[The Structured Information Collection as a data source for AI applications: a qualitative study of data quality].

The Structured Information Collection as a data source for AI applications: a qualitative study of data quality Abstract: Background: Artificial intelligence (AI) to support nursing care planning is gaining attention. However, the effectiveness and reliability of AI-based applications depend on the quality of training data. Objective: This study examines to what extent nursing needs documented within the Structured Information Collection (SIS) can serve as training data for AI-driven decision support systems in care planning. Methods: A qualitative process analysis was conducted. Data were collected through eleven semi-structured interviews on documentation practices in two long-term care facilities. The interviews were analyzed using qualitative content analysis and the METRIC framework was applied to assess data quality. Results: Limitations in data quality were identified across all five clusters of the METRIC framework, especially considering the dimensions of completeness, accuracy, and precision. Conclusions: The data collected within the SIS were found to be only partially suitable for training AI applications for care planning. Improvements in data quality could be achieved through adjustments in documentation practices and an expansion of the data pool. Furthermore, approaches should also be considered to enhance the performance of AI systems. Training nursing professionals is essential to foster the responsible use of AI applications in nursing practice. Limitations: The findings are based on eleven qualitative interviews conducted in two care facilities and therefore do not reflect the full range of documentation culture across care settings.