This study aimed to examine the factorial structure of the Jefferson Scale of Empathy-Medical student version (JSE-S) and assess items' discriminatory ability at higher and lower empathy levels in medical student populations from different countries and languages. JSE-S datasets were retrieved from previous studies of 4113 first- and/or second-year medical students from Switzerland, Portugal, Italy, Brazil, France, Spain, New Zealand, UK, Ireland, Mexico, and Peru. Parallel principal component analyses and item response theory were applied. Excluding item 18, the Compassionate Care subscale emerged in five samples (Switzerland, Portugal, Italy, France and UK/Ireland), Perspective Taking in three (Switzerland, Portugal and UK/Ireland) and Walking in Patient's Shoes in five (Switzerland, Portugal, Spain, UK/Ireland and Mexico/Peru). All subscales had items providing overlapping information. The JSE-S subscales had reliable high test performance at low empathy levels, while when the JSE-S increased, the standard error also increased. There was no consistent support across countries for the JSE-S three-factorial structure. Items provided redundant information and discriminatory power decreased when aiming to differentiate students at high empathy levels. A JSE-S revision to improve cross-cultural comparability, results' generalizability and validity should focus on suppressing or reformulating items that cannot differentiate medical students' empathy.