Large language models (LLMs) could help health researchers analyse large qualitative datasets more efficiently, but they should be treated as a secondary analytic lens rather than a substitute for ...