Introduction
Inflammatory bowel disease (IBD) is a chronic, relapsing, and remitting condition that primarily affects the gastrointestinal tract. Canada has amongst the highest rates of IBD in the world. Efforts are needed to improve the care of IBD – particularly the multidisciplinary care provided, given the impact of IBD on quantity and quality of life. Premature mortality (death <75 years) is a critical indicator of overall population health, socioeconomic inequalities, and health system performance. Multimorbidity is also an emerging topic in, and indicative of, population health, given its association with more complex care needs, poorer health outcomes, and excess mortality.
Researchers at the Population Health Analytics Laboratory sought to explore whether and how non-IBD chronic conditions predicted whether mortality was premature. This recognize the complex interplay of non-IBD chronic conditions on premature mortality, and ultimately, supports identification of potential improvements to multidisciplinary care for those with IBD.
How Machine Larning was used as a Research Tool
With IBD, the relationship between individual non-IBD chronic conditions (as well as the age conditions are developed) and premature mortality is not known. It is also unclear how these conditions interact with each other. Traditional statistical models require pre-specification of the relationship between input and output variables. As such, our prediction challenge was ripe for machine learning (ML) models, as ML can flexibly learn the relationship between input and output variables without pre-specification. Our research team applied ML techniques (random forest and XGboost) to establish whether and how non-IBD chronic conditions (and the age at they are diagnosed) can predict premature mortality.
Impact on Population Health and Clinical Medicine
We showed that conditions developed early in life (age ≤ 60 yr) and their age of onset were important for predicting their health trajectory. Importantly, our models did not use variables directly related to IBD, such as subtype, severity, and medications, demonstrating the importance of non-IBD chronic conditions to the overall health and life expectancy of those with IBD. Clinically, our findings emphasize the need for models of care that ensure people with IBD have access to high-quality, multidisciplinary health care.
Read the full publication in CMAJ here.