The Learning Curve of an Academic Cardiac Surgeon: Use of the CUSUM Method

Publisher Website
Google Scholar
Abstract
Background: Despite the sizeable volume of research on the determinants of outcome after cardiac operations, few articles have analyzed the learning curves of individual cardiac surgeons over time. The objective of our study was to analyze statistically the learning curve of an academic cardiac surgeon in reducing operative morbidity and mortality during a 10‐year interval. Methods: The study cohort of 1347 consecutive and unselected patients undergoing cardiac surgical operations from October 1988 to September 1998 were grouped into five 2‐year blocks (periods 1 to 5) according to the date of operation. The main outcome measures were operative mortality rate and standardized definitions of periopera‐tive myocardial infarction, intra‐aortic balloon pump use, reoperation for bleeding, stroke, sternal wound complications, sepsis, and respiratory insufficiency. Preoperative risk factors and operative results in periods 1 to 5 were compared statistically using a chi‐square test for linear trend (categorical variables) or analysis of variance with linear contrast and lack of fit tests (continuous variables). In addition, the cumulative sum (CUSUM) method was used to determine the association among operative morbidity, mortality, and prespecified 80% alert and 9590 alarm boundary lines in practice years 1, 5, and 9. Results: Of the preoperative risk factors, only patient age showed an important change during the 10 years of the study (61.3 ± 0.7 to 64.3 ± 0.6, p = 0.001). There were no statistically significant changes from periods 1 to 5 in overall operative mortality (4.0% to 2.2%, p = 0.56) or in the rates of perioperative stroke (1.8% to 3.840, p = 0.331, sternal wound complications (0.496 to 0.896, p = 0.97), sepsis (0.9% to 0.856, p = 0.631, or respiratory failure (4‐‐4% to 2.8Y0, p = 0.21). Decreases occurred in a linear fashion during periods 1 to 5 in mortality after coronary artery bypass grafting (5.1% to 1.3%, p = 0.012) and in the rates of perioperative myocardial infarction (7.0% to 2.2%, p = 0.005), intra‐aortic balloon pump use (7.096 to 3.0%, p = 0.051, and reoperation for bleeding (8.4% to 2.2%, p = 0.001). The number of uneventful cases between a death or complication increased from 2.82 ± 0.43 in period 1 to 6.44 ± 1.10 in period 5 (p < 0.001). On CUSUM analysis, the cumulative failure rate in year 1 transgressed the upper 80% alert line after 56 cases and the upper 95% alarm line after 69 cases. During years 5 and 9 the failure rate gravitated around the 80% and 95%“reassurance” lines, respectively, indicating improved results as compared to year 1. Conclusions: The mortality rate after coronary artery bypass grafting and select perioperative morbidity rates improved in a linear fashion from the onset of independent practice to year 10. The CUSUM method was helpful in identifying suboptimal results during the first year of practice and shows promise as a method of prospective quality control in cardiac surgery. These data support mentorship of new consultants by a senior surgeon during the first year or two of independent practice. (J Card Surg 1999; 14:312–320)