Zheng Chang

Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder marked by significant comorbidity and serious outcomes. The use of ADHD medication has surged in Sweden and many other countries, and it prompts questions about its effectiveness and safety amid broadened diagnostic criteria. Despite the general efficacy of these medications, patient responses vary, and high discontinuation rates prevail, highlighting an urgent need for more personalized treatment strategies. This proposal aims to enhance our understanding of the individualized treatment effects of ADHD medication related to patient characteristics and to develop a decision-aid tool for personalized treatment decisions. It focuses on key clinical decisions: whether to initiate medication and the choice of first- and second-line treatments. Utilizing extensive healthcare data from Sweden, Denmark, and the US, we plan to employ cutting-edge study designs and statistical methods. The decision-aid tool, created with end-user collaboration, will be evaluated in a pilot study. The project will be carried out by an interdisciplinary team of researchers from Karolinska Institutet, clinicians from region Stockholm and Skåne, and international experts in the field. This innovative approach aims to improve patient outcomes, reduce side effects, and tackle the challenges associated with the rising use of ADHD medications, marking a substantial step towards personalized ADHD treatment.

The rising prevalence of youth mental health disorders has outpaced the capacity of existing services. To address this, we propose register-based research innovations to support prevention at the societal level and improve individualized treatment in primary and specialist care. We integrate national Registers with novel data from the general population, primary, and specialist care, combined with recent advances in large language models (LLMs) and causal machine learning. Our interdisciplinary team of epidemiologists, health care providers and AI computer scientists aim to advance the use of LLMs in register-based research, by developing a language model architecture for unstructured medical records and prescription data, which will be implemented in three specific aims:1) Identify emerging risk factors to guide prevention using causal inference and longitudinal survey data linked with national registers 2) Strengthen youth mental health services by applying new data and methods to improve patient stratification in primary care3) Optimize treatment in specialist youth mental health services, by better understanding individualized treatment effects based on patient characteristics This research will improve youth mental health, develop innovative methods for register-based research, including a large language model to handle unstructured data, and train the next generation of register-based researchers to address healthcare-relevant questions using cutting-edge AI methods.