Statistical model based on driving data could predict mild cognitive impairment and dementia risk in older people

Researchers used data from the LongROAD project, in which driving trajectories were measured by in-vehicle recording devices for up to 44 months, to develop a statistical model for the prediction of mild cognitive impairment (MCI) and dementia risk in older adults.
Data from 2,977 participants without cognitive impairment at baseline were accumulated to produce 31 variables representing temporal changes in driving parameters, and the Influence Score method was applied to select influential variables for use in the prediction model.
This model had greater accuracy for predicting MCI and dementia risk than random forest and logistic regression methods, at 96% versus 93% and 88%, respectively. The most important driving variables associated with risk prediction were number of hard braking events and right-to-left turn ratio.
The study authors conclude that incorporating the Influence Score method into machine-learning models could improve the ability of statistical methods to predict the likelihood of MCI and dementia development in older drivers.