Prediction model Youth Aid

Municipality of The Hague

The Youth Aid prediction model predicts the number of unique young people with Youth Aid without a stay with a six-year prediction horizon. Besides the number of young people, predictions are also made on costs. Predictions are made at district level, for city districts and for all of The Hague.

Last change on 5th of July 2024, at 11:06 (CET) | Publication Standard 1.0

• Publication category: Other algorithms
• Impact assessment: Field not filled in.
• Status: In use

General information

Theme

Health and Healthcare

Begin date

Field not filled in.

Contact information

[e-mail]

Responsible use

Goal and impact

The algorithm was developed because the Hague municipality wants to be better prepared for the future. If the municipality knows how many young people are expected to use Youth Care without Residence in the next six years and the costs involved, it can better anticipate this demand and continue to provide support to all young people who need it in the future.

The Youth Prediction Model is used in addition to staff knowledge and skills. The tool is mainly intended for policy and making and substantiating longer-term policy choices.

Considerations

The insights from the Youth Forecasting Model support policy and implementation in answering tactical and strategic questions. These are questions and topics that are relevant in the longer term (e.g. 5 to 10 years). The model is not intended to answer operational questions (short-term issues) and does not involve automated decisions.

Human intervention

Yes, there is no automated system. The insights from the prediction model can only be used through human intervention and with the context knowledge and experience of staff involved. If it is decided that the model is no longer usable, this decision will be taken by the model owner (portfolio director of Youth Care) and those directly involved will be informed.

Risk management

The risks of the algorithm were identified in advance and during the construction of the prediction model. Because the prediction model does not make statements about individuals but about the use of facilities in neighbourhoods, there was no risk that the results from the model could violate the privacy of specific individuals. Furthermore, we only worked with neighbourhoods that were sufficiently large (more than 100 inhabitants) and where a sufficient number of people were using Youth Services (more than 70 users). One reason is that this does not allow for disclosure. That is, by combining characteristics, we cannot identify who the potential users of Youth Aid services are. Furthermore, when building the model, we weighed up together with the privacy officer and the ethical officer which variables could and could not be included from a privacy perspective. Here, the AVG was always the starting point. In addition, explainability of the model and the results is a relevant and decisive criterion.

Legal basis

The municipality is responsible for implementing the Youth Act and the model allows the municipality to better anticipate expected developments and better support citizens.

Elaboration on impact assessments

When building the model, the privacy officer and the ethical officer weighed up which variables could and could not be included from a privacy perspective. Here, the AVG always formed the starting point. See further under Risk management.

Operations

Data

Only open data was used, namely the file Kerncijfers Wijken en Buurten CBS and the file 'Young people with youth care and youth care pathways in kind; neighbourhoods' from CBS.

Technical design

The model is a regression model. The performance of this model is measured by the measures RMSE, MAE and MAPE. These are measures that variously determine how large the prediction error is.