Yanyun Liu

Date of Award


Degree Type


Degree Name

Master of Science (MS)




Lehana Thabane




Coronary artely bypass grafting (CABG) is the most commnonly performed "open heart" operation in North America. Complex cardiac surgeries served for a large amount of the cardiac surgery population, but outcomes after these surgeries have been limited by lack of appropriate interpretation. Given the observed trend toward an increasing proportion of complex cardiac surgeries, there is a great need to understand the outcomes and patterns of resource utilization for the population who have had complex cardiac surgery.

The clinical objectives of this thesis are to compare clinical outcomes and resource usage between isolated coronary bypass grafting and complex cardiac surgery and detelmine the difference of outcomes for complex cardiac surgeries among cardiac surgical sites across Canada.

The statistical objective of this thesis is to compare Bayesian and classical methods of analyzing two surgeries difference in outcomes. The classical methods are multivariable logistic regression, matched propensity score method, propensity score weighted regression and stratified propensity score method. The Bayesian method is Bayesian matched propensity score.

For the primary outcome mortality, the odds ratio and 95% confidence interval for the treatment effect is 4.49 (1.92, 10.56) for propensity score matching method, 4.97 (3.62, 6.11) for propensity score weight method, 3.49 (1.91, 6.40) for propensity score strata method, 3.71 (2.10, 6.56) for multivariab1e regression method, and 3.82 (1.23, 13.07) for Bayesian propensity score matching method. Different methods obtained different treatment effect estimates.

We concluded that patients who are undergoing complex cardiac surgery have a greater risk for adverse postoperative events and longer ICU length of stay compared to patients who are undergoing isolated CABG. We also found that there is variability in
outcomes and resource usage among Canadian cardiac centers.

McMaster University Library