One way that doctors predict a person's heart-attack risk may misclassify 5.7 million Americans, a new study suggests. The result is that millions of patients may be either under- or over-treated for coronary problems.

The standard method used, in accordance with national guidelines, is the so-called Framingham model. It takes risk factors such as age, cholesterol levels, blood pressure and smoking into account, and estimates a patient's risk of a heart attack, stroke, or other coronary event in the next 10 years. The calculation sorts patients into three risk groups: moderate, moderately high and high.

The trouble is that a person's risk, when calculated with a simplified version of the Framingham model, doesn't match the risk calculated when the tried-and-true original model is used, according to researchers at the University of California, San Francisco. (The original model uses a more complicated mathematical equation.)

"We thought there might be significant differences between the two methods," which could have significant impacts on patients' treatments for heart disease , study researcher Dr. Michael Steinman, an assistant professor of medicine at UCSF, said in a statement. "And in fact, that turned out to be the case."

The researchers used data from 2,543 people who participated in surveys sponsored by the Centers for Disease Control and Prevention between 2001 and 2006. They calculated each person's risk based on the original Framingham model and on the simplified model, and compared the differences.

The differences "turned out to be substantial for many patients," Steinman said.

Under the simplified model, 15 percent of the subjects were placed at a different level of risk than they were under the original model.

That means that 5.7 million Americans would be placed into different risk groups using the simplified model than they would be using the original model, the researchers said, with 3.9 million misclassified into higher risk groups and 1.8 million misclassified into lower risk groups.

"A lot of individuals would be treated differently either more aggressively or less aggressively using the point-based model," Steinman said.

The simplified model was introduced over a decade ago, when computers and personal digital assistants were less powerful and not so common in private medical practices, Steinman said.

"Just about any computer or PDA in use today can calculate the original Framingham model," Steinman said. "So there's not much reason to use the point-based system anymore in most instances."

Steinman cautions that the study was not designed to determine the benefits or harms for individuals who would be treated differently based on the results of the two models.

"With risk prediction models being increasingly used for many different diseases and conditions, this could be a general problem in the field of medicine," Steinman said. "In creating simplified risk models, we have to be aware of the potential impact on individual patients."

The study's authors disclosed that one study researcher is a lead plaintiff a lawsuit which alleges that improper marketing campaigns promote the simplified version of the Framingham model in an effort to increase sales of lipid-lowering medications.

The study was published online on Sept. 8 in the Journal of General Internal Medicine, and was funded by the National Institute on Aging, the American Federation for Aging Research, the Hartford Foundation, the Department of Veterans Affairs and the National Institutes of Health.