Machine Learning Algorithms to Identify Predictive Variables of Mortality Risk for Dementia Patients: A Longitudinal Cohort Study in Scientific Reports

Researchers have used machine learning algorithms to identify predictive variables for mortality risk following a diagnosis of dementia. Using a longitudinal cohort study of 28,023 patients, the algorithms were able to rank variables that traditional statistical methods may have overlooked, including age, BMI, MMSE score, and time from referral to initiation of work-up. The algorithms also identified multivariable models for predicting mortality risk, as well as time to death. The best performing algorithms used a combination of logistic regression, support vector machines, and neural networks, and consistently selected 20 variables as the most important for predicting mortality risk. These findings could help healthcare practitioners to more accurately predict the outcomes of patients with dementia.