Safe, highly selective use of pulmonary artery catheters in coronary artery bypass grafting: An objective patient selection method

Background. Routine versus selective use of pulmonary artery catheter (PAC) monitoring in coronary artery bypass grafting operations is a topic of significant debate. Accordingly, we retrospectively examined operative outcomes in 2,685 consecutive (1994 to 1998) coronary artery bypass grafting patients in whom PAC use was highly selective. Next, we developed a quantitative model of PAC use in terms of its multivariate predictors as a means of providing an objective criterion for patient PAC use selection. Methods. Safety of the implemented selective PAC use was assessed by comparisons to contemporaneous coronary artery bypass grafting outcome reported by The Society of Thoracic Surgeons' national data. Continuous relations describing PAC use in terms of continuous univariate predictors were obtained using overlappingrange patient cohorts. Next, independent predictors of PAC use were derived by multivariate regression to best fit the categorical variable PAC (Yes = 1, No = 0). Model estimates were a continuous variable (PAC score) with values between 0 and 1. Results. Planned use of PAC was based on collective consideration of preoperative patient variables, and was not limited to low-risk or preserved ejection fraction patients. Planned and unplanned use of PAC was limited to 176 (planned, 6.6%) and 66 (unplanned, 2.4%) patients, respectively, whereas no PAC was used in 2,443 (91%.). Overall patient characteristics and risk factors in this series were comparable to contemporaneous Society of Thoracic Surgeons data, and the incidence of operative deaths was 2.31% (n = 61; observed-to-expected [Society of Thoracic Surgeons risk] mortality = 0.73). Independent predictors of PAC use were ejection fraction, Society of Thoracic Surgeons risk, intraaortic balloon pump, congestive heart failure, reoperative surgery, and New York Heart Association class IV. Expectedly, PAC scores were substantially different for PAC (mean standard deviation, 0.37 +/- 0.20; median, 0.38) and no PAC (0.14 +/- 0.11; median, 0.10) patients (p < 0.001). Area under the receiver operating characteristic curve derived for PAC score was relatively high (area, 0.85). Moreover, the corresponding summed sensitivity (0.68 to 0.91) and specificity (0.85 to 0.62) was maximized at 1.53 for PAC score between 0.15 and 0.31. Conclusions. Our results indicate that highly selective use of PAC in coronary artery bypass grafting can be accomplished safely, and it need not be limited to patients with preserved ejection fractions or low operative risk. Indeed, coronary artery bypass grafting without PAC may be preferable in the vast majority of patients as it reduces catheter-associated risks and resource utilization without incurring an increased operative risk. Also, pending further prospective confirmation, our analysis suggests that collective consideration of PAC use predictors to derive a PAC score provides an objective criterion to minimize unnecessary use of PAC with an acceptably low probability of error.