Partnerships and Governance to Support the Appropriate Application of Predictive Analytic Models

A Prevention Science framework supports important organizational considerations for the application of Predictive Analytics to support practitioner and system decision making. Sustained effective and equitable use of Predictive Analytics to improve child, youth and family outcomes requires a comprehensive, ongoing and institutionalized approach for providing oversight regarding the development of analytic models, evaluating models that have been developed and monitoring the application of Predictive Analytics by practitioners. This paper will describe, using child protection prevention programming as an example, the need and theoretical basis for a continuous oversight and evaluation of Predictive Analytics relative to intervention decision-making.

This discussion will first address the need for collaboration among practitioners, researchers and policy makers throughout the development, evaluation, and application of predictive analytics. The commitment to developing equitable predictive analytic tools, given disparities represented in most system administrative data, as well as the potential for analytic models to be used to challenge present disparities, will be outlined. The importance of institutionalizing guidelines, aimed at ensuring the effective and equitable practice, for the development and application of predictive analytics will round out the discussion. Specific areas for Predictive Analytics guideline development include: 1. ensuring that research and application drive model development; 2. developing sound analytical and technical approaches, including a well-defined outcome, sufficient data, model validity and reliability; 3. Focusing on equity, and 4. Incorporating ongoing model and practice evaluation and improvement.

Final discussions will emphasize how the Prevention Science framework supports successful development, testing and implementation of Predictive Analytics in a practice setting. Prevention Science addresses the previously described needs by: 1. identifying the need for comprehensive, differential levels of treatment based on children’s needs and risks; 2. requiring a multi-partner collaboration and comprehensive, prevention-oriented approach to addressing identified community needs and dysfunctions; and 3. supporting system factors and changes needed such as a sustained learning environments, continuous research, and appropriate service planning and allocation.