Utilizing Topological Clustering on a Traumatic Brain Injury Cohort: The Association of Neighborhood Socioeconomic Deprivation Profiles with Injury Mortality

Nelofar Kureshi, David B. Clarke, Syed Sibte Raza Abidi

Abstract: Traumatic brain injury (TBI) is a significant contributor to global injury burden and a leading cause of injury mortality. While there is an established correlation between neighborhood socioeconomic deprivation and injury incidence, there is conflicting evidence of an association between neighborhood deprivation and injury mortality. We studied the association between neighborhood deprivation and key individual-level covariates with mortality in a provincial cohort of TBI patients. The primary study objective was to segment the population and explore differences in neighborhood deprivation among TBI patient clusters. A secondary objective was to determine the extent to which patient clusters were associated with injury mortality. TBI patients from 2014-2020 were sourced from the provincial trauma registry. Features included key individual-level factors as well as linked neighborhood deprivation. Topological Data Analysis (TDA) was used to identify the complex interactions between features, with respect to the underlying shape of the dataset, to generate TBI patient clusters to predict mortality. Cluster boundaries were defined using Louvain community detection. Differences between features were tested across clusters. There were 1922 patients included in the analysis. TDA was conducted using the Mapper algorithm with L2-norm projection and Isolation Forest as a 2-dimensional lens. Four distinct clusters were identified which showed differences across all features, indicating that this segmentation method effectively distinguished between patients with varying levels of neighborhood deprivation and mortality risks.