SYNTHETIC NARRATIVES IN HEALTHCARE: ADDRESSING BIAS AND REDEFINING VALIDITY IN QUALITATIVE RESEARCH

The integration of artificial intelligence into healthcare research has accelerated the use of synthetic data, particularly in qualitative contexts involving sensitive or ethically restricted information. This study investigates the emergence of synthetic narratives, especially those generated by large language models (LLMs), as tools to augment or replace traditional qualitative data sources. The research aims to assess how synthetic data is applied in healthcare qualitative studies, uncover methodological biases, and highlight innovative practices that enhance validity. A qualitative systematic literature review was conducted, synthesizing peer-reviewed studies across health sciences, AI, and digital ethics from the past five years. Thematic content analysis was applied to identify patterns in data generation techniques, ethical concerns, and validation strategies. Findings show that synthetic narratives are increasingly used to simulate clinical interactions, preserve anonymity, and address data scarcity in ethically sensitive settings. However, concerns about contextual fidelity, algorithmic bias, and representational limitations were prevalent. Many studies lacked transparent reporting of model architectures or validation protocols. The review concludes that synthetic data presents opportunities and challenges for qualitative healthcare research. It requires new standards of interpretive rigor, interdisciplinary engagement, and ethical oversight. Future research should prioritize co-creative model development, robust validation frameworks, and critical examination of how machine-generated narratives influence qualitative interpretation