Stratifying disease severity in patients with non-alcoholic steatohepatitis (NASH) is essential for appropriate treatment and long-term management. Liver biopsy is the reference standard for fibrosis severity in NASH, but less invasive methods are used, eg, Fibrosis-4 Index (FIB-4) and vibration-controlled transient elastography (VCTE), for which reference thresholds for no/early fibrosis and advanced fibrosis are available. We compared subjective physician assessment of NASH fibrosis versus reference thresholds to understand classification in a real-world setting. Data were drawn from Adelphi Real World NASH Disease Specific ProgrammeTM conducted in France, Germany, Italy, Spain and UK in 2018. Physicians (diabetologists, gastroenterologists, hepatologists) completed questionnaires for five consecutive NASH patients presenting for routine care. Physician-stated fibrosis score (PSFS) based on available information was compared with clinically defined reference fibrosis stage (CRFS) determined retrospectively using VCTE and FIB-4 data and eight reference thresholds. One thousand two hundred and eleven patients had VCTE (n = 1115) and/or FIB-4 (n = 524). Depending on thresholds, physicians underestimated severity in 16-33% (FIB-4) and 27-50% of patients (VCTE). Using VCTE ≥12.2, diabetologists, gastroenterologists and hepatologists underestimated disease severity in 35%, 32%, and 27% of patients, respectively, and overestimated fibrosis in 3%, 4%, and 9%, respectively (p = 0.0083 across specialties). Hepatologists and gastroenterologists had higher liver biopsy rates than diabetologists (52%, 56%, 47%, respectively). PSFS did not consistently align with CRFS in this NASH real-world setting. Underestimation was more common than overestimation, potentially leading to undertreatment of patients with advanced fibrosis. More guidance on interpreting test results when classifying fibrosis is needed, thereby improving management of NASH.