An interdisciplinary research team at the University of Alberta is developing machine learning programs using health-related, lifestyle, socioeconomic, and other data to improve the future physical and mental health of older Canadians.
Cao, principal investigator, associate professor of psychiatry, associate professor of computer science and Canadian Research Chair in Computational Psychiatry, said that this approach could one day be used to help medical teams provide personalized care and promote healthy aging.
"Machine learning is a powerful and useful computational method that can harness the abundance of de-identified data," he said. If we want to drive future personalization of patients** for certain health outcomes, we need to leverage machine learning. ”
Cao's team used machine learning to identify patterns and analyze them to benefit patients in two recently published studies from the Canadian Longitudinal Study on Aging (CLSA) involving more than 30,000 Canadians between the ages of 45 and 85 who would be followed for up to 25 years.
"This is top-notch data from Canada, and he is also the co-director of the Interfaculty Computational Psychiatry Group at American universities," Cao said. Our goal is to contribute to the health of Albertans and Canadians alike. We want to chart a healthy aging trajectory for each of us. ”
In the first article**, published in the journal Geriatrics, the team developed a biological age index by applying a machine learning model to blood test data from CLSA.
Just like the comparison of the health levels of 30-year-old Canadians and 60-year-old Swedes in the 70s of the 20th century, your body may be physiologically older or younger than your actual age. Researchers refer to this disparity as the "biological age gap."
They investigated the relationship between biological age gaps and lifestyle, environmental factors, and health status. They report that a positive biological age gap ("older than chronological age") is strongly associated with chronic disease, regular consumption of processed and red meat, smoking, and passive smoking.
Some modifiable factors, such as the consumption of fruits, legumes, and vegetables, are associated with a negative biological age gap ("younger" than chronological age).
"Understanding these associations and identifying risk factors for differential aging can guide effective public health recommendations to promote healthy longevity," the team reported in **. ”
Cao hopes that one day this approach will also affect the health care services that individuals receive. The next step in this research, he says, will try to understand which factors, or combinations of factors, are most important in influencing the biological aging process.
In a second study published in the Journal of Affective Disorders, the team developed a procedure that could pinpoint which people would experience depression over a three-year period.
The machine learning model is trained by working backwards using the records of individuals who are eventually diagnosed with depression. Participants who were previously diagnosed with depression or scored higher on the self-reported depression symptom scale were excluded.
"We identified existing subthreshold symptoms of depression, emotional instability, low life satisfaction, perceived health and social support, and nutritional risk as the most important factors for depressive episodes," the researchers said in the report. ”
Cao said the model was about 70 percent accurate in which study participants would develop generalized depression at the individual level within three years, and the model remained accurate when subthreshold depression symptoms disappeared.
"Interestingly, depression can still be ** even if only personality measures, perceived health or mental health, nutrition and other factors unrelated to depressive symptoms and stress are used," Cao said.
"Neither the mental health machine learning model nor the bioage model are imperfect at this stage to be implemented in the real world, but this is his goal, so more research and testing is planned," Cao said. We're trying to build a dialogue that includes different groups – clinicians, patients, and people with lived experience – to demonstrate that this model can benefit the public. ”