4-year PhD Studentship: Identifying children 10 or younger who are most at risk of abusive head trauma in the aftermath of the pandemic

Paediatric abusive head trauma (AHT) and traumatic brain injury (TBI) is when a child’s head is injured from being hit or shaken. AHT and TBI are uncommon but potentially catastrophic, leading to brain, visual and hearing problems, and about 1 in 5 cases results in death.[1] After the onset of Covid, the Great Ormond Street Hospital saw a 15-fold case increase one month into the March 2020 lockdown, compared to the same period in previous years.[2] Whilst the Covid situation is improving, this is still a dangerous time for vulnerable children as stressors for families continue, whether through rising unemployment, fuel crisis, gas price increase and others.

Identifying predictors of AHT and TBI can help local authorities and other agencies recognise children who are at high risk so that they and their families can be offered support. Understanding those at risk of AHT and TBI can also guide secondary prevention in that social care and health practitioners can more intensively monitor children in at risk groups and may help identify events promptly where primary prevention did not occur. Co-production of prediction models with relevant agencies will help the dissemination and application of this information in real world practice.

Aims and Objectives

The proposed project aims to determine the predictors of AHT and TBI in children under 10 years of age. The objectives are to:

1. Conduct a systematic review of existing prediction models for AHT and TBI in young children.
2. Determine predictors of AHT and TBI using routinely collected datasets and advanced prediction modelling and machine learning techniques
3. Co-produce a prediction model for England with local authority, social care, public health and emergency care practitioners
4. Validate the prediction model using the datasets from Scotland and Wales

Methodology

Literature will be searched for existing prediction models for AHT and TBI. The identified models will be described and critically appraised, and those found to be the most appropriate will be used to inform the development of our model and subsequent stages of the project.

For model development, we will use primary and secondary care records, mortality and census data from the Public Health Research Dataset (PHRD, available at no cost)[3] and multivariable logistic regression models. We will develop models for different age groups: under 1, pre-school and primary school. Potential predictors include child characteristics (socio-demographic, learning disability, primary and secondary care contacts), household (deprivation, household size, whether multi-generational, Covid exposure, key worker status) and area (urban-rural, region). We will use 10-fold cross-validation and compare uniform shrinkage, LASSO, elastic net, ridge regression, decision tree modelling and random forest to adjust for overfitting.

The prediction model will be co-produced with practitioners who will feedback on variables and any omissions from the model, and possibility of personalised predictions. This ensures consideration of care and health practice and maximise acceptability and usability. The final model will be externally validated using datasets from the Scottish National Safe Data Haven and the SAIL Databank (Wales).

Keywords

Abusive head trauma, traumatic brain injury, prediction modelling, co-production

How to apply for this project

This project will be based in Bristol Medical School - Population Health Sciences in the Faculty of Health Sciences at the University of Bristol.

Please visit the Faculty of Health Sciences website for details of how to apply