Verontrustingen - Enabling accurate predictions to detect day-care services which require further inspection

In 2014, the Flemish Agency for Child and Family (Kind en Gezin) developed an AI system which enables more accurate predictions to detect day-care services which require further inspection. These inspections enable the agencies to keep the quality of the day-care services high and to improve the wellbeing of children. The Child and Family agency does not carry out the inspections itself but works together with the Regional Health Care Inspection Unit of the Department of Welfare, Public Health and Family. However, there is limited capacity available to conduct all the inspections. For a long period of time, there has been an interest to figure out how to optimise the inspection capacity. The use of data had been considered as way to enhance existing inspection practices and optimize the scarce amount of inspectors. The predictive system developed uses a supervised machine learning method (logistic regression and XGBoost) to analyse various internal and external data from the Health Care Inspection Unit. Combining the recommendations of the predictive system with existing staff experience and expertise enables more targeted and data-driven interventions.