Machine learning programs developed by a research team from the University of Alberta are being used to predict the future health of aging Canadians. The team, led by principal investigator Bo Cao, is utilizing various data including health-related, lifestyle, socio-economic, and other factors to develop these programs. The goal is to provide individualized care and promote healthy aging.
Cao emphasizes the power and utility of machine learning in utilizing rich de-identified data to push for individualized patient prediction for certain health outcomes. The team used machine learning in two recently published studies using data from the Canadian Longitudinal Study on Aging (CLSA) for more than 30,000 Canadians aged 45 to 85. The aim of this research is to contribute to the health of Albertans and Canadians by developing a healthy aging trajectory for every individual.
In the first study, the team developed a biological age index by applying machine learning models to blood test data from the CLSA. This index helps determine the physiological age of an individual and identifies the difference between their biological age and chronological age, known as the BioAge gap. The research revealed strong associations between a positive BioAge gap (physiologically older than chronological age) and chronic illness, frequent consumption of processed and red meat, smoking, and passive exposure to smoke. On the other hand, a negative BioAge gap (physiologically younger than chronological age) was associated with the consumption of fruits, legumes, and vegetables. These findings can guide effective public health recommendations for promoting healthy longevity.
In the second study, the team developed a program that accurately predicted the onset of depression within three years for individuals. By working backwards and training the machine learning model using records of individuals who were eventually diagnosed with depression, the researchers identified several important predictors such as subthreshold depression symptoms, emotional instability, low levels of life satisfaction, perceived health and social support, and nutrition risk. The model demonstrated about 70% accuracy in predicting the development of full-blown depression within three years at the individual level, even when excluding subthreshold depression symptoms. This indicates the potential of machine learning in predicting depression based on factors not directly related to depressive symptoms and stress.
While these machine learning models are not yet refined enough for real-world implementation, Cao and his team aim to continue further research and testing. Their goal is to refine the models to a point where they can have a positive impact on the healthcare individuals receive. By involving various groups such as clinicians, patients, and individuals with lived experience, they hope to demonstrate the benefits to the general public.
The research for these studies was funded by various organizations including the Canada Research Chairs program, Alberta Innovates, Mental Health Foundation, Mitacs Accelerate program, and the Canadian Longitudinal Study on Aging. Bo Cao is affiliated with several research institutes and networks at the University of Alberta.
In summary, the University of Alberta’s research team is utilizing machine learning programs to predict the future health of aging Canadians. These programs have shown promise in determining an individual’s biological age, identifying factors associated with healthy aging, and predicting the onset of depression. Further research and testing are planned to refine these models and potentially impact the healthcare provided to individuals.
