The classic definition of FUO from the early 1960s was: fever greater than 38 degrees Celsius on several occasions over a 3 week period of time with week worth of hospital investigations. Due to advances in our diagnostic capabilities, the definition of FUO has been changed to a 3 week duration of occasional fever over 38 degrees, but the 1 week of hospital investigations has been replaced with a "first wave" of investigations (history, physical exam, basic blood work, ANA, RF, HIV, CMV, urinalysis, blood and urine cultures, chest imaging, and possibly viral hepatitis serology). I should note that there are a few definitions out there - some say 2 weeks, others say 3...you get the picture.
What is the etiology of FUO? Roughly half of patients do not end up with a diagnosis. About 20% have an underlying inflammatory condition (eg. vasculitis, SLE), 15% have an infectious diagnosis, 7% have an underlying malignancy, and the rest are a hodge-podge (yes, hodge-podge) of miscellaneous conditions - this is from a prospective Dutch study that did not involve immunocompromised patients.
So how do we investigate? A proposed algorithm based on existing evidence was created by local talent and published in Archives of Internal Medicine.