Adapting Coding and QA Practices for Using AI-Assisted Coding Tools in the European Regulated Healthcare Environments
Goszczynski, Blazej (2026)
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-202603033640
https://urn.fi/URN:NBN:fi:amk-202603033640
Tiivistelmä
The increasing use of AI-assisted coding tools is reshaping software development, including European regulated healthcare environments. These environments are shaped by regulations and quality requirements, and put great value on safety, reliability, transparency, and traceability of software systems. AI-assisted coding tools offer potential benefits in productivity or creativity; however, existing regulations and standards provide high-level requirements, however, they offer limited practical guidelines on how such tools should be used in daily coding and quality assurance practices. This can cause uncertainty for developers and organizations aiming to balance productivity and regulatory compliance.
The objective of this thesis was to examine how developers can adapt their coding and QA practices when using AI-assisted coding tools in European regulated healthcare environments. The study focused on identifying benefits and risks associated with AI-assisted coding, analyzing how European regulations influence modern healthcare software development, determining necessary adaptations to maintain safety and compliance, and comparing AI-assisted coding with conventional coding in terms of coequalities coverage, and development efficiency. The theoretical framework of this study was based on European healthcare regulations and standards, including MDR/IVDR, IEC 62304, ISO 13485, ISO 14971, GDPR, and the emerging EU AI Act.
The study was implemented using a mix-method approach. Empirical data was collected through six semi-structured interviews with professionals related to software development and quality assurance and through a comparative coding experiment including a prototype cardiovascular report generation system implemented both manually and with AI assistance. The findings indicate that AI-assisted coding improved development in terms of efficiency, especially for routine tasks, however, it introduces risks in reliability, explainability, data security, and increased QA workload. Test coverage remained similar between AI-assisted and manual implementation, with differences in defects, code duplication, and structural code quality.
Based on these findings, a practical developer-oriented framework was created supporting responsible use of AI-assisted coding tools in European regulated healthcare environments. The framework complements existing regulations and standards and provides practical guidelines in data governance, human oversight, review process, testing, traceability, and developer training.
The objective of this thesis was to examine how developers can adapt their coding and QA practices when using AI-assisted coding tools in European regulated healthcare environments. The study focused on identifying benefits and risks associated with AI-assisted coding, analyzing how European regulations influence modern healthcare software development, determining necessary adaptations to maintain safety and compliance, and comparing AI-assisted coding with conventional coding in terms of coequalities coverage, and development efficiency. The theoretical framework of this study was based on European healthcare regulations and standards, including MDR/IVDR, IEC 62304, ISO 13485, ISO 14971, GDPR, and the emerging EU AI Act.
The study was implemented using a mix-method approach. Empirical data was collected through six semi-structured interviews with professionals related to software development and quality assurance and through a comparative coding experiment including a prototype cardiovascular report generation system implemented both manually and with AI assistance. The findings indicate that AI-assisted coding improved development in terms of efficiency, especially for routine tasks, however, it introduces risks in reliability, explainability, data security, and increased QA workload. Test coverage remained similar between AI-assisted and manual implementation, with differences in defects, code duplication, and structural code quality.
Based on these findings, a practical developer-oriented framework was created supporting responsible use of AI-assisted coding tools in European regulated healthcare environments. The framework complements existing regulations and standards and provides practical guidelines in data governance, human oversight, review process, testing, traceability, and developer training.