This JAMA Guide to Statistics and Methods summarizes the advent, data considerations, utility and unique features, and statistical considerations of the Veterans Affairs National Surgical Quality Improvement Program (VASQIP) for use in surgical quality improvement research.
ADVENT OF THE VETERANS AFFAIRS SURGICAL QUALITY IMPROVEMENT PROGRAM
Since the early 1990s, US Veterans Affairs (VA) has been at the vanguard of national efforts to measure hospital-level performance and ensure quality care for veterans. In response to a congressional mandate that “the VA should report its surgical outcomes in comparison to the national average…with risk adjustment,” the initial VA National Surgical Quality Improvement Program (NSQIP) was created to accurately collect clinical data using standardized methodology and incorporating robust risk adjustment.1 Renamed the VA Surgical Quality Improvement Program (VASQIP) after merging the cardiac and noncardiac surgery components of NSQIP, this mandatory, VA-wide program has remained a model for national quality improvement (QI) efforts and was the template used to develop the private sector American College of Surgeons–NSQIP.
Cases abstracted by trained local VASQIP nurses are identified on an 8-day cycle to ensure data collection periods begin on different days of the week and provide a more representative sample of cases performed at each hospital. Data abstraction involves a thorough review of the electronic medical record and abstraction of preoperative, intraoperative, and postoperative variables based on standardized definitions. All patients are followed for a full 30 days after surgery, regardless of the duration of the postoperative in-hospital length of stay. Abstraction of certain higher-volume cases, such as inguinal hernia repair, is limited to no more than 5 per cycle. Data are collected for up to a maximum of 36 cases per cycle.
The Veterans Affairs Surgical Quality Improvement Program (Box 7) is a rich and accurate source of data for QI and surgical research.2 A complete VASQIP assessment includes more than 200 variables with 78 manually abstracted for each case. In 2009, The American Journal of Surgery dedicated an issue to comprehensively summarize the first 15 years of NSQIP.3 Below is a brief synopsis of the type of data available in VASQIP.
BOX 7 Best Practices for Using VASQIP
For patients with more than 1 operation in the data, only evaluate the first operation in each 30-day follow-up interval.
Review the definitions of each individual data point and outcome to avoid misinterpretation.
Given the large sample size, correlate clinical significance of findings with statistical significance.
Crosslinking of VASQIP to other available databases at the patient level is a unique resource for research within the VA.
Use judicious empirical selection of model covariates based on established evidence or biologic/clinical plausibility to avoid overfitting.
Consider hierarchical models and reliability adjustment to account for patient clustering and statistical noise from small sample size, respectively.