Vulnerable Children Predictive Modelling

Predictive modelling is the use of automated tools to help identify people at risk early enough to allow for effective intervention.  It relies on the combination of historical data around individuals and the application of modelling techniques to that data to help understand potential need for services. 

The Government’s White Paper for Vulnerable Children noted that predictive modelling appeared promising based on preliminary research, but was at that time untried in the context of child maltreatment, carried ethical risks, and warranted careful, staged, development. 

As a result a suite of three reports regarding the feasibility and ethics of applying Predictive Modelling to the prevention of child maltreatment have been undertaken to inform decisions about advancing Predictive Modelling to trial. The Feasibility Study and Ethical Reviews have been subjected to an extensive international and national expert peer review. 

The papers, reports and studies have been canvassed throughout the professional community nationwide, being presented at Expert forums, Expert Advisory Groups and our Māori Reference Group over the last two years. 

The conclusion is that while the application of Predictive Modelling to child maltreatment raises some significant ethical concerns, those concerns can either be significantly mitigated by appropriate implementation strategies or are plausibly outweighed by the potential benefits of such modelling.  

As a result Predictive Modelling will be carefully tested to assess whether it can enhance decision-making at intake for children who are reported to Child Youth and Family because of concerns about abuse or neglect.  The aim is to support, not replace, professional judgement.  

Any testing will take place in a simulated intake setting, using historical case data before any decisions are made about trialling in a real-life setting. 

This reflects the care with which Government and the Ministry is advancing on this front.

Vulnerable Children Predictive Modelling Papers

The feasibility of using predictive risk modelling to identify new-born children who are high priority for preventive services (and companion technical report) - Prepared by Insights MSD, Ministry of Social Development 

Ethics approval for the study was granted in November 2012 and affirmed by the National Ethics Advisory Committee in February 2013. Oversight was provided by He Korowai Tamariki, the Advisory Expert Group on Information Security (AEGIS).  The study has been peer reviewed by international and local experts.

Predictive risk modelling and child maltreatment: An ethical review - Prepared by Associate Professor Tim Dare, University of Auckland

This paper reports on an ethical review of the use of predictive modelling tools in child maltreatment prevention efforts. Associate Professor Tim Dare is a recent Chair of the Health Research Council Ethics Committee. The ethical review has been peer reviewed by international experts.

Ethical issues for Māori in predictive risk modelling to identify new-born children who are at high risk of future maltreatment

Prepared by Anton Blank (Mana Ririki Trust), Dr Fiona Cram (Katoa Ltd), Associate Professor Tim Dare (University of Auckland), Dr Irene De Haan (University of Auckland), Dr Barry Smith (Lakes DHB and current Health Research Council Ethics Committee Chair), and Professor Rhema Vaithianathan (currently Auckland University of Technology).

This paper reports on a review of ethical issues for Māori in the use of predictive risk modelling tools for child maltreatment, undertaken by a group brought together by the University of Auckland.  The review has been peer reviewed by international and local experts.

Privacy Impact Assessment for the research – Prepared by Insights MSD, Ministry of Social Development 

This paper sets out the Privacy Impact Assessment prepared for the programme of research that included the feasibility study.  The Privacy Impact Assessment was discussed with the Office of the Privacy Commissioner, the Central Region Health and Disability Ethics Committee and the National Ethics Advisory Committee.  

MSD has worked with Statistics New Zealand to make the study data available for other approved research through the Statistics New Zealand Data Lab, as recommended by the National Ethics Advisory Committee, and has agreed a new deletion date for the study data held within MSD (this is now scheduled for December 2015).

Vulnerable Children Predictive Modelling: Design for Testing and Trialling

The care and protection of vulnerable children is one of the Ministry of Social Development’s key accountabilities, and is a core government priority.  

MSD is exploring Predictive Modelling as a potential tool to identify vulnerable children and ensure the right services and supports are being provided to them.  We can do better, and have the information to do so.

MSD has sought to generate public discussion and has worked collaboratively with the research community and other government agencies about the use of such tools in better identifying children at risk.

The 2012 Ministerial White Paper for Vulnerable Children had set out the Government’s plan to improve outcomes for New Zealand’s most at-risk children, with a focus on identifying those at-risk children in order to provide them with the protection and support they need.

The White Paper observed that, based on preliminary research, Predictive Modelling appeared to be a promising approach to identify vulnerable children. But the White Paper said more work needed to be done to establish whether Predictive Modelling could feasibly and ethically be used to reduce the risk of child abuse and neglect, and to improve outcomes for children and young people.

The Predictive Modelling which the White Paper discussed was commissioned by MSD and developed specifically for children by the University of Auckland. The researchers who developed this model said it had a ‘fair, approaching good’ accuracy in predicting whether individual children would, in future, be the subject of maltreatment, and they recommended further work around extending the modelling. 

The November 2014 MSD report was a response to that call for further work, and set out options to further test the proposed Predictive Modelling model.

To support the government’s priorities the November 2014 report proposed an ‘observational study’ as an option for testing whether Predictive Modelling would accurately identify those at-risk children.

It should be noted that the ‘Prospective Observational Study’ proposed in this report will not now take place.

The observational study was only one of a number of options which were under consideration in the development of the Predictive Modelling tool, to see if it could be used safely and ethically. 

Had it received Ministerial and ethical approval the observational study would have seen all children born in 2015 being rated for risk, using the Predictive Modelling approach. Data collected over the next two years would have been analysed, to see if the model had accurately identified those at high risk, based on the incidence of recorded maltreatment.

This study, had it proceeded, would not have displaced standard response.  Agencies such as Child Youth and Family and the Police would have, at all times, continued to act on notifications in relation to vulnerable children, as they do now. All children would have received the full range of support available from agencies.

However, as noted, the study is not proceeding. 

Current Approach  

We continue to do work in this area to understand the potential of a risk profiling approach, with the requirement that it remains consistent with concerns in relation to ethical and privacy concerns.

This work will, however, be based on historical anonymised case histories. 

It is currently envisaged this work will be complete in early 2016.