Curtin University

Dataset description

Based on similar research in other countries, feasible outcomes can be generated from population sizes as low as 37 individuals, through to large longitudinal cohorts of more than 10,000. This project aims to recruit 5000 participants across Australia during each the first and second year of recruitment. With wide recruitment we aim to have a population-based capture of adults and children who have experienced a mTBI. Custom machine learning models and related algorithms (such as RandomForests, XGBoost, etc.) will be built to analyse this high-quality dataset to predict outcomes and identify better care pathways for people following mTBI. Model robustness and uncertainty will be tested using extensive sensitivity analyses (including probabilistic sensitivity analyses) and calibration to observed trends in outcome. Baseline characteristics will be compared between men and women using Chi-square tests for categorical variables and t-tests for continuous variables. The analyses will be repeated by age, mechanism of injury, outcome (PPCS or no PPCS), nature of injury (single event or sequelae), and symptoms at presentation. All eligible participants will be included in the analyses. Analysis of data not involving machine learning models will be performed using R, SPSS or Stata V17 statistical software (StataCorp LP, College Station, Tex).
Click to explore relationships graph