Data from routine checkups used to predict dementia

An algorithm based on the data can be used to help doctors rule out risk for developing dementia within five years.
By Stephen Feller  |  Jan. 22, 2016 at 3:01 PM
share with facebook
share with twitter

LONDON, Jan. 22 (UPI) -- Information collected by doctors at basic checkups can be used to accurately predict risk for developing dementia, according to researchers at University College London.

The researchers developed an algorithm, called the Dementia Risk Score, they say can help rule out patients at very low risk for dementia-related conditions such as Alzheimer's disease within five years.

"The score could be especially useful for identifying people at a very low risk of dementia, as recorded by their general practitioner," Dr. Kate Walters, a primary care and population health researcher at University College London, said in a press release. "This could help general practitioners working with people who are anxious about developing dementia."

For the study, published in BMC Medicine, researchers used data collected from general practices in The Health Improvement Network database in England, selecting 930,395 patients from 377 practices between the ages of 60 and 95 who did not have dementia to build the algorithm.

The algorithm is based on four variables as possible predictors of dementia: socio-demographic factors such as age and sex; health and lifestyle elements such as alcohol use and BMI; medical diagnoses; and prescription medication use. The variables were then checked against medical records of the selected patients for dementia diagnoses within a five-year follow-up period from the start of the THIN study.

The algorithm was then validated with another 264,224 patients from 95 other practices.

Researchers found that for people ages 60 to 79, the algorithm performed well, while it was less accurate for people between ages of 80 and 95. At a low threshold of risk for dementia, the algorithm had a sensitivity of 78 percent and specificity of 73 percent. But at risk thresholds of 2 percent or higher, the algorithm had a higher specificity of 85 percent but much lower sensitivity of 58 percent -- which researchers said is similar to previous prediction models based on cohort studies.

"Before this score is widely used we would recommend that it is independently tested in further populations of people, and that the ethical implications of using it in practice are considered," Walters said.

Related UPI Stories
Trending Stories