Sequential logistic models for 30 days mortality after CABG: pre-operative, intra-operative and post-operative...

The goal of this paper was to examine the added effect of operative and post-operative variables on 30 days mortality, in addition to patients' case-mix factors.

A prospective study of 4835 patients, 95% of all Israeli patients who underwent coronary artery bypass grafting (CABG) in 1994. Information related to risk of death was collected at admission to hospital (preceding the operation), at time of the operation and in the immediate post-operative period. Deaths were independently ascertained.

Data collectors followed every patient from admission to discharge. Sequential logistic models were constructed for the 'case-mix', 'operative' and the 'post-operative' periods in chronological order. Each model incorporated and adjusted for the risk estimated at the previous point in time, by forcing individual risk scores.

Significant pre-operative risk factors for 30 days mortality, in the case-mix model included mainly severity of illness characteristics, such as, left ventricular dysfunction and emergency admission, (c-statistic 78.8%). Model 2 (the 'operation' model) included in addition to the case-mix score, excessive duration of the operation per graft, bleeding, etc. (c-statistic 85.3%). The post-operative model showed the added effect of the post-operative factors such as low haemoglobin, additional surgery, and excessive time on respirator, (c-statistic 92.4%).

The sequential analysis was an efficient method for updating patients' risk over time, where the number of events was small, relative to the number of risk factors. The addition of peri-operative factors increased significantly the predictive power of the model, adding clinical insights to the role of the hospital experience on 30 days mortality.