Predicting youth care between 0 to 4 years of age: a 2015–2019 Dutch population register data study

Schreuder, A.; Klaveren, D. van; Nieuwenhuijzen, M. van; Kraaij, W.; Houweling, T.A.J.

doi:doi:10.1016/j.childyouth.2025.108600

Purpose: Youth care services support families facing problems with raising children. Use of such services may be avoided if preventative support were offered. We developed youth care risk prediction models to enable risk stratification approaches. Methods: We used Dutch registry data of births between 2015-2019, including neighbourhood characteristics, socioeconomic status, parental health and behaviours, past birth outcomes, current birth outcomes, and household characteristics. The primary outcome was use of youth care services between the ages of 0-4 years. Multivariable Cox regression models were derived for assessment moments one year before birth, at birth, one year after birth, and two years after birth (models 1-4, respectively). Results: The full cohort consisted of 776,559 Dutch births, of which 32,290 underwent at least one youth care trajectory (4.2 %). Each full model performed equivalently to the respective parsimonious model. Parsimonious model 1 achieved an area...Show morePurpose: Youth care services support families facing problems with raising children. Use of such services may be avoided if preventative support were offered. We developed youth care risk prediction models to enable risk stratification approaches. Methods: We used Dutch registry data of births between 2015-2019, including neighbourhood characteristics, socioeconomic status, parental health and behaviours, past birth outcomes, current birth outcomes, and household characteristics. The primary outcome was use of youth care services between the ages of 0-4 years. Multivariable Cox regression models were derived for assessment moments one year before birth, at birth, one year after birth, and two years after birth (models 1-4, respectively). Results: The full cohort consisted of 776,559 Dutch births, of which 32,290 underwent at least one youth care trajectory (4.2 %). Each full model performed equivalently to the respective parsimonious model. Parsimonious model 1 achieved an area under the receiver operating characteristic curve (AUC) of 0.760 (95 % confidence interval = 0.757-0.763). Model performance improved minimally at each subsequent assessment moment, reaching an AUC of 0.798 (0.794-0.801) for parsimonious model 4. The strongest predictors included prior youth care, parental educational level, maternal psychiatric medication prescription, and maternal job status. When classifying 10 % of the cohort with the highest risk of any youth care as positive, the negative predictive value was high (>= 0.972) and the positive predictive value was low (<= 0.164). Conclusion: If the consequences of false positive tests can be mitigated, then screening may offer relief to families before involvement of the over-encumbered youth care services.Show less

Leiden University Scholarly Publications

View statistics

Documents

In Collections

Predicting youth care between 0 to 4 years of age: a 2015–2019 Dutch population register data study