Implementing Large Language Models in Healthcare Reporting: a Framework for Small Health Technology Companies

Abstract

Large Language Models (LLMs) have showed significant potential in automating and improving healthcare reporting, yet their real-world adoption remains limited due to the lack of practical implementation guidance. This thesis addresses this gap by developing a structured framework for implementing LLMs in healthcare reporting systems, with a focus on small-sized health technology companies. These companies face distinct constraints in resources, expertise, and regulatory capacity that make the adoption of LLMs particularly challenging. The study is guided by two research questions. The first one focuses on what are the key technical and regulatory considerations that are required for deployment, and the second one examines how the developed framework can be utilized to design and implement a real-world LLM-based solution for healthcare reporting. The methodology consists of a qualitative literature analysis for developing the framework and an empirical case study conducted in collaboration with Nuanic Oy. The results indicate that successful LLM implementation is a multidimensional process. It requires a combination of model selection decisions, system design strategies, and alignment with regulatory requirements including data protection, transparency, and system safety. The case study demonstrates that the framework provides structured guidance for real-world deployment and supports informed decision-making in system design. This thesis provides a comprehensive and adaptable framework for the effective, transparent, and compliant integration of LLMs into healthcare reporting systems. The framework is especially valuable for small-sized health technology companies navigating the technical and regulatory complexities of deploying LLM solutions. Collectively, the study advances understanding of the real-world viability of LLMs in healthcare reporting.