<img height="1" width="1" style="display:none" src="https://www.facebook.com/tr?id=648880075207035&amp;ev=PageView&amp;noscript=1">
4th Annual BayesiaLab Conference in Nashville, Tennessee

Presentation on September 29, 2016, at the 4th Annual BayesiaLab Conference:

Causal Attribution of Mortality to Delays in Heart Surgery

Boris Sobolev (Presenter)
Katie Jane Sheehan
Lisa Kuramoto
Guy Fradet

Presenter Biography

Boris SobolevProfessor Boris Sobolev is a Professor at the School of Population and Public Health, University of British Columbia. He leads the Health Services and Outcomes Research Program at the Center for Clinical Epidemiology and Evaluation. Professor Sobolev held the Canada Research Chair in Statistics and Modeling of the Health Care System from 2003-2013. He is an author of Analysis of Waiting-Time Data in Health Services Research and Health Care Evaluation Using Computer Simulation: Concepts, Methods and Applications, and is Editor-in-Chief of the Health Services Research series published by Springer.


Comparing outcomes across treatment groups is a main method of assessing the effectiveness of health care. For ethical, safety and economic reasons, such comparisons are rarely done using experiments. Here, we make causal attribution of the reduction in postoperative mortality to timing of non-emergency heart surgery using observations collected in a population-based cardiac registry. Combining what is known about factors affecting postoperative death and factors affecting time to surgery into a directed acyclic graph, we identify factors producing biasing associations between timing and death. Conditioning on these factors, we block such associations and estimate the potential outcomes of death if all surgeries were delayed and if all were timely. Averaging over these potential outcomes, we estimate that providing surgery to all patients without delay will lead to 4 fewer deaths per 1000 surgeries. We further estimate that 31% of delayed patients would have survived had they not been delayed. These methods advance health services research by allowing estimation of causal effects using data from routine medical care when experimentation is not feasible.