“Bedside” model for predicting an individual's risk of developing dementia

International Conference on Prevention of Dementia
Washington, D.C., June 10, 2007

Physicians currently have a variety of tools to predict the likelihood that someone will develop heart disease over the next 10 years. Deborah E. Barnes, Ph.D., M.P.H., Assistant Professor of Psychiatry at the University of California, San Francisco, and colleagues sought to develop a similar tool to predict an individual’s risk of developing dementia.

The researchers studied 3,375 subjects in the Cardiovascular Health Cognition Study who did not have dementia at the beginning of the study and determined which combination of factors most accurately predicted whether or not they would develop dementia over the next six years. They developed two models: the “best” model included all possible factors and the “bedside” model included only those factors that could be easily assessed by a physician during a clinic visit.

The factors in the “best” model included: older age, lower cognitive test scores (Modified Mini-Mental State Exam, Digit Symbol Substitution Test), low body mass index (<18), slower physical function (time to put on a button a shirt, time to walk 15 feet), history of by-pass surgery, lack of alcohol consumption, possession of 1 or more apolipoprotein E e4 alleles, evidence of enlarged ventricles or white matter disease on cerebral magnetic resonance imaging scans, and thickening of the carotid artery wall on ultrasound. The “bedside” model included older age, lower cognitive test scores, low body mass index (<18), slower physical function, history of by-pass surgery, and lack of alcohol consumption

The overall accuracy of the “best” model was 88 percent. Six percent of people with low scores on the “best” model developed dementia within six years compared to 25 percent of people with moderate scores and 52 percent of people with high scores. The overall accuracy of the “bedside” model was almost as good: 87 percent. Using the “bedside” model, the risk of dementia was six percent for people with low scores, 25 percent for people with moderate scores, and 54 percent for people with high scores.

“In this study population, it was possible to predict an individual’s six-year risk of developing dementia with high accuracy,” Barnes said. “The ‘best’ model was significantly better than the ‘bedside’ model, although the absolute difference between the models was relatively small. Future studies should validate these tools in other study populations.”

_____________________________________________________________________