Predictive risk modelling and homelessness

Predictive risk models (PRMs) are statistical models that exploit patterns in large historical data to generate a prediction that a new client might have for a specific event. There are two use cases: proactive and reactive. In their proactive use, the PRM identifies people likely to become homeless while the reactive use is to prioritize already homeless people based on their risk of adverse outcomes if unhoused. We summarize the existing literature and provide a case study of an application of PRM in the form of the Allegheny Housing Assessment Tool (AHA). We emphasize ethical guidelines and ongoing validation when adopting PRMs.