The diagnosis and treatment of fever of unknown origin (FUO) are huge challenges to clinicians. Separating the etiologies of FUO into infectious and non-infectious disease is conducive to clinical physicians not only on making decisions rapidly concerning the prescription of suitable antibiotics but also on further analysis of the final diagnosis. In order to develop and validate a diagnostic tool to efficiently distinguish the etiologies of adult FUO patients as infectious or non-infectious disease, FUO patients from the departments of infectious disease and internal medicine in three Chinese tertiary hospitals were enrolled retrospectively and prospectively. By using polynomial logistic regression analysis, the diagnostic formula and the associated scoring system were developed. The variables included in this diagnostic formula were from clinical evaluations and common laboratory examinations. The proposed tool could discriminate infectious and non-infectious causes of FUO with an area under receiver operating characteristic curve (AUC) of 0.83, sensitivity of 0.80 and specificity of 0.75. This diagnosis tool could predict the infectious and non-infectious causes of FUO in the validation cohort with an AUC of 0.79, sensitivity of 0.79 and specificity of 0.70. The results suggested that this diagnostic tool could be a reliable tool to discriminate between infectious and non-infectious causes of FUO.