the number of obstructed coronary arteries as count data may better reflect disease severity than conventional binary approaches, particularly when excess zeros are present. To identify determinants of coronary artery obstruction and compare count regression models for analyzing disease severity.
Methods: This retrospective cross-sectional study included 902 adults aged ≥50 years who underwent coronary angiography at Imam Reza Hospital, Iran. The outcome was the number of obstructed coronary arteries (0–3). Poisson, Negative Binomial (NB), Zero-Inflated Poisson (ZIP), Zero-Inflated Negative Binomial (ZINB), and Hurdle Negative Binomial (HNB) models were compared using AIC, BIC, log-likelihood, RMSE, MAE, and the Vuong test.
Results: Among participants, 28.9% had no coronary obstruction and 34.5% had three-vessel disease. No significant overdispersion was detected. Zero-modified models outperformed conventional Poisson and NB models. Although the HNB model showed the best fit, the ZIP model was selected because of its comparable performance and greater parsimony. Female sex was associated with fewer obstructed arteries (IRR = 0.82, P = 0.016), whereas younger age, female sex, and absence of diabetes increased the probability of having no coronary obstruction.
Conclusion: The ZIP model provided a simple and clinically meaningful approach for modeling coronary artery obstruction counts. Age, male sex, diabetes, and smoking were the main determinants of CAD severity.