Substantial Variation in Post-PCI Bleeding Observed Across US Hospitals

Bleeding after PCI varies widely across the United States at the hospital level even after adjustment for patient and procedural characteristics, according to a study published online November 25, 2014, ahead of print in Circulation: Cardiovascular Quality and Outcomes. Thus, opportunities may exist for poor-performing sites to learn best practices from hospitals with low bleeding rates, the authors say.

Eric D. Peterson, MD, MPH, of the Duke Clinical Research Institute (Durham, NC), and colleagues analyzed data from almost 2 million PCIs performed at 1,292 National Cardiovascular Data Registry hospitals from July 2009 to September 2012. All hospitals were required to perform a minimum of 50 procedures over the study period for inclusion. Bleeding was defined as any of the following events occurring within 72 hours after PCI:

  • Arterial access site bleeding, either overt external bleeding or hematoma > 10 cm for femoral access, > 5 cm for brachial access, or > 2 cm for radial access
  • Retroperitoneal, gastrointestinal, or genitourinary bleeding
  • Intracranial hemorrhage
  • Cardiac tamponade
  • Decrease of ≥ 3 g/dL in hemoglobin post-PCI with preprocedure hemoglobin ≤ 16 g/dL
  • Postprocedure non–CABG-related blood transfusion with pre-PCI hemoglobin of ≥ 8 g/dL

Hospital Practices Affect Bleeding

The median bleeding rate was 5.2% overall, but there was little uniformity across hospitals. Bleeding rates in the 5th, 10th, 25th, 75th, 90th, and 95th percentiles were 2.6%, 3.0%, 3.9%, 6.9%, 8.8%, and 10.4%, respectively.

The investigators found “the need for red blood cell transfusion remains somewhat subjective,” so they looked at whether differences in the composite endpoint were related to differential thresholds in transfusion. However, “hospital pretransfusion hemoglobin values were highly consistent across all 3 bleeding tertiles,” they write.

Hospitals in the highest tertile versus the lowest and average tertiles of hospital-level bleeding were more often teaching hospitals and had lower median annual PCI volumes. Unfractionated heparin and GPIs were used more often during procedures at centers in the highest tertile of hospital-level bleeding, as well. In contrast, bleeding avoidance strategies such as bivalirudin (Angiomax; The Medicines Company), radial access, and vascular closure devices were used most frequently during procedures performed at hospitals in the lowest bleeding tertile.

Compared with patients treated at hospitals in the lowest and average bleeding tertiles, those treated at hospitals in the highest tertile were less often white and more likely to have a history of MI and prior congestive heart failure, but they were less likely to have undergone prior coronary revascularization. Patients in the highest hospital tertile also more frequently presented with STEMI and heart failure and more frequently underwent PCI for emergency and salvage indications than patients in the lower 2 tertiles.

Adjustment for Case Mix Did Not Erase Differences

Variation remained even after accounting for case mix however, with hospital risk-adjusted bleeding rates ranging from 2.8% in the 5th percentile to 9.4% in the 95th percentile. The overall pattern of hospital rankings was largely unaffected by adjustment (Spearman correlation coefficient 0.88), but the position of individual hospitals on the list of nearly 1,300 institutions rose or fell.

Considering both unadjusted and adjusted bleeding rates, among the hospitals there were 300 “low outliers” and 370 “high outliers.” Adjustment for case mix shifted the outlier status for 286 sites (22.1%), with changes in category for 29.3% of low outliers, 16.1% of nonoutliers, and 26.5% of high outliers.

Specifically, 81.5% of hospitals whose unadjusted bleeding rates were lower than expected and 83.4% of those whose bleeding rates were higher than expected still fell into their respective categories after adjustment.

The median bleeding rates at academic vs nonacademic hospitals were similar (5.4% vs 5.1%). Median annual hospital PCI volume was 391.2 cases, and this value was not associated with risk-adjusted bleeding (Spearman correlation coefficient 0.02). Nevertheless, bleeding rates were higher with greater hospital use of heparin and GPIs (P < .0001 for both) and lower with increased use of bleeding avoidance strategies (Spearman correlation coefficient -0.26). Various sensitivity analyses supported the results.

A Call for Nationwide Survey on Hospital Bleeding

In an email with TCTMD, study co-author Connie N. Hess, MD, MHS, also of the Duke Clinical Research Institute, explained that her team aimed to formally investigate whether post-PCI bleeding represents an appropriate quality metric.

“Studying [both high and low] outlier sites may provide an opportunity to educate low-performing sites to improve clinical outcomes and to study high-performing sites to identify strategies that may help other hospitals to achieve better patient outcomes,” she said.

Much of the observed variation can be attributed to differences in patient case mix, “with some hospitals treating sicker and higher bleeding risk patients,” Dr. Hess said. Residual variation after adjustment “may be partly due to provider choices, such as arterial access site or anticoagulation” as well as reporting bias or misreporting of bleeding, she suggested. “Thus, recognizing and addressing such modifiable factors may help to improve clinical outcomes.”

Although several PCI bleeding risk models have been developed, Dr. Hess said, “it may be that hospitals do not routinely implement these models when assessing which patients are at highest bleeding risk and may benefit the most from bleeding reduction strategies. Raising awareness about bleeding is an important part of the solution; to this end, including bleeding as a PCI performance measure may help to focus attention on the issue.”

She suggested that low-performing sites could learn best practices from nationwide hospital surveys, such as those used to reduce door-to-balloon times in STEMI patients. “Incorporation of bleeding risk models into patient care algorithms and standardized order sets might also help to reduce bleeding rates,” Dr. Hess concluded.




Hess CN, Rao SV, McCoy LA, et al. Identification of hospital outliers in bleeding complications after percutaneous coronary intervention. Circ Cardiovasc Qual Outcomes. 2014;Epub ahead of print.


  • Dr. Peterson reports receiving research funding from the American College of Cardiology, the American Heart Association, Eli Lilly, Ortho-McNeil-Janssen Pharmaceuticals, and the Society of Thoracic Surgeons and consulting for AstraZeneca, Boehringer Ingelheim, Genentech, Johnson & Johnson, Ortho-McNeil-Janssen Pharmaceuticals, Pfizer, Sanofi-Aventis, and WebMD.
  • Dr. Hess reports no relevant conflicts of interest.

Related Stories:

We Recommend