Developing a System Architecture for Integrating Patient Data Analytics into PICU Capacity Management

Abstract

Pediatric Intensive Care Unit (PICU) capacity management plays a vital role in healthcare systems, especially during dynamic and high-demand situations. However, current planning approaches rely on static models and retrospective hospital data, which are limited in their ability to respond to sudden changes in demand. These models largely depend on internal patient flow and length of stay predictions, without considering external influences or real-time developments, leading to delayed responses and inefficient resource allocation. This research explores how new approaches to patient data analytics can improve PICU utilization planning and what kind of system architecture could enable this shift. A qualitative approach was employed, involving interviews with various PICU capacity management work-related professionals. The research process was structured using the Double Diamond model to guide the progression from problem exploration to solution development. Thematic analysis was used to identify key themes related to data use, legal and technical constraints, and opportunities for integrating new data sources such as wearable devices, remote monitoring systems, and data from other departments. This research was conducted in the context of PICU capacity management in the Netherlands. The findings show that while relevant data is being generated, it remains fragmented and unintegrated in current systems, limiting forecasting capabilities. Legal frameworks such as GDPR and the EU AI Act further complicate data exchange but also provide a basis for designing compliant systems. The study proposes a new system architecture that integrates diverse data sources while taking into account legal constraints, to enhance forecasting accuracy and operational decision-making. This research contributes to PICU capacity management by bringing together data analytics and automation while addressing legal constraints. It offers a scientific contribution by exploring how predictive modeling from various data sources can be integrated within a compliant system architecture and by addressing the challenge of fragmented data for effective PICU planning.