DAIsy

The burden of mental diseases, including major depressive disorder and eating disorders, is huge and continues to grow, with far-reaching social and economic consequences. Current diagnosis and treatment are insufficient and time-consuming, and could profit from integrated AI-driven solutions. In eating disorder care, much of the cost and workload is concentrated in clinician time, particularly in documenting complex intake assessments. These time-intensive processes can limit both diagnostic depth and continuity of care.

Within the ITEA project DAIsy, GGZ Oost-Brabant collaborated with multiple partners to explore how artificial intelligence (AI) can support different stages of the patient journey. The project focused on reducing administrative burden, improving clinical insight and supporting long-term treatment outcomes through a combination of AI applications, including automated documentation, interpretable prediction models, patient engagement tools and continuous monitoring.

One of DAIsy’s eating disorder use cases, involving Dutch project partners GGZ Oost-Brabant, Semlab, and MEDrecord, addressed clinical documentation, using large language models (LLMs) to support intake reporting through AI-assisted summarisation and structured diagnostic templates. Clinicians from GGZ Oost-Brabant’s Centre for Eating Disorders tested two speech-to-text applications, developed by Semlab and MEDrecord (HealthTalk), in routine care. An on-premise dedicated server from Semlab handled real-time intakes and processed them directly within the clinicians’ workflow.

In total, 32 healthcare professionals used the AI-driven speech-to-text platform HealthTalk to record 363 clinical conversations, representing more than 229 hours of real-life dialogue. The tool was applied across different consultations such as intake assessments and progress evaluations.

Clinicians reported that AI-generated summaries can support documentation by structuring and recalling key information, while maintaining clinical oversight. Within the pilot, AI-generated summary and eating disorder feedback contributed to a significant reduction of intake time, with the goal to ultimately reach approximately 75% reduction compared to the manual process. At the same time, implementation in daily practice sometimes proved more complex than anticipated. The pilot highlighted the importance of specificity, flexibility and alignment with clinical workflows - insights that directly informed further development of the technology.

For end users, the benefit is clear. By supporting time-consuming documentation tasks, AI has the potential to significantly reduce administrative burden and create more time for direct patient care. In a field where careful listening and clinical judgement are essential, this shift is meaningful. GGZ Oost-Brabant’s contribution to DAIsy demonstrates how AI can move from experimental tools to practical, workflow-integrated solutions, supporting clinicians in delivering high-quality, patient-centered care.

More information:
- https://itea4.org/project/daisy.html

This development of GGZ Oost-Brabant, Semlab and MEDrecord in the ITEA project DAIsy is supported by RVO.