Research volume against patients' priorities
Not ready: a calibration check is still pending
How much published research goes to each question that women with PCOS ranked as most important?
The method
- The questions: the 27 ranked research questions from the James Lind Alliance PCOS priority-setting partnership (2025), word for word. experiments/mismatch_map/pmos/PREREG.md:11-12
- The papers: MEDLINE, title or abstract, 1 January 2016 to 24 October 2026, under a condition search Dan approved before anything was counted. experiments/mismatch_map/pmos/PREREG.md:13-14 palace/threads/mismatch-map-pmos.md:35
- The count: code counts papers per question with two independently drafted searches. Readers then judge 25 sampled papers per search, to estimate how many of the hits truly address the question. experiments/mismatch_map/pmos/PREREG.md:15-19
- The plan was written and committed before any count was run. experiments/mismatch_map/pmos/PREREG.md:3-5
Why it is held back
A calibration check asks whether the method gets a known answer right on this condition. That check has been built for PCOS and has not been scored. Showing results before it is scored would invite people to rely on a number we have not tested.
Other limits are already known. The priorities come from a UK partnership. Keyword counts overcount questions whose words are common in other papers. palace/threads/mismatch-map-pmos.md:31 On the test condition, the counts ran low on 3 of 10 questions, so every count would be labelled as likely an undercount. experiments/mismatch_map/pmos/PREREG.md:38-40 And how many papers address a question says nothing about how good they are.