Barking & Dagenham works on ‘at risk’ algorithms
Borough’s Insight Hub team is building a predictive data model aimed at supporting people before they fall into crisis
The data science team at the London Borough of Barking & Dagenham is working on a predictive data model aimed at identifying people who could be close to the need of support from its social services department.
Pye Nyunt, founder of the Insight Hub at the council and an Agilisys consultant, outlined the plan at the annual conference of public sector IT association Socitm in Leicester yesterday.
He said the team is currently working on the learning algorithms and that there is no firm launch date, but that it hopes it will help to protect people with interventions before they reach a state of crisis. This could also help to ensure the council does not have to spend as much money in supporting some at risk people.
The effort is being made in response to the heavy proportion of the council’s budget taken by social care – a factor in virtually all local authorities – and the formation of an umbrella service for the sector named Community Solutions.
“It has looked at all the traditional services, taken some of the components such as access points, triages and interventions and put them into one service which has focused on customer fulfilment rather than departmental blocks,” Nyunt said.
“It’s focused on understanding why people need help rather than what help they need.”
“We were tasked with coming up with a predictive model using existing data to help the service to understand where to intervene before it’s too late and the individual becomes a heavy social and economic cost to the council.”
The model is using machine learning techniques to calculate proportionate levels of risk, allocated to individuals based on expert opinion.
He pointed out that a small number of people could become “persons in crisis” if the council does not intervene quickly, and the model is aimed at understanding who they are before they go into crisis.
“We are hoping we can get these predictive models right so we can help frontline staff do their job better, and ultimately through helping vulnerable residents become more self-sufficient the council will become less vulnerable to the decreasing budgets for services,” he said.
Nyunt pointed to one difficult issue facing the team in the form of multiple sets of data related to the same issue. This could come, for example, from somebody presenting themselves as homeless on a handful of occasions, and he said that at the moment it is a challenge to consolidate the data.
He also outlined a test for an organisation to establish whether or not it is driven by data. If people are using data they will begin a sentence with “I know”, whereas if not they will say “I think”. He pointed out there are likely to be a lot more of the latter than the former.
“If you want to know how data driven your organisation is, try this simple exercise,” he said. “I’m sure you will be surprised.”