Polishing the Crystal Ball: Risk Prediction Methods Need Update

Calculating the risk that a heart attack patient will die or have another heart attack is physicians' attempt at peering into a crystal ball, but their view is clouded. Current calculation methods omit important facts that could better predict a person's future health and guide treatment, researchers report in today's rapid access issue of Circulation: Journal of the American Heart Association.

Risk stratification "scores" allow physicians to predict the likelihood of a person having future health problems associated with a condition like heart disease. Here, researchers examined risk stratification after a heart attack.

"In our study, we found that scores used to predict risk in heart attack patients should consider a patient's other health problems (comorbidity) and a measure of the heart's pumping strength (ejection fraction)," says senior author Veronique L. Roger, M.D., M.P.H., of the division of cardiology and the department of health sciences research at the Mayo Clinic Foundation in Rochester, Minn. "We believe that the current scoring systems should be re-evaluated because most scores do not consider either variable."

The most recent scores that physicians have been using to measure risk in heart attack patients were derived from clinical trials. Roger and colleagues believe, however, that scores should be validated in a more general population to verify their accuracy. In addition, the various scoring methods had not been compared to one another to determine their accuracy.

"Our study has given these scoring systems a reality check and determined that more information should be gathered to predict risk more accurately in the community," Roger says.

Researchers reviewed heart attack records of the population of Olmsted County, Minn., and categorized each patient based on the type of heart attack they had -- ST segment elevation (STEMI) or non-ST segment elevation (NSTEMI), according to their electrocardiographic presentation.

The researchers selected two of the scoring systems -- one derived from a clinical trial known as TIMI, and one created from a community-based study called PREDICT -- and examined their accuracy at predicting death or a recurrent heart attack. In addition, they assessed the value of adding a patient's ejection fraction as a variable in each of the scores.

From 1983 to 1994, there were 1,279 heart attacks. After an average follow-up of about 6¸ years, researchers found that the survival was similar for the 562 STEMI and 717 NSTEMI patients. The accuracy of the TIMI score was good in STEMI but only fair in NSTEMI, researchers say. The PREDICT score was consistently superior to that of the TIMI scores, mostly because PREDICT incorporates the effect of other diseases such as diabetes or hypertension (comorbidity). The study also showed that ejection fraction provided important additional information over that provided by the scores alone, including the comorbidity variable.

The researchers believe that the PREDICT score faired better because it was developed from a population-based study that used methods similar to their own, and its use of comorbidity data makes its risk predictions more suitable to a community with complex medical histories. They write that the "scores developed in trials are influenced by the selection process inherent to trials such that comorbidity is not integrated in such systems."

They explain that as heart disease shifts toward older age groups, incorporating a patient's other diseases, such as diabetes and hypertension, are integral to accurately predicting risk. Clinical trials typically involve younger people, they add.

Because most Olmsted County residents are white, the findings may not apply to other populations.

In an editorial on the study, William S. Weintraub, M.D., professor of medicine at Emory University in Atlanta, notes that risk stratification is an important endeavor.

"How many times do we hear from our patients 'Doctor, what is going to happen to me?' Patients, physicians and other providers would like to be able to predict what will happen after a major event in order to more rationally plan future care," he writes.

He agrees that the current scoring systems are useful, but better ways should be developed to calculate scores to consider a wider range of outcomes. The simplicity of the current scores allow them to be calculated at a patient's bedside, but he believes that handheld devices can make more complex risk prediction possible at the bedside.

Co-authors are Mandeep Singh, M.D.; Guy S. Reeder, M.D.; Steven J. Jacobsen, M.D., Ph.D.; Susan Weston, M.S.; and Jill Killian, B.S.