A groundbreaking Canadian study suggests that machine learning could soon help predict premature death in people with inflammatory bowel disease (IBD) — offering new hope for better care and early intervention.

IBD and the Burden of Comorbidities

IBD, which includes Crohn’s disease and ulcerative colitis, affects hundreds of thousands of people in Canada and millions globally. But the story doesn’t end with gastrointestinal symptoms. People with IBD often develop other chronic conditions — a situation known as multimorbidity — which can complicate treatment and worsen long-term outcomes.

In fact, nearly half of all deaths among IBD patients in Ontario between 2010 and 2020 occurred before the age of 75, a threshold commonly used to define premature mortality.

Using AI to Uncover Risk Patterns

Researchers from ICES and the University of Toronto tapped into Ontario’s vast health databases to analyze 9,278 IBD patients who had passed away. They trained several machine learning models to determine whether the presence — and age of onset — of other chronic diseases could predict early death.

The models performed impressively, with their best version achieving an AUC score of 0.95 — a near-excellent rating.

So what stood out?

• Mental health conditions, including mood disorders, were top predictors.

• Osteoarthritis and hypertension also played a significant role.

• Most critically, the age of diagnosis mattered. Conditions appearing before age 60 were far more predictive of premature death.

A Call for Better Multidisciplinary Care

This study doesn’t just offer a data science breakthrough — it underscores a serious healthcare gap. While IBD care often focuses on gastroenterology, the findings suggest that early detection and management of other chronic illnesses (especially mental health and cardiovascular issues) could be lifesaving.

Multidisciplinary care isn’t just a buzzword anymore; it’s a data-backed imperative.

What’s Next?

While these AI models aren’t ready for bedside use just yet, they show promise for future clinical decision-making. As researchers continue to refine the tools — and investigate causal relationships — health systems can begin planning how to proactively identify and support high-risk patients earlier in life.

Source: Gemma Postill, Ijeoma Uchenna Itanyi, M. Ellen Kuenzig, Furong Tang, Vinyas Harish, Emma Buajitti, Laura Rosella, and Eric I. Benchimol.
https://www.cmaj.ca/content/197/11/E286