Economist's maths aim to forecast which children will be abused

Associate Professor Rhema Vaithianathan led an Auckland University study using past data to predict child abuse. Photo / Sarah Ivey

Accuracy rate close to that of mammograms

An economist who has developed a mathematical method of predicting child abuse says it is about as accurate as a mammogram is in spotting breast cancer.

Auckland University associate professor Rhema Vaithianathan, who led a team that developed the technique for the Government's white paper on vulnerable children, hopes it can be introduced through the country next year.

Sri Lanka-born Dr Vaithianathan, 46, who came to New Zealand at the age of 7, has also written papers applying economic techniques to the "marriage market" for female genital mutilation in Africa and to the likelihood of patients who are admitted to New Zealand hospitals being readmitted in the following year.

The three Auckland health boards are using her hospital formula to tell family doctors how likely their patients are to be readmitted.

Professor Vaithianathan contested the Labour nomination for Parliament in the 2009 Mt Albert byelection, won by David Shearer, but she said she was happy to do research for the National Government.

"I think New Zealand is too small, and our problems are too grave, for partisanship."

Her team found that children born between January 2003 and June 2006 whose parents were on welfare at some point in their first two years accounted for 83 per cent of total New Zealand substantiated findings of abuse or neglect of children under 5 during the subsequent five years.

The rate of substantiated abuse or neglect in children on welfare (13 per cent) was almost 10 times the 1.4 per cent rate in families that were never on welfare in those five years.

However, the formula, based on 132 factors such as parents' ages and their own history of being abused as children, cannot predict abuse with certainty.

Almost two-thirds (63 per cent) of the children who scored in the top fifth for being most likely to be maltreated did not suffer substantiated maltreatment over the next five years.

Conversely, 56 per cent of the children on benefits who were maltreated were not in the top fifth of children who were predicted to be maltreated.

"This is similar to the predictive strength of mammograms for detecting breast cancer," the team found.

But Professor David Fergusson, who heads a study that has tracked children born in Christchurch in 1977 for the past 35 years, warned against rushing into using the method without careful testing.

"How do you practically implement the service in a way that is non-stigmatising and doesn't end up with social workers labelling people as child abusers?" Professor Fergusson asked.

"The appropriate way of doing this would be a small pilot study of perhaps 300 or 400 beneficiary families in one region, see if you could identify them [those who will abuse their children], then experiment with methods of selling the idea.

"Then you can widen the thing, and then conduct a randomised trial, in which you take a much larger group in which half of them get the programme and half of them don't, and see if that had any impact on child abuse statistics."