The first batch came easy.
Recently-published abstracts. Readers still argued.
what we noticed
field
reader a
reader b
study_type
other
not applicable
study_design
cohort
single-arm trial
consort_flag
exclude ✓
exclude ✓
Readers reached the same final answer. They argued about the fields in between.
what we cut
- study_design · 14 categories→removed
- study_type · 11 categories→manuscript_type · 4
We didn’t need hard papers to learn the codebook was too big. The easy batch showed us what was redundant.
@karlrohe