Towards automated risk prediction of persistent pain: Exploring psychosocial factors from electronic health records after breast cancer surgery.

Records came from one Finnish university hospital, so documentation style and language may not apply elsewhere. Only free-text narratives were analysed, so information recorded only in structured fields was not included. The records were written by healthcare professionals and may not reflect patients' own experiences or thoughts. There was no clear systematic documentation policy, so some factors may have been discussed but not recorded. Anxiety, depression and sleep were not measured systematically in all patients, so the recorded expressions do not show how common these symptoms were. No patient pain outcome was analysed, so the study cannot show which documented factors predicted persistent pain. Only 101 records were analysed after random selection from a larger set, and saturation was used rather than a powered sample.

There was no funding for this work, and the authors declared no potential conflicts of interest.

Do not read the category percentages as evidence that these factors caused or predicted persistent pain. This study reviewed existing records and did not test a prediction model or patient outcomes.