This article in the American Journal of Medicine argues for a focused clinical approach rather than a shotgun approach to the diagnosis of fever of undetermined origin. Due to the complexities of the condition there are difficulties with such an approach. There is certainly no algorithm to apply. Nevertheless certain clinical clues can help guide the diagnostic approach.
Despite the use of outmoded terminology (eg hypernephroma for renal cell carcinoma and periarteritis nodosa for polyarteritis nodosa) and questionable assumptions (eg that fever curve patterns reliably point to the correct diagnosis) the review, which is available as free full text, makes some helpful points:
First, confirm that your patient's illness meets the criteria for FUO
The coding calculators in your EMR may try and force you into this designation prematurely. Do not succumb!
Attempt to classify the patient's FUO according to the two bases for classification
These are classical (infectious, neoplastic, rheumatic/inflammatory) and host based (immunocompetent, HIV, immunosuppressive therapy, travel history).
Look for historical features that suggest or eliminate certain categories
Rigors, for example, eliminate the rheumatologic category.
Look for organ involvement clues
SLE, for example, tends to spare the liver.
Laboratory clues may help
From the paper:
Diagnostic specificity of nonspecific laboratory abnormalities is increased when considered together. Degree of test abnormality itself limits diagnostic possibilities, for example, an elevated erythrocyte sedimentation rate is very sensitive/not specific, but a highly elevated erythrocyte sedimentation rate (greater than100 mm/h) narrows diagnostic possibilities to very few entities. Similarly, 6% atypical lymphocytes (drug fever, toxoplasmosis) have a different differential than 36% atypical lymphocytes (Epstein-Barr virus, cytomegalovirus). Nonspecific findings may be exclusionary clues, for example, eosinophilia argues strongly against typhoid/enteric fever...
In a fever of unknown origin, an isolated alkaline phosphatase elevation suggests lymphoma.21, 23, 26 Serum protein electrophoresis also may provide diagnostic clues, for example, elevated α1/α2 globulin elevations (lymphoma, systemic lupus erythematosus); monoclonal gammopathy (multiple myeloma, hyper-IgD syndrome, multicentric Castleman's disease); and polyclonal gammopathy (human immunodeficiency virus, cytomegalovirus, cirrhosis, sarcoidosis, malaria).