Outcomes reported by the Vascular Quality Initiative and the National Surgical Quality Improvement Program are not comparable

Objective: The Vascular Quality Initiative (VQI) and National Surgical Quality Improvement Program (NSQIP) have emerged as the primary vascular surgery quality measurement tools with the purpose of evaluating perioperative outcomes and assessing hospital and physician quality. VQI uses self-reporting to capture all index vascular procedures during the inpatient period. NSQIP employs nurse abstractors to capture a sample of procedures and covers 30-day events. We hypothesize that patients undergoing lower extremity bypass (LEB) will exhibit high concordance for preoperative variables and low concordance for postoperative variables between these data sets. Methods: All patients undergoing LEB for peripheral arterial disease at the University of Massachusetts captured in both VQI and NSQIP databases were reviewed (2007-2012). Concordance between categorical variables was assessed by kappa correlation coefficient. All postoperative variables were compared during equivalent inpatient stay. Events between discharge and 30 days postoperatively were tabulated with use of the NSQIP data set. Results: We identified 240 patients undergoing LEB captured in both VQI and NSQIP. Comparison of this identical patient cohort between VQI and NSQIP revealed a moderate to strong agreement for most preoperative variables except for congestive heart failure (kappa=0.14) and hypertension (kappa=0.35), which showed poor agreement. Concordance for inpatient postoperative variables was high for mortality (kappa=1.0) and myocardial infarction (kappa=0.86) but moderate for pulmonary complications (kappa=0.57) and poor for unplanned return to the operating room (kappa=0.41), wound infection (kappa=L0.01), and change in renal function (kappa=L0.01). A majority of postoperative events (71%) occurred between discharge and 30 days postoperatively, with a significantly higher incidence of wound infections in the outpatient setting (4.2% vs 95.8%; P <.0001). Conclusions: VQI and NSQIP demonstrate substantial concordance for most preoperative variables and poor concordance for most postoperative variables, even at identical collection periods. This discordance is a result of differences in data collection methods and variable definitions. On the basis of these findings, VQI and NSQIP data sets cannot be used to directly compare risk-adjusted patient outcomes between institutions.