Wide variability has been reported in the administrative and quality improvement databases that are used to guide evidence-based decision-making in vascular surgery. For example, researchers at Brigham and Women’s Hospital recently found mortality after open abdominal aortic aneurysm repair (OAAA) was lower in the Vascular Quality Initiative database than in reports from other national registries. As they explained in the Journal of Vascular Surgery, this raises concern that benchmarks for complication and mortality rates may be too permissive if based on large registry data.

Extending that research, the team has detected numerous differences in patient populations and outcomes between three national data repositories, again using OAAA repair as a model. These data were published in the Journal of Vascular Surgery by Rebecca E. Scully, MD, MPH, clinical fellow in the Division of Vascular and Endovascular Surgery at the Brigham, Michael Belkin, MD, chief of the division, and colleagues.

Methods

The researchers identified individuals who underwent elective OAAA repair between 2013 and 2016 and had records recorded in the:

• Vascular Quality Initiative (VQI) database, maintained by the Society for Vascular Surgery—Captures 350 clinical, procedural and outcomes variables from 14 registries of common arterial and venous interventions (number of patients included in this study=3,912)

• National Surgical Quality Improvement Program (NSQIP) database, maintained by the American College of Surgeons—Captures preoperative and periprocedural data on >1 million cases annually at more than 700 institutions; it includes follow-up data for up to 30 days but, like the VQI, it lacks cost data and reporting is voluntary (n=1,667)

• National Inpatient Sample (NIS), maintained by the Agency for Healthcare Research and Quality—Nationally representative sample of more than 35 million U.S. hospitalizations annually; contains procedure codes and cost information but not anatomic or procedural data (n=3,196)

Unadjusted Analyses

Significant differences existed between groups with regard to the primary outcomes:

• Post-procedure inpatient mortality—3.3% in the VQI, 5.2% in the NSQIP, and 5.5% in the NIS (P<0.001)

• 30-day mortality—3.5% in the VQI and 5.9% in the NSQIP (P<0.001) (percentage in the NIS unavailable)

• Length of stay—9.4 days in the VQI, 9.5 in the NIS, and 10.3 in the NSQIP (P=0.003)

• Discharge disposition—73% of individuals in the NIS and NSQIP were discharged to home vs. 76% in the VQI (no P value reported)

The groups also differed significantly in racial distribution of patients, primary insurer and comorbidities.

Adjusted Analyses

Some differences persisted after adjustment for age, gender, race, comorbidities and smoking:

• Inpatient mortality—adjusted OR (aOR), 1.79 for the NSQIP vs. VQI, and 1.52 for the NIS vs. VQI (P<0.001 for both comparisons)

• Discharge to home—aOR, 0.82 for the NIS vs. VQI (P=0.001), and 0.75 for the NSQIP vs. VQI (P<0.001)

Implications for Policymakers

These findings imply that when formulating practice guidelines and consensus statements, experts in vascular surgery should:

• Avoid directly comparing the information derived from different data repositories

• Interpret data within the context of the respective strengths and limitations of the source dataset

• Strive to account for differences between databases when setting quality thresholds in order to guarantee equal access to high-quality care for all patients