Inflammatory bowel disease (IBD), including ulcerative colitis and Crohn’s disease, is a chronic inflammatory condition of the gastrointestinal system. Canada has one of the highest IBD rates globally, with 470,000 individuals (1 in 91) projected to have IBD by 2035. People with IBD have a shorter lifespan (up to 8 years less for females and 6 for males) and a higher mortality rate (17 versus 12/1000 person-years) compared to those without IBD. In IBD, multimorbidity (≥2 chronic conditions) is becoming more prevalent. It worsens health outcomes and complicates treatment. Machine learning models can predict premature mortality and identify trends that warrant further investigation. A recent study published in the Canada Medical Association Journal (CMAJ) investigated the premature mortality rates as well as the association between multimorbidities and premature death in Ontario’s IBD population.
This population-based retrospective study used administrative health data from Ontario, Canada. Individuals with IBD who died from January 1, 2010, to January 1, 2020, in Ontario were included. The study also included individuals with a history of 17 chronic conditions, including hypertension, chronic obstructive pulmonary disease, rheumatoid arthritis, asthma, dementia, diabetes, congestive heart failure, cancer, osteoporosis, renal failure, chronic coronary syndrome, myocardial infarction, cardiac arrhythmia, stroke, mood disorders, other mental disorders, and osteoarthritis or other forms of arthritis (non-rheumatoid). People with invalid health cards, non-Ontario residents, and missing information about their age and death date were excluded from this study. People’s deaths, ages, and sex were identified from the Registered Persons Database. Â
The primary endpoint was premature mortality. Statistical and machine learning analysis was developed in order to determine premature mortality using the age of the patient at diagnosis and the presence of 17 chronic illnesses. These models (n = 7) were assessed based on sensitivity, the area under the receiver operating curve (AUC), explainability plots, calibration plots, positive predictive value, accuracy, and F1 scores.
These study results were reported according to assessing the performance and potential risks of AI systems through evaluation (APPRAISE) artificial intelligence (AI) framework, 2024 transparent reporting of a multivariable prediction model for individual prognosis or diagnosis plus artificial intelligence (TRIPOD+AI) checklist, and reporting of studies conducted using observational routinely collected health data (RECORD) guidelines.
A total of 9278 decedents with IBD (median age of death = 76 (64–85) years and 49.3% female) were included in this cohort study. Around 47.2% of included participants were reported with premature deaths (55% male and 44% female). At death, conditions with the highest prevalence were non-rheumatoid arthritis with 76.8%, hypertension with 72.8%, mood disorders with 69%, renal failure with 49.6% and cancer with 46.1%.
The model that included age at diagnosis for each chronic condition developed at the age of ≤60 years performed the best with an AUC of 0.95 (95% confidence interval [CI] 0.94-0.96), although all models performed well with AUC 0.81-0.95 and calibration on testing data which included 1,856 samples. Key predictors of premature mortality included male sex and early-age diagnosis of hypertension, mood disorders, other mental health conditions, and osteoarthritis or other forms of arthritis.
This study’s limitations include the risk of misclassification bias using administrative health data, which is minimized by the validated algorithm method. This study did not include the population-level risk prediction tool. Causality required alternative analytical methods.
This study concluded that the early onset (≤60 years) of chronic conditions significantly influences health trajectories, as demonstrated by machine learning models estimating their impact on premature mortality in IBD patients. Future research is necessary for care models that provide access to high-quality and interdisciplinary medical treatment for individuals with IBD.
Reference: Postill G, Itanyi IU, Kuenzig ME, et al. Machine learning prediction of premature death from multimorbidity among people with inflammatory bowel disease: a population-based retrospective cohort study. CMAJ. 2025;197(11):E286-297. doi:10.1503/cmaj.241117
