Eyeball test may not see risk accurately

Twitter icon
Facebook icon
LinkedIn icon
e-mail icon
Google icon
 - eye, eyeball

Statistical estimates of mortality risk after cardiac surgery are more accurate than physician estimates based on clinical assessments—also known as the “eyeball test”—although both methods tend to overestimate risk, a study published online Jan. 14 in Circulation: Cardiovascular Quality and Outcomes found.

Renuka Jain, MD, of the University of Minnesota Medical School in Minneapolis, and colleagues used a study population of more than 5,000 consecutive patients who underwent cardiac surgery between 1993 and 2010 at the Minneapolis VA Medical Center. They calculated a risk estimate and also asked cardiac surgeons to estimate risk based on their clinical assessments. They excluded patients who did not have CABG or valve surgery.

The primary outcome was 30-day operative mortality, defined as either less than 30 days after surgery from any cause or more than 30 days but due to a complication from surgery. They also evaluated long-term mortality at one and five years after surgery.

There were a total of 168 deaths at 30 days (3.3 percent), 360 deaths at one year (7.1 percent) and 942 deaths at five years (18.5 percent). There was a modest correlation between the eyeball test and the statistical estimates. Statistical risk estimates were significantly better than the eyeball test at predicting mortality at 30 days, at one year and at five years. Physician risk estimates were higher for elderly patients and all types of surgeries, except in high-risk patients.

Their findings, the authors argued, “suggest that neither statistical nor physician’s risk estimate is accurate but rather provides a general basis for a meaningful discussion with the patient.”

In an accompanying editorial, Karl F. Welke, MD, MS, of the University of Illinois College of Medicine in Peoria, argued that statistical risk estimates fall short. “We need to improve the validity of statistical risk models by expanding them to include additional predictive variables and keep them up to date,” he wrote.

The most accurate risk estimate, he added, may come from “our assessment of the patient and our knowledge of the care environment added to that derived from a statistical risk model.”