A model of health: using data modelling techniques to improve health outcomes for older Canadian adults by optimizing the development and delivery of physical activity interventions

Older Canadian adults do not get enough physical activity, increasing their risk of developing numerous chronic diseases. The economic burden of these diseases exceeded $10 billion in Canada in 2015, much of which is avoidable if older adults were more active. Our research will quantify the relationship between various levels of physical activity, disease outcomes, and healthcare costs. We will use methods from epidemiology and marketing analytics to build statistical models that use individual-level data from the CLSA in order to predict health and cost outcomes of physical inactivity. We will show how these models can be used to optimize and target the delivery of physical activity interventions. Our research will be the first to predict health and cost outcomes for older adults based on individual-level data. This will allow healthcare decision makers to offer, and to make an economic case for, cost-beneficial physical activity interventions for older Canadian adults.

Nearly 90% of older Canadians have at least one chronic disease; 65% have two or more. In this project, business analytics techniques were used to predict the presence of an exemplar chronic disease, heart disease, among older Canadians, and to calculate the corresponding expected healthcare costs. Neural networks were used to develop logistic regression models of heart disease using demographic, lifestyle, and health information for 15,599 older adults from the Canadian Longitudinal Study on Aging. The Economic Burden of Illness in Canada provided healthcare cost data. The best model identified 65.8% of heart disease cases from 40% of participants with the highest predicted probabilities of heart disease, accounting for $2.7 million more expected annual healthcare costs than a randomly sampled 40%. Among all older Canadians, this difference would be $1.1 billion. These methods could assist healthcare decision makers to optimize the delivery of chronic disease prevention interventions.