Thousands of people who experience homelessness turn to their councils for help every day. But what if these residents, who are already in such dire straits, could be found and supported before hitting such crisis points in their lives? That would seem to require a futuristic power of prediction.

But this is exactly what four councils in England are doing as part of the novel Test and Learn project, Using Data to Prevent Homelessness, one of eight being delivered and evaluated by the Centre for Homelessness Impact and funded by the Ministry of Housing, Communities, and Local Government. Test Valley, Newham, Stockport, and Barking & Dagenham councils are using a data analysis platform that includes tools that use artificial intelligence to predict which of their residents are at most risk of homelessness in the future.

Sarah Merriam is leading the project at the London Borough of Barking & Dagenham, one of England's most deprived authorities. Currently homelessness prevention teams assist residents when they are close to crisis point. This predictive model aims to identify households which are at a higher risk of becoming homeless in the medium term: up to nine months’ time.“This step into the field of advanced analytics and AI will allow us to offer support to households before they come to the attention of our homelessness prevention team.” 

Using existing data on people who have experienced homelessness in the past enables the model to pinpoint risk factors for homelessness and subsequently help identify who is at risk from today's population. 

Data confidentiality is key in this project. The data platform uses an ‘Information Governance (IG)’ bridge to safeguard council data. This digitally “splits off” any identifiable information in the data and records such as ages, addresses, and names. The bridge attaches pseudonyms to each record before passing it across to the supplier Xantura’s servers. The predictive model has been developed using this pseudonymised dataset and excluded any information on protected characteristics.

The identifiable data is then re-attached to the pseudonymised records on the council side, at the point when individual records are accessed by council officers working with those residents.

The project is being run across four authorities for good reason. It is anticipated that the “predictors of homelessness” which the platform identifies will be different in London boroughs like Barking & Dagenham from those in more rural authorities such as Test Valley, Hampshire.

Like some of the other Test and Learn projects, Using Data to Prevent Homelessness is being evaluated with a randomised controlled trial, a “gold standard” method more common to medicine than public policy. Over the course of this three-month trial the platform is expected to generate lists of around 500 people at risk of homelessness for each of the participating authorities. These lists are sent to Verian Group, the research firm appointed to carry out the evaluation, which splits the list into “control” and “intervention” groups of equal size. The intervention half of each list is dispatched to the four local authorities.

In Barking & Dagenham, people in the intervention group are proactively contacted by “community navigators”, council officers trained to have holistic conversations with residents about their strengths and needs., Ms Merriam says, ”People aren’t used to the council working in this way, but outreach we have done in the past has been received really positively by our community. Ultimately it’s about offering people some extra support and most people appreciate that” She expects most of the people they contact will be in financial difficulties. Others may be young people who have recently left the care system or residents living alone with little support from family or friends.

Residents in the “control” group can also access their councils’ usual support and homelessness services if they approach them for help. But they won’t be contacted and offered proactive support during the trial period. By comparing outcomes of residents in the control with those in the treatment group, the evaluators hope to work out the impact of stepping in months early thanks to the predictive power of the platform.

Ms Merriam says the project fits well with Barking & Dagenham’s own shift from a “reactive to preventative” approach and hopes its participation will help others follow its lead. “There’s a broad consensus across local government that we need to do prevention in lots of different areas because that achieves better outcomes for residents and saves money,” she says. “It’s not very easy to do but it’s better for councils and it's better for residents,” Ms Merriam adds. “If we can build the evidence with a randomised controlled trial which suggests this approach is effective, it can then support the scaling up of this kind of approach and help prevent homelessness for many households.”

‍Keith Cooper is a freelance journalist