We aimed to develop a simple prediction score to identify fluconazole non-susceptible (Flu-NS) candidaemia using simple clinical criteria. A derivation cohort was extracted from the CANDIPOP study, a prospective, multicentre, population-based surveillance programme on candidaemia conducted in 29 hospitals in Spain from April 2010 to May 2011. The score was validated with an external, multicentre cohort of adults with candidaemia in six tertiary hospitals in three countries. The prediction score was based on three variables selected by a logistic regression model together with the severity of disease. In total, 617 and 297 cases of candidaemia were included in the derivation and validation cohorts, respectively; of these, 134 (21.7%) and 57 (19.2%) were caused by Flu-NS strains. Factors independently associated with Flu-NS were transplant recipient status (adjusted odds ratio (AOR) 2.13; 95% CI 1.01-4.55; p 0.047), hospitalization in a unit with a high prevalence (≥ 15%) of Flu-NS strains (7.53; 4.68-12.10; p < 0.001), and previous azole therapy for at least 3 days (2.04; 1.16-3.62; p 0.014). The area under the receiver operating characteristics curve (AUC) was 0.76 (0.72-0.81), and using 2 points as the Flu-NS prediction score cut-off gave a sensitivity of 82.1%, a specificity of 65.6%, and a negative predictive value of 93%. The AUC in the validation cohort was 0.72 (95% CI 0.65-0.79). Hence, the Flu-NS prediction score helped to exclude Flu-NS Candida strains. This could improve the selection of empirical treatments for candidaemia in the future.