Introduction
Adolescent violent offending in England accounts for a notable proportion of offences committed by adolescents each year. There is a growing need to be able to identify adolescents at an increased risk of committing a first violent offence so that appropriate support can be targeted more effectively to help them overcome this risk. Routinely collected police data is administrative data which to date has been used predominately to consider geospatial risk and hyper-specific forms of violent offending (knife crime). It has not been identified, in the context of England, to have been used to explore associations between recorded exposures occurring during adolescence and first-time violent offending outcomes occurring during the same developmental phase.
Methods
A routinely collected police dataset provided by Thames Valley Police was used to model directional associations between adolescence-limited exposures and first-time violent offending outcomes during two phases of development, early adolescence (10-13-years) and middle adolescence (14-16-years). Inclusion in the dataset required data subjects be aged 16-years by the 25th March 2021. The dataset was scaled to a full cohort level using 2021 UK Census data. The modelling involved multivariable and ridge Logistic Regression analysis.
Results
Multivariable Logistic Regression analysis identified consistent statistically significant positive associations between first-time violence and physical victimisation, neglect and other welfare-related incidents, and non-violent incidents and offences. Multivariable Ridge Logistic Regression using multiway exposure interactions was used to develop a predictive risk tool with a high level of out-of-sample predictive accuracy, and good discriminatory ability for outcomes occurring in early adolescence and in middle adolescence (calibration slope = 1.031-1.036, calibration-in-the-large = 0.000-0.000, AUC = 0.706-0.720).
Conclusion
Routinely collected police data offers opportunities to explore and model individual-level factors for first-time violent offending outcomes in adolescents with a view to using such insights to provide timely support to mitigate such outcomes.
