Data science

Making youth care costs predictable with data science

July 8, 2021 - 4 minutes reading time
Article by Anp Expert Support

By means of data science, ICT advisor Kees van den Tempel helps municipalities gain more insight into the costs of youth care. By analyzing data, it is not only possible to predict costs, but also to take a more targeted approach to prevention. "When the costs of youth care are clear, you can influence them too."

Independent data scientist Kees van den Tempel works together with municipalities to provide insight into the costs of youth care. In order to be able to analyze the available data, he enlisted the help of Centric. According to Van den Tempel, an important first step has been taken, but there is still a lot of progress to be made.

How big is the problem surrounding the costs of youth care for municipalities?

"It is a huge expense for municipalities. A total of 5.5 billion euros goes to youth care every year. But of that amount there is a deficit of no less than 1.6 billion. A large part of the municipalities in the Netherlands does not have a balanced budget, partly because of the costs of youth care. And these are so-called open-ended arrangements.

It works like this: as a young person you have a problem. Then you go to the doctor or the center for Youth and Family of the municipality, and there they decide that you need something. This can vary from a dyslexia test to severe crisis admissions. When a specialist says: 'You have to get treatment', the municipality has to pay for it. One way or the other, it has to be paid for."

Previously, the budget for youth care was based on the costs in previous years, and in the end that was never correct

‐ Kees van den Tempel

And data science provides a better picture of that cost flow?

"It is important for municipalities to gain insight into how much costs they can expect in the field of youth care. Data science offers the possibility to predict this. Previously, such a budget was based on the costs in previous years, and in the end that was never correct.

One of the big questions that we ask ourselves via data science is: can you predict whether a specific child will end up in youth care on the basis of family characteristics in a child who is now 8 or 9 years old? By analyzing the data we receive from municipalities, we can make accurate estimates."

What do you look for in a prediction like that?

"The composition of a family can say a lot. Youth care problems are often linked to debt problems. If you grow up in a family with poverty, or if you have a father who drinks, you are more likely to get into trouble. So what you want as a municipality is to set up policy accordingly and then predict what the effect will be on youth care."

How accurate are those estimates?

"Right now, those predictions are about 80 percent accurate. That is high, but not high enough: I believe that with the right data we can predict with 95 percent certainty.

The problem lies with the datasets of the municipalities; they are polluted. For example, within municipal software systems it is possible to enter 1899 as the year of birth. Or you can enter zip codes that are not correct at all. That is why I turned to Centric to remove the noise and errors from the data. If we ensure that the data from municipalities is pure, the precision of our predictions can easily increase."

How privacy sensitive are those datasets?

"Thanks to the GDPR legislation, the privacy of young people in the municipalities is always preserved. Otherwise I would not have started working on the project in the first place - after all, it often concerns severe cases: young people with depression, eating disorders, difficult home situations. So we have no social security numbers, no names or addresses. The data is protected in such a way that it can never be traced back to who it is, the data we analyze is completely anonymous."

Most bills are under a thousand euros, but if there are a lot of them, you still lose a lot of money

‐ Kees van den Tempel

What did analyzing that data yield?

"Everyone knows that the serious cases, such as crisis care or long-term depression, cost a lot of money. So that's often the focus. Yet the bulk of the money is spent on very small interventions: dyslexia tests, speech therapy, that sort of thing. Most bills are under a thousand euros, but if you have a lot of them, you will still lose a lot of money. Municipalities often have no control over such flows. But it is precisely those small interventions that you can relatively easily implement policy."

So in the end you not only visualize the costs via data science, but data science can also help municipalities to improve their policies.

"Certainly. It is of course very important for municipalities to map out this cost flow, with a view to balancing the budgets.

But what is also interesting for municipalities is prevention. When the influx to youth care is clear via our data, you can influence it. In this way you as a municipality can ensure that there are excellent facilities for young people. Think of setting up a community center, appointing youth workers, ensuring that sports associations can run smoothly. Especially those kinds of facilities, which offer solutions for light cases, are relatively easy to set up. And when we see that those light cases take care of most of the costs, this can save a municipality a lot of money."

Related articles
Why data science goes beyond Business Intelligence, but can't do without it
Data science Retail Finance Public Logistic
Business Intelligence analysts have been working with data for years. Since the rise of data science, org ...
Data science: get over your threshold fears
Data science Retail Finance Public Logistic
It is sometimes referred to as one of the pillars of the 'Fifth Industrial Revolution': data science. Wit ...
Data-oriented working in nine steps
Data science Retail Finance Public Logistic
The rise of data science offers an enormous amount of opportunities to map and solve problems within your ...