Unlocking data to inform public health policy and practice

Background: Responses to crimes like domestic abuse cannot be separated from the social context in which they are embedded and responses need a multi-agency preventative approach and a holistic understanding of risk factors and predictors in order to develop early preventative measures. Data sharing at the individual and local level is part of the current practice.

However, arguably this can create a postcode lottery and inequality in response. This work aims to facilitate sharing of national datasets to improve recognition of clustered risk factors and patterns which predict future problems and to improve the ability to evaluate the success of programmes and interventions (both local and national) not just in terms of reoffending but also in terms of substance abuse, mental health problems, parenting and wider household health.

Aims and objectives: To facilitate a multi-agency partnership approach in data sharing of national datasets.

Methods: This work addresses four known barriers to national dataset sharing with police data. WP1: Data collected is often in report form, and although rich data, it is text and so can not easily be coded or summarised and extracted for sharing. We will examine if text mining can be used to automatically extract information from reports such as information on; other agencies already involved, referral to MARAC (multi-agency risk assessment conference), police action taken.

This work will focus predominately on information extraction methods from natural language processing. WP2. Different software systems are used within and between police forces, the software systems are quite often structured differently so not possible to harmonise easily.

This work will undertake a scoping study of the different systems used across Wales, what data is collected, quality of data, similarities and differences in the systems. The scoping study will examine potential for a minimum core dataset created from combining existing datasets and separately the possibility of some forces migrating to the most commonly used system (e.g. building on the NICHE collaborative which already exists).

WP3. The police data has not been combined with other national datasets such as education, justice/courts data, health (GP/hospital/A&E), social services before and so it is hard to envisage the potential advantages and easier to see the problems and risks. This work will undertake an exemplar case study to demonstrate the potential of national data sharing.

It will bring together the data already linked in Wales in SAIL (GP, hospital, A&E, education, maternity indicators, national community child health dataset, substance misuse, family courts etc) with Public Protection Notification data from the Police dataset, to profile risk factors for domestic violence. WP4. Identifiers in different agencies datasets are different, for example the pupil ID number in Education data, NHS number in health data and individual police identifier are not compatible for linking across datasets, this work will examine if datasets can be linked using other identifiable data captured in the datasets.

Timelines for delivery: April- June: Acquisition of data, defining terms for WP2, defining variables for WP4. July-Sep: analysis, scoping of datasets Oct-Dec: finalise reports, multi-agency workshops to discuss findings and next steps. Anticipated impact and dissemination: evidence base of potential ways forward, e.g. core minimum dataset