The proposed study is intended to address a number of important methodological and health policy issues in assessing the quality of cardiac surgical care in Ontario, Canada. Using a new, population-based database from Ontario, this study will contain three major components. First, a longitudinal study will be conducted assessing trends in risk- adjusted cardiac surgery outcomes (mortality, ICU length of stay, overall postoperative length of stay) in the period 1991 to 1993, among the 9 hospitals performing adult cardiac surgery in Ontario. This study will enable us to determine whether most of the variation in outcomes is due to random cancer alone or due to systematic differences in quality of care. Second, we will compare the volume, case mix, and outcome of coronary artery bypass grafting (CABG) in Ontario with those in New York State for the period 1991 to 1992. This study will enable us to determine whether there are any significant international differences in risk-adjusted outcomes and will enable us to determine in which case mix groups, CABG surgery is being performed at higher rates in the United States. Third, we will evaluate the potential role of artificial neural networks as a method for risk adjustment in assessing the quality of cardiac surgical care using the Ontario database. We will compare the performance of neural network models with logistic regression-based models developed using the same data base. From this study, we will be able to determine what the advantages and disadvantages are of using neural networks as a risk-adjustment tool. The results from the proposed project should be of international interest and will have important health policy implications in both Canada and the United States.
