A research team at the University of Alberta is using machine learning to predict Canadian’s future mental and physical health as they age. 

Cloud Cao, an associate professor in the U of A’s department of psychiatry, is the principal investigator on this research team. He is also a Canada Research Chair in Computational Psychiatry and an adjunct professor in computing science.

This research aims to use machine learning to “make predictions [about] health conditions, as well as outcomes.” Cao hopes these predictions can help doctors make better decisions about patient care. 

Two recent studies show that machine learning could give insight to predicting future health 

Cao’s team recently published two studies involving predictions about the future health of Canadians. The first study focuses on comparisons between a person’s age and a biological age index developed by Cao and his team. 

The biological age index uses blood markers to identify a person’s BioAge. This BioAge is then compared to their chronological age. For example, poor lifestyle choices such as smoking can lead to a positive BioAge. A person could have a BioAge of 70 when they are actually only 60, leading to significant health challenges. 

On the contrary, healthy lifestyle choices such as exercising could result in a negative BioAge. Using the same example, if a person is 60 but their BioAge is 50, they have a negative BioAge. 

The study had two purposes: to create the biological age index and identify the associated factors that influence a positive or negative BioAge. Cao wants to determine which factors are most significant when compared to other factors. 

“We try to incorporate as many variables [as possible],” Cao said. These included “lifestyle, social economics, and cognitive [function].”

The second study that Cao’s team conducted focused on predicting whether or not people would experience an onset of depression within three years. This study suggests that there is potential for predicting other mental health conditions for aging Canadians. 

For this study, the research team began by collecting baseline data, such as personality measures and perceived health. Then they conducted a follow-up. The researchers tried to determine whether or not we can use “baseline data to [predict] future depression onset.”

The model developed by Cao’s team was roughly 70 per cent accurate in predicting which participants would develop depression within three years, based solely on the baseline data. 

“Once we have the prototype of such models, can we actually use them? Can we bring them beyond the research domain?” Cao says 

According to Cao, machine learning for predicting future health is still far from being implemented in practice throughout Canada. 

“The ultimate goal is to use this data … to try and make predictions about our health and our aging status so we can try to reflect the best in us,” Cao said. 

Cao hopes to “improve the models using [more] data, a larger population, [and] more factors” in the next three to five years. He believes that this will help improve the accuracy of these models and their predictions. 

“Once we have the prototype of such models, can we actually use them? Can we bring them beyond the research domain? That’s one of the major goals for my group.”