This PhD thesis focused on the question of the extent to which machine learning can be applied to healthcare data, with the notion that it could ultimately improve the Dutch mental healthcare system. We answered this question by carrying out several studies in which we endeavoured to predict the characteristics of treatments in the mental healthcare system. Machine learning is a technique that uses patterns from past data to predict future elements or events. We primarily focused on predicting levels of healthcare utilisation in relation to treatment outcomes. The findings from the seven studies in this PhD thesis provided relevant insights both for the different healthcare-specific themes and in more methodological areas. One of the key lessons from this exploratory PhD thesis is that there is no one-size-fits-all solution. The development of machine learning algorithms invariably involves the consideration of different options as well as the critical examination of what works, when, and why. Algorithms are not 100% accurate, which means that a nuanced interpretation is also required during their application. If all of these steps are carried out correctly, machine learning can help to support effective decision-making and consequently contribute to creating a healthcare system that offers accessible, affordable and high-quality care.