pipette
ENEnglish

Label-free, low-cost AI ranking for systematic review screening

T. Jo, D. Lee, L. Lee

PreprintUso en el mundo real

En palabras de los autores

Title and abstract screening is among the most labor-intensive steps of systematic reviews. Jev, a low-cost AI decision model, scores records against eligibility criteria without screening labels. We scored 169,982 records from 128 reviews and, after tuning on development reviews, evaluated 23 held-out reviews once and 31 Cochrane reviews post hoc. Jev outranked the best free ranker in all 23 held-out reviews (mean AUC 0.966 vs 0.831) and came within 0.011 AUC of Claude Opus 5.5 on 1999 records at about 1/700 of its list-price cost (US$0.02/1000 records). Combined with active learning, it raised work saved at 95% recall from 0.659 to 0.806. A fixed score threshold fell below 95% recall in 3 of 28 Cochrane reviews, whereas Jev ranking with statistical stopping (post hoc) reached 95% recall in all 51 reviews, reading 86-88% of records. Label-free ranking with Jev was accurate and inexpensive, but stopping safely required screening decisions.

Resultado principalLimitación que admiten los autores

Apareció: martes, 13 de octubre. medRxiv. Preprint, todavía sin revisión por pares.

DOI: 10.64898/2026.10.09.26365173